HIGHLY CONFIDENTIAL

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL (STA)

Dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

Appraisal Committee Meeting – 2 November 2021 1[st] Committee meeting

The following documents are made available to the Company:

The final scope and final stakeholder list are available on the NICE website.

Pre-technical engagement documents

1. Company submission summary from GlaxoSmithKline

2. Clarification questions and company responses 2a. Clarification questions

• 2b. Company responses

3. Patient group, professional group and NHS organisation submissions from:

• a. NCRI-ACP-RCP-RCR

4. Evidence Review Group report prepared by Warwick Evidence

5. Evidence Review Group report – factual accuracy check

Post-technical engagement documents

6. Technical engagement response from company

7. Technical engagement responses and statements from experts:

• a. Dr Susana Banerjee, Consultant Medical Oncologist – clinical expert, nominated by GSK (company)

• b. Andrew Clamp, Consultant and Honorary Senior Lecturer in Medical Oncology – clinical expert, nominated by GSK (company)

• c. Hilary Maxwell, Gynae-Oncology Clinical Nurse Specialist – patient expert, nominated by Go Girls

8. Evidence Review Group critique of company response to technical engagement prepared by Warwick Evidence

9. Appraisal Committee Meeting presentation slides (uploaded separately)

Please note that the full submission, appendices to the company’s submission and

HIGHLY CONFIDENTIAL

company model will be available as a separate file on NICE Docs for information only.

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Dostarlimab for previously treated advanced and recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

Document B

Company evidence submission

17 May 2021

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 1 of 222

Instructions for companies

This is the template for submission of evidence to the National Institute for Health and Care Excellence (NICE) as part of the single technology appraisal (STA) process. Please note that the information requirements for submissions are summarised in this template; full details of the requirements for pharmaceuticals and devices are in the user guide.

This submission must not be longer than 150 pages, excluding appendices and the pages covered by this template. If it is too long it will not be accepted.

Companies making evidence submissions to NICE should also refer to the NICE guide to the methods of technology appraisal and the NICE guide to the processes of technology appraisal.

In this template any information that should be provided in an appendix is listed in a box.

Highlighting in the template (excluding the contents list)

Square brackets and **** highlighting are used in this template to indicate text that should be replaced with your own text or deleted. These are set up as form fields, so to replace the prompt text in ******************* with your own text, click anywhere within the highlighted text and type. Your text will overwrite the highlighted section.

To delete grey highlighted text, click anywhere within the text and press DELETE.

Grey highlighted text in the footer does not work as an automatic form field, but serves the same purpose – as prompt text to show where you need to fill in relevant details. Replace the text highlighted in ****** in the header and footer with appropriate text. (To change the header and footer, double click over the header or footer text. Double click back in the main body text when you have finished.)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 2 of 222

Contents

Instructions for companies ...................................................................................................................... 2 Contents .................................................................................................................................................. 3 List of tables ............................................................................................................................................ 5 List of figures ........................................................................................................................................... 7 Abbreviations ........................................................................................................................................ 10 B.1 Decision problem, description of the technology and clinical care pathway .................................. 12 B.1.1 Decision problem ..................................................................................................................... 13 B.1.2 Description of the technology being appraised ....................................................................... 17 B.1.3 Health condition and position of the technology in the treatment pathway ............................. 19 B.1.3.1 Disease overview ............................................................................................................... 19 B.1.3.2 Disease classification, progression and recurrence ........................................................... 20 B.1.3.3 Epidemiology ...................................................................................................................... 24 B.1.3.4 Clinical care pathway.......................................................................................................... 25 B.1.3.4.1 Initial management of EC .............................................................................................. 26 B.1.3.4.2 Treatment for recurrent or advanced EC ...................................................................... 27 B.1.3.5 Limitations of current treatment and unmet need ............................................................... 29 B.1.3.6 Dostarlimab ........................................................................................................................ 30 B.1.3.6.1 Anticipated positioning of dostarlimab in UK clinical practice ....................................... 31 B.1.4 Equality considerations ........................................................................................................... 32 B.2 Clinical effectiveness ...................................................................................................................... 33 B.2.1 Identification and selection of relevant studies........................................................................ 35 B.2.2 List of relevant clinical effectiveness evidence ........................................................................ 35 B.2.2.1.1 Dostarlimab (GARNET) ................................................................................................. 35 B.2.2.1.2 Comparators (NICE final scope) ................................................................................... 35 B.2.2.1.3 Current clinical management (UK RWE study) ............................................................. 36 B.2.3 Summary of methodology of GARNET and the UK RWE study ............................................. 39 B.2.3.1 Dostarlimab (GARNET) ...................................................................................................... 39 B.2.3.1.1 Summary of trial methodology ....................................................................................... 39 B.2.3.1.2 Baseline characteristics ................................................................................................. 43 B.2.3.1.3 Statistical analysis and definitions of study groups ....................................................... 45 B.2.3.1.4 Quality assessment ....................................................................................................... 46 B.2.3.2 Current clinical management (UK RWE study) .................................................................. 48 B.2.3.2.1 Summary of study methodology .................................................................................... 48 B.2.3.2.2 Baseline characteristics ................................................................................................. 52 B.2.4 Clinical effectiveness results from GARNET and the UK RWE study..................................... 55 B.2.4.1 Baseline characteristics (GARNET versus UK RWE study) .............................................. 55 B.2.4.2 Objective response rate (ORR) .......................................................................................... 57 B.2.4.2.1 Dostarlimab (GARNET) ................................................................................................. 57 B.2.4.2.2 Dostarlimab versus current clinical management ......................................................... 58 B.2.4.3 Duration of response (DOR) ............................................................................................... 59 B.2.4.3.1 Dostarlimab (GARNET) ................................................................................................. 59 B.2.4.3.2 Dostarlimab versus current clinical management ......................................................... 60 B.2.4.4 Change in target lesion size ............................................................................................... 60 B.2.4.4.1 Dostarlimab (GARNET) ................................................................................................. 60 B.2.4.4.2 Dostarlimab versus current clinical management ......................................................... 61 B.2.4.5 Progression-free survival (PFS) ......................................................................................... 61 B.2.4.5.1 Dostarlimab (GARNET) ................................................................................................. 61 B.2.4.5.2 Dostarlimab versus current clinical management ......................................................... 63 B.2.4.6 Overall survival (OS) .......................................................................................................... 65 B.2.4.6.1 Dostarlimab (GARNET) ................................................................................................. 65 B.2.4.6.2 Dostarlimab versus current clinical management ......................................................... 67 B.2.4.7 Immune-related endpoints (dostarlimab only) .................................................................... 68 B.2.4.8 Health-related quality of life (dostarlimab only) .................................................................. 70 B.2.5 Subgroup analysis (dostarlimab only) ..................................................................................... 73 B.2.6 Meta-analysis .......................................................................................................................... 75

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 3 of 222

B.2.7 Indirect and mixed treatment comparisons ............................................................................. 75 B.2.7.1 Dostarlimab versus current clinical management (UK RWE MAIC versus GARNET) ....... 75 B.2.7.2 Supportive comparative evidence for dostarlimab versus individual chemotherapy regimens ......................................................................................................................................... 85 B.2.7.2.1 ITC between dostarlimab (GARNET) and doxorubicin (ZoptEC study) ........................ 85 B.2.7.2.2 MAICs versus carboplatin plus paclitaxel, paclitaxel monotherapy and doxorubicin monotherapy ................................................................................................................................. 90 B.2.7.2.3 Strengths, limitations and conclusions of the supportive comparative efficacy evidence ...................................................................................................................................................... 98 B.2.7.3 Comparative efficacy evidence versus hormone therapy ................................................ 101 B.2.8 Adverse reactions .................................................................................................................. 102 B.2.8.1 Treatment exposure ......................................................................................................... 103 B.2.8.2 TEAEs .............................................................................................................................. 104 B.2.8.3 Treatment-related TEAEs ................................................................................................. 107 B.2.8.4 Serious TEAEs ................................................................................................................. 108 B.2.8.5 Treatment-related serious TEAEs .................................................................................... 109 B.2.8.6 Deaths .............................................................................................................................. 110 B.2.8.7 Comparative safety ........................................................................................................... 110 B.2.9 Ongoing studies .................................................................................................................... 112 B.2.10 Innovation ............................................................................................................................ 113 B.2.11 Interpretation of clinical effectiveness and safety evidence ................................................ 114 B.2.11.1 Principal findings from the clinical evidence base .......................................................... 114 B.2.11.2 Strengths of the clinical evidence base .......................................................................... 117 B.2.11.3 Limitations of the clinical evidence base ........................................................................ 119 B.2.11.4 Conclusion ...................................................................................................................... 121 B.2.11.5 End-of-life criteria ........................................................................................................... 121 B.3 Cost-effectiveness ........................................................................................................................ 125 B.3.1 Published cost-effectiveness studies .................................................................................... 126 B.3.2 Economic analysis ................................................................................................................. 128 B.3.2.1 Patient population ............................................................................................................. 128 B.3.2.2 Model structure .............................................................................................................. 128 B.3.2.3 Intervention technology and comparators ........................................................................ 131 B.3.2.4 Subsequent therapies .................................................................................................... 134 B.3.3 Clinical parameters and variables ......................................................................................... 135 B.3.3.1 Patient characteristics ...................................................................................................... 135 B.3.3.2 Clinical outcomes ............................................................................................................. 136 B.3.3.3 Survival inputs and assumptions ...................................................................................... 136 B.3.3.4 Treatment waning ............................................................................................................. 137 B.3.3.5 Progression-free survival .................................................................................................. 138 B.3.3.6 Overall survival ................................................................................................................. 145 B.3.3.7 Time on treatment ............................................................................................................ 152 B.3.3.8 Adverse events ................................................................................................................. 157 B.3.4 Measurement and valuation of health effects ....................................................................... 159 B.3.4.1 Health-related quality-of-life data from clinical trials ......................................................... 159 B.3.4.2 Mapping ............................................................................................................................ 160 B.3.4.3 Health-related quality-of-life studies ................................................................................. 160 B.3.4.4 Adverse reactions ............................................................................................................. 160 B.3.4.5 Age-adjusted utility values ................................................................................................ 161 B.3.4.6 Health-related quality-of-life data used in the cost-effectiveness analysis ....................... 162 B.3.5 Cost and healthcare resource use identification, measurement and valuation .................... 162 B.3.5.1 Intervention and comparator drug acquisition costs and resource use ............................ 163 B.3.5.2 Drug administration costs ................................................................................................. 167 B.3.5.3 Follow-up and monitoring costs and resource use ........................................................... 167 B.3.5.4 Adverse event costs and resource use ............................................................................ 170 B.3.5.5 Subsequent therapies costs and resource use ................................................................ 171 B.3.5.6 End-of-life costs and resource use ................................................................................... 174 B.3.5.7 Diagnostic testing costs and resource use ....................................................................... 174 B.3.6 Summary of base case analysis inputs and assumptions .................................................... 175

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 4 of 222

B.3.6.1 Summary of base case analysis inputs ............................................................................ 175 B.3.6.2 Assumptions ..................................................................................................................... 179 B.3.7 Base case results .................................................................................................................. 187 B.3.7.1 Base case incremental cost-effectiveness analysis results ............................................. 187 B.3.8 Sensitivity analyses ............................................................................................................... 188 B.3.8.1 Probabilistic sensitivity analysis ....................................................................................... 188 B.3.8.2 Deterministic sensitivity analysis ...................................................................................... 194 B.3.8.3 Scenario analysis ............................................................................................................. 198 B.3.8.4 Summary of sensitivity analyses results........................................................................... 211 B.3.9 Subgroup analysis ................................................................................................................. 211 B.3.10 Validation ............................................................................................................................. 211 B.3.10.1 Validation of cost-effectiveness analysis ........................................................................ 211 B.3.11 Interpretation and conclusions of economic evidence ........................................................ 212 B.4 References ................................................................................................................................... 215 B.5 Appendices ................................................................................................................................... 222

List of tables

Table 1: The decision problem ...................................................................................................................... 13 Table 2: Technology being appraised ........................................................................................................... 17 Table 3: FIGO cancer staging for EC ............................................................................................................ 21 Table 4: Ten most common chemotherapy regimens received by patients with recurrent or advanced EC following disease progression on platinum-based doublet chemotherapy .................................................... 28 Table 5: Summary of the clinical effectiveness evidence presented in this submission ............................... 37 Table 6: Summary of methodology for GARNET trial ................................................................................... 39 Table 7: Baseline demographics and disease characteristics of patients in GARNET ................................. 44 Table 8: Statistical analysis sets in GARNET ............................................................................................... 46 Table 9: Risk of bias assessment of GARNET trial using the CASP risk of bias tool for non-RCTs ............. 46 Table 10: Initial inclusion and exclusion criteria of the UK RWE study to identify patients with EC .............. 49 Table 11: Inclusion and exclusion criteria of the UK RWE study to identify patients with recurrent or advanced EC ................................................................................................................................................. 49 Table 12: Inclusion and exclusion criteria of the UK RWE study to identify patients for a GARNET-like cohort ............................................................................................................................................................ 50 Table 13: Patient demographics and clinical characteristics in the GARNET-like UK RWE study cohort .... 52 Table 14: Most common chemotherapy regimens received by patients in the UK RWE study GARNET-like cohort ............................................................................................................................................................ 54 Table 15: Patient demographics and clinical characteristics in the GARNET-like UK RWE study cohort and ITT population of the GARNET trial .............................................................................................................. 55 Table 16: Summary of the BOR to dostarlimab by RECIST v1.1 in GARNET (efficacy population) (BICR) . 57 Table 17: KM analysis of DOR based on BICR in GARNET (efficacy population; patients with objective response) ...................................................................................................................................................... 60 Table 18: KM analysis of PFS from GARNET (efficacy population and ITT population) (BICR) .................. 62 Table 19: Naïve PFS comparison for patients with recurrent or advanced EC that have progressed on or after platinum-based chemotherapy (UK RWE study versus GARNET) ....................................................... 64 Table 20: KM analysis of OS from GARNET (efficacy population and ITT population) ................................ 66 Table 21: Naïve OS comparison for patients with recurrent or advanced EC that has progressed on or after platinum-based chemotherapy (UK RWE study versus GARNET) ............................................................... 68 Table 22: irPFS per irRECIST KM analysis in GARNET (immune-related efficacy population) (Investigator Assessment) ................................................................................................................................................. 69 Table 23: Scenarios considered in the UK RWE study MAICs versus GARNET ......................................... 78 Table 24: OS for patients in the GARNET ITT population (before and after matching) and the UK RWE GARNET-like cohort ...................................................................................................................................... 79 Table 25: PFS for patients in the GARNET ITT population (before and after matching) and the UK RWE GARNET-like cohort ...................................................................................................................................... 82 Table 26: Results for the safety analysis data set on OS with adjusting stabilised-IPTW ............................ 87

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 5 of 222

Table 27: Summary of OS with adjusting stabilised-IPTW for the main analysis data set ............................ 88 Table 28: Summary of PFS for the main analysis data set ........................................................................... 89 Table 29: PFS hazard ratios derived from the MAICs ................................................................................... 93 Table 30: OS for dostarlimab versus chemotherapy comparators based on MAICs .................................... 96 Table 31: Duration of treatment with dostarlimab (ITT population) ............................................................. 103 Table 32: Overall summary of TEAEs (ITT population) .............................................................................. 104 Table 33: Common TEAEs by system organ class and preferred term (≥10% of patients) (ITT population) .................................................................................................................................................................... 105 Table 34: Grade 3 or greater TEAEs occurring in ≥3 patients (ITT population) .......................................... 105 Table 35: Overall summary of treatment-related TEAEs (ITT population) .................................................. 107 Table 36: Treatment-related TEAEs experienced by ≥5% of patients (ITT population) .............................. 107 Table 37: Treatment-related[a] Grade ≥3 TEAEs occurring in ≥2 patients (ITT population) .......................... 108 Table 38: Serious TEAEs by system organ class and preferred term experienced by >1% of patients (ITT population) .................................................................................................................................................. 108 Table 39: Treatment-related serious TEAEs (ITT population) .................................................................... 109 Table 40: Summary of deaths in GARNET (ITT population) ....................................................................... 110 Table 41: Summary of treatment-related TEAEs in dostarlimab versus chemotherapy studies identified in the clinical SLR (naïve comparisons only) .................................................................................................. 111 Table 42: Treatment-related TEAEs experienced patients in GARNET and ZoptEC (naïve comparison only)[8-10] ........................................................................................................................................................ 112 Table 43: End-of-life criteria ........................................................................................................................ 123 Table 44: OS estimates from trials identified in the clinical SLR and included in ITCs ............................... 123 Table 45: Summary list of published cost-effectiveness studies identified in the economic SLR ............... 127 Table 46: Key features of the economic analysis ........................................................................................ 130 Table 47: Weighting of the individual treatment regimens included as part of current clinical management .................................................................................................................................................................... 132 Table 48: Dosing regimens for dostarlimab and the treatment regimens included as part of current clinical management in the base case cost-effectiveness analysis ........................................................................ 133 Table 49: Patient baseline characteristics of the base case economic analysis ......................................... 136 Table 50: Summary of goodness-of-fit data for dostarlimab PFS (GARNET ITT population); standard parametric and spline models ..................................................................................................................... 140 Table 51: Summary of clinical expert estimates[a] for the percentage of patients who would be progressionfree at various time intervals following treatment with dostarlimab ............................................................. 141 Table 52: Summary of goodness-of-fit data for PFS (based on TTNT) for current clinical management (GARNET-like cohort in the UK RWE study) – standard parametric and flexible models ........................... 143 Table 53: Summary of clinical expert estimates for the percentage of patients who would be progressionfree at various time intervals following treatment with current clinical management ................................... 144 Table 54: Summary of the base case extrapolations for PFS for dostarlimab and current clinical management ............................................................................................................................................... 145 Table 55: Summary of goodness-of-fit data for dostarlimab OS (GARNET ITT population) standard parametric and spline models ..................................................................................................................... 147 Table 56: Summary of clinical expert estimates[a] for the percentage of patients who would be alive at various time intervals following treatment with dostarlimab ........................................................................ 148 Table 57: Summary of goodness-of-fit data for OS for current clinical management (GARNET-like cohort in the UK RWE study) – standard parametric and flexible models ................................................................. 150 Table 58: Summary of clinical expert estimates[a] for the percentage of patients who would be alive following treatment with current clinical management ................................................................................................ 151 Table 59: A summary of the base case extrapolations for OS for dostarlimab and current clinical management ............................................................................................................................................... 152 Table 60: Summary of goodness-of-fit data for dostarlimab ToT (GARNET ITT population) standard parametric and spline models ..................................................................................................................... 154 Table 61: Summary of goodness-of-fit statistics for current clinical management ToT – standard parametric and flexible models ..................................................................................................................................... 156 Table 62: A summary of the assumptions in the base case for ToT for dostarlimab and current clinical management ............................................................................................................................................... 157

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 6 of 222

Table 63: AE rates included within the base case cost-effectiveness analysis ........................................... 158 Table 64: Health state utility values predicted from GARNET .................................................................... 159 Table 65: Adverse event disutilities ............................................................................................................. 161 Table 66: Summary of utility values used in the base case cost-effectiveness analysis ............................ 162 Table 67: Drug acquisition unit costs for dostarlimab and relevant comparators ........................................ 165 Table 68: Drug administration costs ............................................................................................................ 167 Table 69: Routine care and monitoring unit costs ....................................................................................... 168 Table 70: AE costs ...................................................................................................................................... 170 Table 71: Subsequent therapy costs included within the model ................................................................. 172 Table 72: Total subsequent therapy costs in the base case cost-effectiveness analysis ........................... 174 Table 73: End-of-life care costs .................................................................................................................. 174 Table 74: Number of patients requiring IHC testing for dMMR (considered in a scenario analysis only) ... 175 Table 75: IHC testing for determining MMR/MSI status – sensitivity and specificity and costs (considered in a scenario analysis only) ............................................................................................................................. 175 Table 76: Summary of variables applied in the base case economic analysis ........................................... 175 Table 77: List of assumptions for the base case cost-effectiveness analysis ............................................. 179 Table 78: Base case deterministic economic analysis results[a] (dostarlimab list price) ............................... 187 Table 79: Base case deterministic economic analysis results[a] (dostarlimab PAS price) ............................ 187 Table 80: PSA inputs and distributions ....................................................................................................... 188 Table 81: Base case PSA results[a] (dostarlimab list price) .......................................................................... 191 Table 82: Base case PSA results[a] (dostarlimab PAS price) ....................................................................... 192 Table 83: One-way DSA inputs ................................................................................................................... 194 Table 84: Scenarios considered in the UK RWE study MAICs versus GARNET ....................................... 199 Table 85: Scenario analysis results[a] ........................................................................................................... 207 Table 86: Comparison of PFS model results[a] with current clinical management ........................................ 212 Table 87: Comparison of OS model results[a] with current clinical management .......................................... 212

List of figures

Figure 1: Mechanism of action of dostarlimab binding with PD-1 receptor ................................................... 17 Figure 2: The mechanism of action of the DNA MMR system and dMMR/MSI-H ........................................ 22 Figure 3: Estimated dostarlimab-eligible patient population numbers in England ......................................... 25 Figure 4: Treatment pathway for patients with EC in the UK and the anticipated positioning of dostarlimab ...................................................................................................................................................................... 26 Figure 5: UK RWE OS for patients with recurrent or advanced EC who received further chemotherapy following disease-progression on or after platinum-based chemotherapy .................................................... 29 Figure 6: GARNET trial design ...................................................................................................................... 39 Figure 7: Cohorts in the GARNET trial .......................................................................................................... 40 Figure 8: Patients included in the UK RWE study ......................................................................................... 52 Figure 9: DOR (from time of first PR or CR) based on RECIST v1.1 in GARNET (efficacy population) (BICR) ........................................................................................................................................................... 59 Figure 10: Waterfall plot of the maximum percentage change in target lesions compared with baseline measurements based on BICR per RECIST v1.1 in GARNET (efficacy population) .................................... 61 Figure 11: PFS KM curves from GARNET (efficacy population and ITT population) (BICR) ........................ 63 Figure 12: TTNT as a proxy for PFS for patients with recurrent or advanced EC that have progressed on or after platinum-based chemotherapy receiving current clinical management (UK RWE GARNET-like cohort) ...................................................................................................................................................................... 64 Figure 13: OS KM curves from GARNET (efficacy population and ITT population) ..................................... 66 Figure 14: OS for patients with recurrent or advanced EC that has progressed on or after platinum-based chemotherapy receiving current clinical management (UK RWE GARNET-like cohort) ............................... 67 Figure 15: irPFS per irRECIST KM curve in GARNET (immune-related efficacy population) (Investigator Assessment) ................................................................................................................................................. 70 Figure 16: Adjusted Mean Change from Baseline in EQ-VAS (GARNET efficacy population) ..................... 71 Figure 17: Mean change in Global Health Status/QOL from Baseline (GARNET ITT population) ............... 72

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 7 of 222

Figure 18: Pain, fatigue and physical functioning mean change from baseline (GARNET ITT population) .. 72 Figure 19: Symptomatic AE change in response from baseline (GARNET ITT population) ......................... 73 Figure 20: Forest plot of ORR (CR or PR) and 95% CI by subgroup by RECIST v1.1 in GARNET (efficacy population (N=) (BICR) ............................................................................................................................. 74 Figure 21: Histogram of the weights assigned to patients in the GARNET ITT population in the MAIC versus the UK RWE GARNET-like population using the matching variables in Scenario 1 in addition to grade ............................................................................................................................................................. 78 Figure 22: OS KM curves – dostarlimab (unadjusted GARNET ITT population, N=129) versus current clinical management (UK RWE GARNET-like cohort, N=) ....................................................................... 80 Figure 23: OS KM curves – dostarlimab (adjusted GARNET population, Scenario 1, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=) ........................................................... 81 Figure 24: OS KM curves – dostarlimab (adjusted GARNET population, Scenario 2, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=) ........................................................... 81 Figure 25: PFS KM curves – dostarlimab (unadjusted GARNET ITT population, N=) versus current clinical management (UK RWE GARNET-like cohort, N=) ....................................................................... 83 Figure 26: PFS KM curves – dostarlimab (adjusted GARNET population, Scenario 1, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=) ........................................................... 84 Figure 27: PFS KM curves – dostarlimab (adjusted GARNET population, Scenario 2, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=) ........................................................... 84 Figure 28: OS KM curves – dostarlimab (adjusted GARNET main analysis set, N=) versus doxorubicin monotherapy (adjusted ZoptEC main analysis set, N=** following adjustments based on IPTW[8-10] .......... 87 Figure 29: PFS KM curves – dostarlimab (GARNET main analysis set, N=) versus doxorubicin (ZoptEC main analysis set, N=*)[8-10] .......................................................................................................................... 89 Figure 30: MAIC PFS KM curves for dostarlimab based on GARNET versus doxorubicin based on Makker et al. (2013)[11] ................................................................................................................................................. 94 Figure 31: MAIC PFS KM curves for dostarlimab based on GARNET versus carboplatin plus paclitaxel based on Rubinstein et al. (2019)[59] ............................................................................................................... 94 Figure 32: MAIC PFS KM curves for dostarlimab based on GARNET versus carboplatin plus paclitaxel based on Mazgani et al. (2008)[60] .................................................................................................................. 95 Figure 33: MAIC OS KM curves for dostarlimab versus paclitaxel or doxorubicin[a] based on McMeekin et al. (2015)[6] ........................................................................................................................................................... 96 Figure 34: MAIC OS KM curves for dostarlimab versus doxorubicin based on Makker et al. (2013)[11] ......... 97 Figure 35: MAIC OS KM curves for dostarlimab versus PLD based on Julius et al. (2013)[7] ........................ 97 Figure 36: MAIC OS KM curves for dostarlimab versus carboplatin plus paclitaxel based on Rubinstein et al. (2019)[59] ..................................................................................................................................................... 98 Figure 37: MAIC OS KM curves for dostarlimab versus carboplatin plus paclitaxel based on Mazgani et al. (2008)[60] ......................................................................................................................................................... 98 Figure 38: PSM structure schematic ........................................................................................................... 129 Figure 39: Log cumulative hazard plot between PFS in GARNET and TTNT in the UK RWE study .......... 138 Figure 40: Schoenfeld residual plot between PFS in GARNET and TTNT in the UK RWE study .............. 139 Figure 41: Dostarlimab PFS extrapolations up to five years (GARNET ITT population) (treatment waning applied) ....................................................................................................................................................... 140 Figure 42: Dostarlimab PFS extrapolations up to 40 years (GARNET ITT population) (treatment waning applied) ....................................................................................................................................................... 141 Figure 43: UK RWE GARNET-like cohort PFS extrapolations (based on TTNT) up to five years .............. 143 Figure 44: UK RWE GARNET-like cohort PFS extrapolations (based on TTNT) up to 40 years ............... 144 Figure 45: Log cumulative hazard plot between OS in GARNET and the UK RWE study ......................... 145 Figure 46: Schoenfeld residual plot between OS in GARNET and the UK RWE study .............................. 146 Figure 47: Dostarlimab OS extrapolations up to five years (GARNET ITT population) (treatment waning applied) ....................................................................................................................................................... 148 Figure 48: Dostarlimab OS extrapolations up to 40 years (GARNET ITT population) (treatment waning applied) ....................................................................................................................................................... 148 Figure 49: UK RWE GARNET-like cohort OS extrapolations up to five years ............................................ 150 Figure 50: UK RWE GARNET-like cohort OS extrapolations up to 40 years .............................................. 151 Figure 51: Log cumulative hazard plot between ToT in GARNET and TTD in the UK RWE study ............ 153

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 8 of 222

Figure 52: Schoenfeld residual plot between ToT in GARNET and TTD in the UK RWE study ................. 153 Figure 53: Dostarlimab ToT extrapolations up to five years (GARNET ITT population) ............................. 154 Figure 54: Dostarlimab ToT extrapolation adjusted for anticipated real-world prescribing of dostarlimab .. 156 Figure 55: Current clinical management ToT extrapolations up to five years ............................................. 157 Figure 56: Cost-effectiveness acceptability curve for dostarlimab versus current clinical management (dostarlimab list price) ................................................................................................................................. 192 Figure 57: Cost-effectiveness acceptability curve for dostarlimab versus current clinical management (dostarlimab PAS price) .............................................................................................................................. 193 Figure 58: Cost-effectiveness plane for dostarlimab versus current clinical management (dostarlimab list price) ........................................................................................................................................................... 193 Figure 59: Cost-effectiveness plane for dostarlimab versus current clinical management (dostarlimab PAS price) ........................................................................................................................................................... 194 Figure 60: DSA tornado plot for dostarlimab versus current clinical management (dostarlimab list price) . 197 Figure 61: DSA tornado plot for dostarlimab versus current clinical management (dostarlimab PAS price) .................................................................................................................................................................... 198

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 9 of 222

Abbreviations

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 10 of 222

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 11 of 222

B.1 Decision problem, description of the technology and clinical care pathway

Recurrent or advanced dMMR/MSI-H EC

• Dostarlimab is a treatment for patients with recurrent or advanced DNA mismatch repair deficient (dMMR)/microsatellite instability-high (MSI-H) endometrial cancer (EC) that has progressed on or following prior treatment with a platinum-containing regimen.[1]

• • This patient population, which equates to approximately 124 women each year in England (see Section B.1.3.3), reflects a small, well-defined proportion of the total EC population, and represents those with the greatest critical unmet need.

• Once patients are diagnosed with recurrent or advanced EC, they face an extremely poor prognosis; only 15% and 20% of patients diagnosed with advanced EC and recurrent EC, respectively, will survive for longer than five years.[2-5]

• First line platinum-based chemotherapy is the mainstay of initial treatment for patients with recurrent or advanced EC, but regrettably, disease progression is inevitable and only one in every three patients will ever receive another line of treatment – the remaining patients will have died or be too unwell to withstand further treatment. Following progression on or after platinum-based chemotherapy, patients face a bleak prognosis with no recognised standard of care treatments and a median overall survival (OS) of less than one year.[6-11]

Current clinical pathway of care

• With no standard of care treatments available following disease progression on platinumbased chemotherapy, patients in this setting are left with extremely limited and inadequate treatment options, based on unclear and inconsistent treatment guidelines. • Many patients will receive further lines of chemotherapy but by this stage, EC is largely considered to be a chemotherapy-resistant disease.[12] Real-world evidence (RWE) in the UK shows that patients receiving further chemotherapy face a median OS of just **** months *** **** **** **** **** **** **** **** (95% CI: , ).[13] Only % (95% CI: , ) and % (95% CI: , ) of patients were alive after one and two years, respectively.[13] A small number of patients may alternatively receive hormone therapy, despite the lack of published evidence that it provides any benefit in this setting.[14]

• The lack of effective treatment options has a detrimental psychological impact on patients, leaving them feeling underserved and abandoned. The critical unmet need for a more effective treatment is so severe that unlicensed nivolumab monotherapy is temporarily available via the Cancer Drugs Fund (CDF) through a COVID-19 response programme, despite there being no clinical evidence to support its use in this patient population.[15]

Dostarlimab

• Dostarlimab is a novel and innovative immuno-oncology (I-O) therapy that represents a significant step-change for patients with recurrent or advanced dMMR/MSI-H EC tht has progressed on or after a platinum-containing regimen.

• • In the single-arm pivotal GARNET trial (see Section B.2.3.1); interim analysis data when compared with RWE on current clinical management shows that dostarlimab potentially improves survival for this patient group. Overall ****% (95% CI: ****, ****) of patients treated with dostarlimab were alive after one year, and ****% of patients were still alive after two years, *************** the proportion of patients alive at two years in current UK clinical practice based on RWE.[13, 16]

• • Dostarlimab represents a critical addition to the treatment armamentarium for patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or after a platinumcontaining regimen, who will otherwise continue to face an extremely bleak prognosis with a limited life expectancy and almost no hope of receiving effective treatment.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 12 of 222

B.1.1 Decision problem

This submission demonstrates the clinical and cost-effectiveness of dostarlimab within its full marketing authorisation as monotherapy for the treatment of adult patients with recurrent or advanced mismatch repair deficient (dMMR)/microsatellite instability-high (MSI-H) endometrial cancer (EC) that has progressed on or following prior treatment with a platinum-containing regimen.

The decision problem addressed within this submission is broadly consistent with the NICE final scope for this appraisal as outlined in Table 1. The principal difference relates to the comparators considered relevant to this appraisal as detailed in Table 1.

Table 1: The decision problem

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

©GlaxoSmithKline (2021). All rights reserved

==> picture [85 x 437] intentionally omitted <==

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

©GlaxoSmithKline (2021). All rights reserved

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

©GlaxoSmithKline (2021). All rights reserved

==> picture [85 x 357] intentionally omitted <==

Abbreviations: BSC: best supportive care; CDF: Cancer Drugs Fund; DG: diagnostics guidance; dMMR: DNA mismatch repair deficiency; EC: endometrial cancer; GSK: GlaxoSmithKline; IHC: immunohistochemistry; KM: Kaplan-Meier; MSI-H: microsatellite instability-high; NA: not applicable; NCRAS: National Cancer Registry Analysis System; NHS(E): National Health Service (England); NICE: National Institute for Health and Care Excellence; PLD: pegylated liposomal doxorubicin; RWE: real-world evidence.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

©GlaxoSmithKline (2021). All rights reserved

B.1.2 Description of the technology being appraised

A description of the technology being appraised (dostarlimab [Jemperli][®] ) is provided in Table 2.

Table 2: Technology being appraised

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 17 of 222

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 18 of 222

Abbreviations : CHMP: Committee for Medicinal Products for Human Use; dMMR: DNA mismatch repair deficiency; DG: diagnostics guidance; EC: endometrial cancer; ECDRP: European Commission Decision Reliance Procedure; EMA: European Medicines Agency; GSK: GlaxoSmithKline; IHC: immunohistochemistry; IV: intravenous; MHRA: Medicines and Healthcare Regulatory Agency; MSI-H: microsatellite instability-high; NA: not applicable; NHS(E): National Health Service (England); PAS: patient access scheme; PD-1/2: programmed cell death protein 1/2; PDL-1/2: programmed cell death-ligand 1/2; QXW: once every X weeks; SmPC: Summary of Product Characteristics.

B.1.3 Health condition and position of the technology in the treatment pathway

B.1.3.1 Disease overview

Overview of endometrial cancer

EC is a type of uterine cancer that originates in the lining of the womb (uterus), known as the endometrium. The term EC is frequently used synonymously with uterine cancer, since a large majority (~94%) of uterine cancers are EC.[24] However, other types of uterine cancer are clinically distinct and are treated differently to EC.[25]

EC contributes to an estimated 2,162 deaths every year in the UK, with an age-adjusted mortality rate (the number of deaths due to EC occurring in a specified population over a given period of time) of 2.6 per 100,000 patients in 2018.[26, 27] In the UK, EC is responsible for approximately one woman’s death every four hours.[24, 26]

This submission focusses on patients with recurrent or advanced dMMR/MSI-H EC (see Section B.1.3.2) who have progressed on or following prior treatment with a platinum-containing regimen. This patient population, which equates to approximately 124 patients each year in England, focusses on a small, well-defined proportion of the total EC population, and reflects the population of patients with the greatest critical unmet need.

Patients with recurrent or advanced EC face an extremely poor prognosis – only 15% of patients diagnosed with advanced (Stage IV) disease will survive longer than five years, compared to 92.2% of patients with Stage I disease.[2 ] Fewer than half (46.5%) of patients with Stage IV EC will survive for more than one year.[14, 28] Similarly, only 20% of patients who experience disease recurrence from earlier stages of disease will survive for five years, versus 89% of patients without disease recurrence.[3-5] Advanced EC is also associated with a range of debilitating symptoms, deteriorations in physical functioning and health-related quality of life (HRQoL).[29-31 ]

For patients with recurrent or advanced EC, first-line platinum-based chemotherapy is the standard of care, and is currently their last chance to receive effective treatment. Regrettably, most patients will progress past this first line of therapy, and feedback from UK clinical experts is

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 19 of 222

that only one in every three patients will ever receive another line of treatment – the remaining patients will have died or be too unwell to withstand further treatment.[16]

Disease progression carries devastating and distressing consequences and an extremely bleak prognosis. No evidence based standard of care treatments are available in the post-platinum chemotherapy or subsequent settings, leaving patients with extremely limited and inadequate treatment options based on unclear and inconsistent treatment guidelines. The lack of subsequent effective treatment options has a detrimental psychological impact on patients, leaving them feeling underserved and abandoned.

Most patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or following prior treatment with a platinum-containing regimen will go on to receive further lines of chemotherapy, either as monotherapy or as a doublet chemotherapy regimen. However, each subsequent line of chemotherapy results in increased chemoresistance and is associated with a substantial burden of toxicity.[12, 16] Data from a GSK-initiated UK RWE study show that further **** *** chemotherapy in this setting is associated with a median OS of just months (95% CI: , ****), with only ****% and ****% of patients alive after one and two years, respectively (see Section B.2.3.2).[13]

Alternatively, some patients who have high oestrogen or progesterone receptor expression in the tumour (known as hormone receptor positive [HR+]) may receive hormone therapy, such as letrozole or medroxyprogesterone acetate, despite no evidence that it provides any survival benefit in this post-platinum setting.[14] This highlights the significant unmet need clinically, that when faced with extremely limited treatment options, clinicians are willing to administer nonevidence based treatments in the hope that they will provide some benefit for patients in this setting.

Patients with recurrent or advanced EC who progress on or after platinum-based chemotherapy unequivocally deserve an effective, evidence-based treatment option. This critical unmet need is so severe that currently unlicensed nivolumab is available via the Cancer Drugs Fund (CDF) for patients with metastatic or locally advanced dMMR/MSI-H EC through a COVID-19 response programme, despite there being no available clinical evidence in support of its use in this patient population.[15]

B.1.3.2 Disease classification, progression and recurrence

Dostarlimab is a treatment option for adult patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or following prior treatment with a platinum-based chemotherapy regimen, in line with its marketing authorisation.

It is important to note that, whilst patients with recurrent EC and patients with advanced EC may have different treatment histories, they are viewed as one patient population in clinical practice following the treatment of recurrent or advanced EC with platinum-based chemotherapy. As such, patients with recurrent EC and patients with advanced EC in the post-platinum chemotherapy setting are viewed together as one within this appraisal. This population represents the patients that face the worst prognosis, and those with a critical unmet need.

Advanced disease

Upon diagnosis, EC is staged according to the International Federation of Gynaecology and Obstetrics (FIGO) system.[32] FIGO Stages I–II are considered early stage EC, at which point the

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 20 of 222

disease has not spread outside of the uterus. The majority of patients with EC (approximately 80%) are diagnosed at an early stage (Table 3).[33 ]

A smaller number of EC patients (15–20%) will be diagnosed with advanced stage cancer (Stage III and IV), at which point the disease has spread beyond the uterus (Table 3).[33] Patients with advanced stages (Stage III and IV) of disease, have a much poorer prognosis. Only 50% of patients with Stage III EC will survive for five years or more, and this declines drastically at Stage IV, with only 15% of patients surviving longer than five years.[2] Fewer than half (46.5%) of patients with Stage IV EC will survive for more than one year.[14, 28 ]

Table 3: FIGO cancer staging for EC

Abbreviations: EC: endometrial cancer; FIGO: International Federation of Gynaecology and Obstetrics. Source: American Cancer Society.[34]

Recurrent disease

Irrespective of stage, patients with EC can experience disease recurrence, defined as disease that cannot be detected after primary treatment, but then is radiologically or histologically detected again at a later point in time.[35] Overall, an estimated 13% of EC patients will experience disease recurrence in their lifetime, with the majority of recurrences occurring within three years post-treatment.[3] Prognosis drastically worsens in the recurrent setting with only 20% of patients surviving for five years or more, versus 89% of patients without disease recurrence.[3-5]

dMMR/MSI-H EC

dMMR/MSI-H is a molecular biomarker indicating the presence of a defective DNA repair process (Figure 2).[36] EC is reported to have the highest incidence of dMMR/MSI-H across all solid tumours, with approximately 23% of all EC cases classified as dMMR/MSI-H.[18, 37, 38]

DNA mismatch repair (MMR) is a cellular process responsible for identifying and repairing mismatched bases that occur during DNA replication and genetic recombination.[23, 36] The system consists of DNA MMR proteins, which repair insertions or deletions of abnormal DNA within microsatellites (repetitive non-coding DNA sequences) (Figure 2).[36]

Mutations in the genes that code for these proteins can result in a defective MMR process which results in the accumulation of abnormal mutations.[36] This can be caused by sporadic mutations in the genes encoding the MMR proteins, or through inherited conditions such as Lynch

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 21 of 222

syndrome, which is the result of a germline mutation in the genes encoding several MMR proteins.[23] When one or more of these MMR proteins are dysfunctional, or are not expressed, this results in dMMR. Otherwise, the cancer is considered MMR-proficient [pMMR]) and microsatellite stable (MSS) (Figure 2).[36 ]

Microsatellite instability (MSI), a change in the length of repetitive sequences in tumour DNA compared with normal DNA, is the phenotypic (observable characteristic) result of dMMR.[39] MSI can be further characterised as high or low: if two or more DNA repeats are altered, this is specifically defined as MSI-H; if there is only one mutated sequence, this is considered microsatellite instability-low (MSI-L).[36 ]

Figure 2: The mechanism of action of the DNA MMR system and dMMR/MSI-H

==> picture [409 x 352] intentionally omitted <==

Abbreviations: dMMR: DNA mismatch repair deficiency; MSI-H: microsatellite instability-high; MSS: microsatellite stable; pMMR: DNA mismatch repair proficient. Source: Adapted from Eso et al. (2019).[40]

dMMR/MSI-H disease has one of the highest mutational loads versus other molecular subtypes, and is highly immunogenic, with increased levels of circulating tumour infiltrating lymphocytes and high expression of immune checkpoint molecules.[23, 39] This is important, as dMMR/MSI-H tumours are therefore more likely to respond to immuno-oncology (I–O) treatment, including antiprogrammed cell death protein 1 (PD-1) or anti-programmed cell death ligand 1 (PD-L1) therapy such as dostarlimab.[23 ]

Evidence for this has been observed in other cancers, where patients with dMMR/MSI-H disease have experienced improved responses to I-O therapy, compared to patients with pMMR disease.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 22 of 222

In KEYNOTE-016, a Phase 2 study of pembrolizumab in patients with progressive metastatic colorectal cancer, as well as other cancers, pembrolizumab demonstrated an objective response rate (ORR) of 40% in patients with dMMR/MSI-H disease versus 0% in patients with pMMR/MSS disease.[41 ]

Consequently, dMMR/MSI-H EC represents a subgroup where I-O therapy with dostarlimab is most effective.[23]

dMMR/MSI-H testing in the UK

In the UK, recently published NICE diagnostics guidance DG42 recommends that individuals with EC should undergo genetic testing for Lynch syndrome.[18] Lynch syndrome type II (also known as hereditary non-polyposis colorectal carcinoma [HNPCC] syndrome), accounts for up to 3% of all EC cases.[42] Lynch syndrome is a hereditary condition caused by mutations in the MMR genes, that predisposes women to developing EC throughout their lifetime. Whilst the general population risk of developing EC is 2%, women with Lynch syndrome have a 30-60% lifetime risk of developing EC.[12, 43]

NICE DG42 states that all patients with EC should be tested for dMMR/MSI-H using immunohistochemistry (IHC) testing, and if indicative of dMMR/MSI-H, patients are offered further germline genetic testing to confirm Lynch syndrome.[18] IHC testing is a simple, inexpensive technique already routinely used within the NHS to test for dMMR/MSI-H in other cancers, including colon cancer.[44]

UK clinical expert opinion sought by GSK for this submission agreed that IHC would be the preferred testing approach for dMMR/MSI-H EC and that all patients eligible for treatment with dostarlimab would receive dMMR testing as a result of this guidance.[16] Consultation with NHS England (NHSE) via an NHSE surgery also confirmed the availability of dMMR testing across England, and that access to testing will not be a barrier to accessing dostarlimab upon reimbursement. Consequently, dMMR testing via IHC will soon become standard of care for all patients with EC, and no additional diagnostic tests will be required to facilitate the prescribing of dostarlimab beyond those already conducted for patients with EC in UK NHS clinical practice.

Use of immunotherapy in dMMR/MSI-H disease

dMMR/MSI-H disease has one of the highest mutational loads versus other molecular subtypes, and is highly immunogenic, with increased levels of circulating tumour infiltrating lymphocytes and high expression of immune checkpoint molecules.[40] This high number of tumour antigens within the tumour microenvironment observed in dMMR/MSI-H tumours suggests these cancers may be more likely to respond to immuno-oncology (I–O) treatment, including anti-programmed cell death protein 1 (PD-1) or anti-programmed cell death ligand 1 (PD-L1) therapy such as dostarlimab.[23]

Evidence for this has been observed in other cancers, where patients with dMMR/MSI-H disease have experienced improved responses to I-O therapy, compared to patients with pMMR disease. KEYNOTE-016 was a Phase 2 single arm study of pembrolizumab in patients with progressive metastatic colorectal cancer, as well as other cancers. In the study pembrolizumab demonstrated an ORR of 40% in patients with dMMR/MSI-H disease versus 0% in patients with pMMR/MSS disease.[41]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 23 of 222

Consequently, dMMR/MSI-H EC represents an extremely promising, biomarker selected patient population for dostarlimab.[23]

B.1.3.3 Epidemiology

There are an estimated 7,539 new patients diagnosed with EC in England every year.[12] This submission focusses on a small proportion of these patients – those with recurrent or advanced dMMR/MSI-H EC that has progressed on or after platinum-based chemotherapy. This is a small, well-defined proportion of the total number of patients with EC, and reflects those patients with the poorest prognosis, and critical unmet need.

Approximately 18% of EC patients will be diagnosed with advanced EC (Stage III or IV) at first presentation, additionally approximately 13% of patients who are diagnosed at early stages of EC will later experience disease recurrence.[3, 5, 33] This means that approximately 2,337 patients will be diagnosed with recurrent or advanced EC in England each year. Of these, UK clinical expert opinion sought by GSK estimates that approximately two in three of these patients (64%) will receive first-line treatment with platinum-based chemotherapy for their disease.[16]

Regrettably, almost all patients who receive first-line platinum-based chemotherapy for recurrent or advanced EC will subsequently experience disease progression – in this setting, patients are no longer treated with curative intent, and disease progression is inevitable. The majority of patients receiving first-line platinum-based chemotherapy will never receive a subsequent line of treatment, and will have already died, or will be too severely unwell to receive further treatment with chemotherapy or hormone therapy. UK clinical expert opinion indicates that only one in every three patients (36%) who receive first-line platinum based chemotherapy will be eligible for further treatment with chemotherapy or hormone therapy, equating to approximately 538 patients in England every year.[16]

UK RWE evidence collected by GSK found that ****% of patients received a chemotherapy subsequent to first-line platinum based chemotherapy.[13] It is likely that the discrepancy between the ****% derived from the RWE study and the 36% from UK clinical expert opinion may be due to patients receiving hormone therapy in this setting within UK clinical practice, which could not be captured in the RWE study (see Section B.2.3.2).[13, 16]

Of the 538 patients with recurrent or advanced EC in England that suffer disease progression on or after platinum-based chemotherapy and are eligible for further treatment with chemotherapy or hormone therapy, approximately 23% will be classified with dMMR/MSI-H EC and will be eligible for treatment with dostarlimab.[18] This means that approximately 124 new patients will be eligible for treatment with dostarlimab in England each year. These assumptions are detailed in Figure 3 below.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 24 of 222

Figure 3: Estimated dostarlimab-eligible patient population numbers in England

==> picture [468 x 263] intentionally omitted <==

Footnotes:[a ] 7,862 incident uterine cancer cases in 2017, adjusted to 2021 using an annual general population growth rate of 0.5%.[18, 45]

Abbreviations: DG: diagnostics guidance; dMMR: DNA mismatch repair deficiency; EC: endometrial cancer; MSI-H: Microsatellite instability-high; NA: not applicable.

Source: Fung-Kee-Fung et al. (2006);[3] Odagiri et al . (2011);[5][a] Cancer Research UK;[33][b] Cancer Research UK;[46] c Cancer Research UK;33 d GSK Data on File;16 e NICE DG42.18

B.1.3.4 Clinical care pathway

Clinical guidelines for EC

There are a number of clinical guidelines available for the management of EC, however, they present very limited guidance for patients with recurrent or advanced EC who have progressed on or after platinum-based chemotherapy. Clinical guidelines for the management of EC are available from the British Gynaecological Cancer Society (BGCS), the European Society of Medical Oncology (ESMO), the European Society of Gynaecological Oncology (ESGO), the European Society for Radiotherapy and Oncology (ESTRO) and the European Society of Pathology (ESP).[14, 47, 48] There are no published NICE guidelines for EC.

Details of the current treatment pathway for patients with EC in the UK are presented in Figure 4 and below. These are based on the recommendations from key clinical guidelines, as well as UK clinical expert opinion and a RWE study conducted by GSK on clinical treatment patterns for patients with recurrent or advanced EC using linked patient-level health data available through the National Cancer Registry Analysis System (NCRAS) in England (hereafter referred to as the UK RWE study).[14, 47, 48]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 25 of 222

Figure 4: Treatment pathway for patients with EC in the UK and the anticipated positioning of dostarlimab

==> picture [390 x 272] intentionally omitted <==

Footnotes:[a] At any stage of disease, patients may also receive neoadjuvant or adjuvant radiotherapy, chemotherapy or hormone therapy, in addition to surgery.[b] Further chemotherapy may consist of carboplatin plus paclitaxel, doxorubicin or gemcitabine, carboplatin monotherapy, paclitaxel monotherapy, doxorubicin monotherapy, among others.[c ] UK clinical expert opinion sought by GSK indicated that although not licensed, hormone therapy would consist of either letrozole or medroxyprogesterone acetate in this setting. Abbreviations : dMMR: DNA mismatch repair deficiency; EC: endometrial cancer; MSI-H: microsatellite instability-high.

To provide context, a brief summary of the treatment pathway for early EC is provided in the sections below. This is followed by a more detailed summary of the treatments available for patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or following prior treatment with a platinum-containing regimen, as this is the indication of relevance to this submission.

B.1.3.4.1 Initial management of EC

The initial management of EC typically involves surgical treatment, which may include total hysterectomy (removal of the uterus and cervix) and bilateral salpingo-oophorectomy (removal of both ovaries and the fallopian tubes) with or without lymphadenectomy (removal of one or more groups of lymph nodes), depending on the stage of disease at diagnosis.[47, 48]

Patients with early stage EC (Stages I and II) receive surgery with curative intent. However, 13% of patients with early stage disease will experience disease recurrence in their lifetime, with the majority of recurrences occurring within the first three years of treatment.[3-5] Approximately 15– 20% of patients are diagnosed with advanced EC.[33] For these patients, surgical treatment may still be considered, however, it is not likely to cure their disease.[47]

After surgery, patients at any stage of disease may also receive (neo)adjuvant radiotherapy or chemotherapy, depending on their risk factors.[47 ] Hormone therapy, including progestogens (e.g. megestrol, medroxyprogesterone acetate), tamoxifen and aromatase inhibitors (e.g. anastrozole,

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 26 of 222

letrozole) represent alternative treatment options for some patients at this early stage in the treatment pathway. These would typically be offered to patients who have high oestrogen or are HR+.[14]

B.1.3.4.2 Treatment for recurrent or advanced EC

First-line treatment for recurrent or advanced EC

For patients with recurrent or advanced EC, first-line platinum-based chemotherapy currently represents the last available standard of care treatment. In the UK, doublet chemotherapy with carboplatin plus paclitaxel is accepted as the standard of care in this setting. The RWE study conducted by GSK showed that ****% of patients with recurrent or advanced EC received platinum doublet therapy as first-line treatment for recurrent or advanced EC, and carboplatin plus paclitaxel was the most commonly prescribed regimen.[13] However, as described previously, a majority of patients will progress after their first-line platinum based-chemotherapy.

All future lines of anti-cancer therapy – positioning of dostarlimab in this appraisal:

It is estimated that approximately one in three patients will be eligible for further treatment following disease progression on or after platinum-based chemotherapy for recurrent or advanced EC. In this setting, disease progression carries devastating and distressing consequences, and leaves patients facing a bleak prognosis with almost no hope of receiving further effective treatment.[48]

The lack of standard of care, or even a licensed treatment option at this stage, leads to patients feeling abandoned, with only extremely limited and inadequate treatment options based on unclear and inconsistent treatment guidelines. With a distinct lack of treatment options, the vast majority of patients are either treated with further chemotherapy, hormone therapy, or are enrolled into clinical trials. Each of these options is described in the following sections.

Chemotherapy

In UK clinical practice, most patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after platinum-based chemotherapy will receive treatment with further chemotherapy however there is extremely limited consensus on which chemotherapy regimens to prescribe. BGCS guidelines recommend that re-challenge with carboplatin plus paclitaxel can be considered in fit patients.[14] However, ESMO guidelines note that there are no standard of care treatments in this setting, and currently used treatments are associated with disappointing response rates; only paclitaxel has consistently shown a response rate above 20% (and these response rates predate the use of paclitaxel as a prior treatment, meaning that they likely represent optimistic estimates for what might be achieved in UK clinical practice today).[14, 47] One study by Lincoln et al. (2003), reported an ORR for paclitaxel monotherapy of 27.3%.[49]

The UK RWE study conducted by GSK, using data from the NCRAS (see Section B.2.3.2) highlights the lack of consensus in UK clinical practice, with patients receiving a wide range of alternative chemotherapy regimens (Table 4).[13, 16]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 27 of 222

Table 4: Ten most common chemotherapy regimens received by patients with recurrent or advanced EC following disease progression on platinum-based doublet chemotherapy

Abbreviations: EC: endometrial cancer; PLD: pegylated liposomal doxorubicin. Source: GSK Data on File.[13]

Hormone therapy

UK clinical expert opinion sought by GSK suggests that hormone therapy, such as medroxyprogesterone acetate or letrozole, may also be a treatment option for a small select number of patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after platinum-based chemotherapy.[16] These patients may be treated with hormone therapy despite the fact that BGCS guidelines highlight that there is no evidence that hormone therapy confers any survival benefit in the post-platinum setting.[14] This is substantiated by the results of a targeted literature review conducted by GSK (detailed in Appendix L), which did not identify any published evidence for hormone therapy in patients with recurrent or advanced EC that have received prior platinum-based chemotherapy.

Furthermore, UK clinical experts indicated that survival with hormone therapy would not be expected to exceed that observed in the UK RWE study, estimating that the median PFS and OS for hormone therapy in this setting would be approximately 3 months and approximately 6 months,[16] respectively, whereas median PFS associated with the chemotherapy regimens that *** *** *** constitute current clinical management in the UK RWE study is (95% CI: , ) months and **** *** **** median OS is months (95% CI: , ).[13]

Clinical trials

Finally, UK clinical expert opinion sought by GSK during the development of this submission highlighted that at this stage of treatment, given the distinct absence of an established standard of care, they would actively seek to enrol their patients in a clinical trial, due to the disappointing outcomes and toxicity associated with the limited treatment options that are currently available in clinical practice.[16] ESGO/ESTRO/ESP guidelines also note that clinical trial participation should be offered to all patients with relapsed disease, highlighting the inadequacy of currently available treatments.[48]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 28 of 222

B.1.3.5 Limitations of current treatment and unmet need

Limited survival associated with current treatments

Sadly, patients with recurrent or advanced dMMR/MSI-H EC that have experienced disease progression on or after platinum-based chemotherapy face a distressing and bleak prognosis, regardless of which current treatment regimen they receive, and there is no evidence that hormone therapy confers any survival benefit in the post-platinum chemotherapy setting.[14] UK clinical experts have agreed that there is little expectation that either chemotherapy or hormone therapy are effective in this setting.[16]

The UK RWE study conducted by GSK (see Section B.2.3.2) found that patients in this setting **** *** **** have a median OS of just months (95% CI: , ) following the initiation of further chemotherapy, and only ****% and ****% of patients were still alive after one and two years, respectively (Figure 5).[13]

Figure 5: UK RWE OS for patients with recurrent or advanced EC who received further chemotherapy following disease-progression on or after platinum-based chemotherapy

==> picture [384 x 262] intentionally omitted <==

Abbreviations: CI: confidence interval; EC: endometrial cancer; OS: overall survival; RWE: real-world evidence. Source: GSK Data on File.[13]

Patients face even worse PFS; the UK RWE study found that patients had a median PFS of just *** *** *** months (95% CI: , ) from the initiation of 2L chemotherapy (using time to next treatment [TTNT] as a proxy for PFS). Only ****% of patients were progression-free one year after the initiation of treatment, and this number dropped to just ****% of patients after two years (Section B.2.4.5.2).[13]

Data from the published literature paint a similarly devastating picture for patients with recurrent or advanced EC following disease progression on platinum-based chemotherapy. Various clinical trials have reported that patients have a median OS of less than one year, with median PFS ranging from three to six months.[6-11] Published clinical outcomes for patients in this setting according to treatment are summarised in Section B.2.4 and Appendix D.5.2 and D.5.3.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 29 of 222

Toxicity and HRQoL burden of further chemotherapy

Alongside the extremely limited clinical benefit, cytotoxic chemotherapy regimens are also associated with a number of harmful and debilitating side effects, including leukopenia, neutropenia, anaemia, fatigue, nausea, vomiting and alopecia.[30,10] UK clinical expert opinion sought by GSK has highlighted the toxicity burden of chemotherapy that lasts well beyond the duration of treatment, and has a detrimental impact on HRQoL for patients with EC.[16 ]

The substantial detrimental impact of chemotherapy was highlighted in the PORTEC-3 trial, where patients completed the European Organisation for the Research and Treatment of Cancer (EORTC) QLQ-C30, an internationally validated HRQoL questionnaire for cancer.[50] Patients with EC that received first-line platinum-based chemotherapy (in addition to radiotherapy) reported significantly lower scores (worse functioning) across most EORTC QLQ-C30 functioning scales, and significantly higher symptom scores (worse symptoms), versus patients receiving radiotherapy alone, following completion of radiotherapy and at Month 6.[50] Notably, these patients were only receiving their first-line of platinum-based chemotherapy. There is little reason to suggest that the patients relevant to this submission, who have already received prior treatment with platinum-based chemotherapy, would not experience at least a comparable decline in HRQoL on initiation of further chemotherapy treatment.

The toxicity burden of chemotherapy for patients with recurrent or advanced EC in the postplatinum setting was demonstrated in the ZoptEC trial, which found that ****% of patients receiving doxorubicin experienced either a treatment emergent adverse event (TEAE) or treatment-related TEAE of ≥ Grade 3 severity.[8-10] According to the Common Terminology Criteria for Adverse Events (CTCAE), a Grade 3 AE represents an event which is “severe or medically significant but not immediately life-threatening; hospitalisation or prolongation of hospitalisation indicated; disabling; limited self-care activities of daily living”.[51]

Severe unmet need for patients

The lack of effective treatment options for patients with recurrent or advanced dMMR/MSI-H EC that have progressed on or after platinum-based chemotherapy may have a substantially detrimental impact on patients and their families, leaving them feeling underserved and abandoned. There is a crucial unmet need for the introduction of new effective treatment options to be routinely available for these patients.

The interim introduction and availability of nivolumab via the CDF, despite a lack of data for its use in this patient population, highlights the severity of the unmet need for dMMR/MSI-H EC patients, and the enthusiasm within the UK clinical community for access to an I-O therapy for these patients. The introduction of a licensed, scientifically supported treatment option would be preferred by UK clinicians, given the additional data available when making a prescribing decision.

B.1.3.6 Dostarlimab

Dostarlimab is an I-O therapy and the first licensed PD-1 inhibitor monotherapy for patients with EC in the UK. Dostarlimab is positioned as a treatment option for adult patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or following prior treatment with a platinum containing regimen .[1] This is aligned with the marketing authorisation for dostarlimab, and the patient population included within the pivotal GARNET trial (see Section B.2.3.1).

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 30 of 222

As a novel and innovative I-O therapy in EC, the introduction of dostarlimab represents a clinically significant change in the management of patients with recurrent or advanced dMMR/MSI-H EC in the post-platinum chemotherapy setting.

In the GARNET trial (see Section B.2.3.1), treatment with dostarlimab resulted in additional survival for patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after platinum-based chemotherapy, which has led to substantial excitement in the clinical community. Overall, ****% of patients treated with dostarlimab were still alive after one year, and ****% of patients were still alive after two years. In comparison, RWE in the UK for patients receiving a range of chemotherapies found that only ****% and ****% of patients were alive after one and two years, respectively. These survival results represent a truly clinically meaningful benefit for patients, who are otherwise at the end of their lives in current UK clinical practice.

Notably, in other cancers, I-O therapies have been shown to result in extended treatment benefits and long-term remission even after treatment discontinuation, offering a substantially improved prognosis for many patients.[52] Indeed, the long-term benefits of I-O therapies have been demonstrated across multiple indications including melanoma, lung, head and neck, where patients who discontinued therapy had durable responses that extended beyond the end of treatment.[53] Given this trend, it is reasonable to believe some patients who respond to dostarlimab may continue to experience extended treatment benefits and long-term remission beyond the two-year follow-up in the GARNET trial to date.

The introduction of dostarlimab would represent a shift in the treatment armamentarium for clinicians and patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or after platinum-based chemotherapy. Patient groups have shared that dostarlimab would provide hope to those who currently feel abandoned and currently face an extremely distressing prognosis with almost no chance of receiving effective treatment.

B.1.3.6.1 Anticipated positioning of dostarlimab in UK clinical practice

At a recent advisory board UK clinicians agreed that in clinical practice, dostarlimab is expected to be a treatment for patients who would otherwise receive further chemotherapy or hormone therapy, which would represent the most relevant comparators for this submission. Based on this anticipated positioning, a range of further chemotherapy regimens represent the most relevant comparators to dostarlimab in this submission, alongside hormone therapy in some patients.

As a result of the lack of definitive standard of care in this setting and the wide range of treatments used, in the base case economic analysis for this appraisal, GSK have primarily considered a comparison with a basket of the most commonly used chemotherapy regimens representing current clinical management in UK clinical practice, using data collected via the UK RWE study detailed in Section B.2.3.2.[13]

The UK RWE comparison incorporates the NICE final scope comparators as well as the most utilised chemotherapy regimens in UK clinical practice, which include carboplatin plus doxorubicin and carboplatin plus gemcitabine, amongst others. The identification of treatment regimens in the UK RWE study other than those included in the NICE final scope demonstrates the lack of consensus for the management of recurrent or advanced EC that has progressed on or after platinum-based chemotherapy.

In addition, GSK have considered individual comparisons in scenario analyses between

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 31 of 222

dostarlimab and the individual chemotherapy regimens listed in the NICE final scope for which published data exist.

Hormone therapy is included as a comparator in this submission, since it is included in the NICE final scope and has been confirmed as an appropriate comparator by UK clinical experts.[16] UK clinical experts agreed that medroxyprogesterone acetate and letrozole represent the most relevant hormone therapies used in this setting in UK clinical practice. However, as detailed in Section B.2.2 and B.2.7.3, there was no data identified for hormone therapy in patients with recurrent or advanced EC who have progressed on or after platinum-based chemotherapy, meaning that any indirect comparisons are difficult, and would be associated with substantial limitations. It is also important to reiterate that there is no published evidence that hormone therapy provides any survival benefit in the post-platinum setting for this defined group of patients.[14 ]

The NICE final scope also lists BSC as an additional comparator, but GSK does not consider this to be relevant to this submission. BSC may be given as add-on therapy for patients with current or advanced EC in the post-platinum setting receiving further chemotherapy treatment, and it is anticipated that it may also be given to patients receiving dostarlimab, thus would in effect cancel one another out within the context of an economic analysis. Dostarlimab would not replace the use of BSC, and therefore BSC should not be considered a relevant comparator to dostarlimab in this submission.

B.1.4 Equality considerations

Due to the lack of effective treatment options for patients with recurrent or advanced dMMR/MSIH EC who progress on or after platinum-based chemotherapy, feedback from UK clinical experts suggests that many clinicians have no choice but to seek to enrol these patients into clinical trials[.16] However, clinical trials are time bound, and the distribution, enrolment criteria and number of trial sites are restricted. Moreover, clinical trials are generally only available to patients treated in larger hospitals or near larger trial sites and are unlikely to be an option for patients in more rural areas of the UK. As such, the lack of nationally funded treatment options represents a possible equity concern.

Pembrolizumab is an alternative I-O therapy to dostarlimab, that has demonstrated efficacy for patients with recurrent or advanced EC who progress on or after platinum-based chemotherapy. However, it is an unlicensed therapy in Europe, and during the recent NICE scoping workshop, it was highlighted that pembrolizumab is currently only available in the private market in the UK. This represents an equity issue, whereby it is only available for patients who are able to afford the cost of treatment or are covered via private health insurance.

The availability of dostarlimab as a licensed, nationally funded treatment option on the NHS for patients whose only alternative options for receiving active treatment are via entry into a clinical trial or via private treatment with pembrolizumab would therefore help to address these equity issues.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 32 of 222

B.2 Clinical effectiveness

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 33 of 222

a disease control rate (DCR) of % (). This represents a marked increase compared to current clinical management. European Society of Medical Oncology (ESMO) guidelines highlight that for patients with advanced EC recurring after first-line chemotherapy, only paclitaxel has consistently shown a response rate >20%, less than half the ORR achieved by dostarlimab.[47] However, one study by Lincoln et al. (2003), reported an ORR for paclitaxel monotherapy of 27.3%.[49] • Importantly, the responses to dostarlimab were durable; ****% (*****) of responders maintained a response up to Month 12, and by Month 18, ****% (***) of responders were still experiencing an ongoing response to dostarlimab. PFS and OS for patients receiving dostarlimab versus current clinical management • Results for PFS from GARNET show a clear benefit in favour of dostarlimab and paint current clinical management in a harrowing light. At Month 12, ****% of patients treated with dostarlimab in the ITT population (n=129) were progression-free, compared to just ****% of patients treated with current clinical management from the UK RWE study. At Month 24, ****% of patients treated with dostarlimab remained progression-free, compared with just ****% in the UK RWE study (see Section B.2.4.5). • Patients in GARNET also experienced a remarkable survival benefit compared with current clinical management in the UK. By Month 12, ****% of patients treated with dostarlimab in the ITT population (n=129) were still alive versus just ****% of patients in the UK RWE study. By Month 24, ****% of patients treated with dostarlimab were still alive. In contrast, the UK RWE study showed no evidence of a long-term survival benefit with current clinical management; by Month 24, just ****% of patients were still alive – ************** of the proportion of patients treated with dostarlimab that were still alive by Month 24 (see Section B.2.4.6). The results of the OS MAIC were consistent with this naïve comparison, demonstrating a similar magnitude of OS benefit for patients treated with dostarlimab versus current clinical management (unadjusted hazard ratio [HR]: ****; 95% CI: **********; adjusted HR: ****; 95% CI: **********) (see Section B.2.7.1). HRQoL, safety and tolerability associated with dostarlimab • In terms of HRQoL, the GARNET study showed that treatment with dostarlimab preserved patient-reported HRQoL from baseline. Key disease-related symptom subscales, such as pain and fatigue showed a positive trend of improvement throughout the study. • Dostarlimab was also shown to be well-tolerated and associated with a manageable AE profile in line with other currently licensed anti-programmed cell death ligand 1 (PD-L1) therapies. Treatment emergent adverse event (TEAEs) related to treatment were generally low grade (only % of patients reported any Grade ≥3 treatment-related TEAE), and discontinuation as a result of treatment-related AEs was low (%). The most frequent treatment-related TEAEs were diarrhoea (%) and asthenia (*%). • In contrast, cytotoxic chemotherapy regimens are associated with debilitating AEs. Whilst safety data were not collected in the UK RWE study, data from the ZoptEC study report that 96.4% of patients receiving doxorubicin monotherapy experienced treatment-related TEAEs (compared to ****% in GARNET), and ****% of patients reported any Grade ≥3 TEAE (compared to ****% in GARNET).[8-10] Conclusion • The remarkable survival outcomes for patients in GARNET, combined with a well-tolerated safety profile, mean that the introduction of dostarlimab in the UK would provide hope to patients with recurrent or advanced dMMR/MSI-H EC in the post-platinum chemotherapy setting who currently feel abandoned and face an extremely distressing prognosis, with almost no chance of receiving effective treatment.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 34 of 222

B.2.1 Identification and selection of relevant studies

A systematic literature review (SLR) was conducted to identify relevant clinical evidence for the efficacy and safety of dostarlimab and the chemotherapy comparators listed in the NICE final scope for the treatment of recurrent or advanced EC that has progressed on or after platinumbased chemotherapy .

In total, 3,077 publications were screened, of which 148 publications were reviewed at the fulltext stage. After exclusion of publications not meeting the eligibility criteria, 23 publications (reporting on 13 unique studies) were included in the SLR. Full details of the SLR, including the search strategy, study selection process and detailed results, are presented in Appendix D.

B.2.2 List of relevant clinical effectiveness evidence

B.2.2.1.1 Dostarlimab (GARNET)

Of the 13 studies included in the clinical SLR, one clinical trial, GARNET (NCT02715284), was identified for dostarlimab and this represents the pivotal clinical trial for dostarlimab in this indication.[56-58] Further details on the GARNET trial are presented in Section B.2.3.1.

B.2.2.1.2 Comparators (NICE final scope)

The remaining 12 studies included in the clinical SLR investigated relevant chemotherapy regimens in patients with recurrent or advanced EC who have progressed on or after platinumbased chemotherapy. These trials included patients receiving carboplatin plus paclitaxel, paclitaxel monotherapy and doxorubicin monotherapy. No studies were identified for carboplatin monotherapy.

In addition, hormone therapy was not included in the original SLR, because it was not considered to be a relevant comparator to dostarlimab at the time the review was conducted. Following the inclusion of hormone therapy in the NICE final scope and discussions with UK clinical experts that indicated that hormone therapy would be considered in a small subset of patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or after platinum-based chemotherapy, a targeted literature review of PubMed was conducted to identify relevant evidence for hormone therapy, but no suitable studies were identified (see Appendix L).

For the studies that were identified in the clinical SLR for comparator therapies, there was a distinct paucity of reported data. Most studies in the relevant patient population were observational studies, where patient characteristics and KM survival data were poorly reported, limiting the quality, and therefore increasing the uncertainty, of any potential ITCs. Nevertheless, for completeness, a series of ITCs have been conducted between dostarlimab and the individual chemotherapy comparators listed in the NICE final scope where possible, based on available published data identified in the clinical SLR. Given the limitations associated with these analyses, they are provided for completeness as supportive comparative efficacy evidence only and are presented in Section B.2.7.

It should be noted that despite efforts made to identify alternative sources of data for hormone therapy and carboplatin monotherapy (as no relevant studies were identified in the literature), feedback from UK clinical experts strongly indicated that any data for patients not in the postplatinum chemotherapy setting would not be suitable to use as a proxy for these therapies. As

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 35 of 222

such, it was not possible to conduct any individual comparisons in terms of comparative efficacy between dostarlimab and carboplatin monotherapy or hormone therapy.

B.2.2.1.3 Current clinical management (UK RWE study)

Given the limitations of the studies identified in the clinical SLR for the comparators listed in the NICE final scope, and to provide a more accurate representation of the current clinical management for recurrent or advanced EC that has progressed on or after platinum-based chemotherapy in the UK, a RWE study was conducted by GSK using NCRAS data.[13] This study included a population of patients closely aligned to the patients in GARNET that received a range *** of chemotherapy regimens that represent current clinical treatment paradigms in the UK (N= ).

Given the large sample size of this study, together with the close alignment to the patient characteristics in the GARNET trial, and the real-world representation of current clinical management in this difficult-to-treat population, this UK RWE study serves as the primary comparative efficacy evidence in this submission and informs the base case cost-effectiveness analysis.

Full details of the UK RWE study are presented in Section B.2.3.2. To explore the impact of any potential remaining differences between the GARNET and UK RWE study patient populations, an ITC was conducted between GARNET and the UK RWE study for OS, and details of this analysis are presented in Section B.2.7.1.

An overview of the comparative efficacy evidence presented in this submission is presented in Table 5 .

Note that not all of the comparative efficacy evidence was considered suitable for decisionmaking and therefore not all of the analyses have been included within the economic model. Full details of the comparative efficacy evidence included within the economic model are presented in Section B.3.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 36 of 222

Table 5: Summary of the clinical effectiveness evidence presented in this submission

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

©GlaxoSmithKline (2021). All rights reserved

Footnotes:[a] Due to the differences in PFS in GARNET versus TTNT in the UK RWE study, it was only possible to conduct a MAIC using a Cox proportional hazards model for OS between GARNET and the UK RWE study. [b ] Patients in the McMeekin et al. (2015)[6] study received either paclitaxel or doxorubicin. Clinical expert opinion indicated that the efficacy between the two treatments is likely to be similar, and therefore it is appropriate to consider this as one combined arm that provides evidence for both paclitaxel monotherapy and doxorubicin monotherapy.[c] PFS curves for PFS were not reported in McMeekin et al. (2015) and Julius et al. (2013), meaning it was only possible to conduct OS MAICs versus these studies.[6, 7][d] It was not possible to use IPTW to estimate a HR for PFS between dostarlimab and doxorubicin, due to differences in the definition of PFS and the timepoints of tumour assessments between GARNET and ZoptEC (detailed in Appendix D.5.2).[8-10]

Abbreviations: BSC: best supportive care; IPTW: inverse probability treatment weighting; ITC: indirect treatment comparison; MAIC: matching-adjusted indirect comparison; NA: not applicable; OS: overall survival; PFS: progression-free survival; RWE: real-world evidence.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802]

©GlaxoSmithKline (2021). All rights reserved

B.2.3 Summary of methodology of GARNET and the UK RWE study

B.2.3.1 Dostarlimab (GARNET)

The clinical SLR identified one trial for dostarlimab in patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after platinum-based chemotherapy: the GARNET trial, an open-label, single-arm, multicentre, non-randomised Phase I trial (NCT02715284).

GARNET was conducted in two parts: Part 1 of the study, which established the recommended dose for dostarlimab (dose escalation), and Part 2, which was conducted in two subparts (2A and 2B). Part 2A evaluated the safety and tolerability of dostarlimab in fixed-dose safety evaluation cohorts. Part 2B (the extension phase) investigated the efficacy of dostarlimab in five expansion cohorts according to the following tumour types: dMMR/MSI-H EC (Cohort A1), MMRproficient/MSS EC (Cohort A2), non-small cell lung cancer (NSCLC) (Cohort E); dMMR/MSI-H or POLE-mutated non-EC (Cohort F) and platinum-resistant ovarian cancer (PROC) without known breast cancer susceptibility gene mutation (BRCA).[54]

This submission focuses solely on Part 2B, Cohort A1 of GARNET, the cohort of patients with recurrent or advanced dMMR/MSI-H EC, in alignment with the indication of relevance to this submission and the licensed indication for dostarlimab (Figure 6).

Figure 6: GARNET trial design

==> picture [446 x 167] intentionally omitted <==

Abbreviations : BRCA: breast cancer susceptibility gene; dMMR: DNA mismatch repair deficiency; EC: endometrial cancer; MSS: microsatellite stable NSCLC: non-small cell lung cancer; pMMR: mismatch repair proficient; POLE: polymerase ε; PROC: platinum-resistant ovarian cancer. Source: GSK Data on File.[54]

B.2.3.1.1 Summary of trial methodology

A summary of the trial methodology for the GARNET trial is presented in Table 6.

Table 6: Summary of methodology for GARNET trial

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 39 of 222

• France • Italy • Spain • United States

• Trial design Ongoing, open-label, single-arm, multicentre, non-randomised Phase I trial Population The extension phase of the GARNET trial enrolled five cohorts of patients: • Cohort A1: patients with recurrent or advanced dMMR/MSI-H EC that has progressed after treatment with a platinum-containing chemotherapy regimen

• • Cohort A2: patients with recurrent and advanced MMRproficient/microsatellite stable (MSS) EC

• • Cohort E: non-small cell lung cancer (NSCLC)

  • Cohort F: dMMR/MSI-H or polymerase ε (POLE)-mutated non-EC

  • Cohort G: platinum-resistant ovarian cancer without known BRCA mutations

This submission will focus solely on Cohort A1, the cohort of patients with recurrent or advanced dMMR/MSI-H EC, in alignment with the indication of relevant to this submission and the licensed indication for dostarlimab. Cohorts A2, E, F and G are not within the scope of this submission and they do not align with the marketing authorisation for dostarlimab. Therefore, the results of these cohorts will not be presented or discussed further within this submission.

Figure 7: Cohorts in the GARNET trial

==> picture [281 x 105] intentionally omitted <==

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 40 of 222

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 41 of 222

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 42 of 222

Footnotes:[a ] Additional details/timings and results for these outcomes can be found in the GARNET Clinical Study Report and are not presented within this submission.

Abbreviations: BICR: blinded independent central review; DCR: disease control rate; DOR: duration of response; EC: endometrial cancer; ECOG: Eastern Cooperative Oncology Group; EOT: end-of-treatment; EORTC QLQ-C30; European Organization for Research and Treatment of Cancer Quality of Life Questionnaire; EQ-5D-5L: EuroQoL 5-dimensions 5-levels; irAE, immune-related adverse event; irCR: immune-related complete response; irDCR: immune-related disease control rate; irDOR: immune-related duration of response; irPD: immune-related progressive disease; irPFS: immune-related progression-free survival; irORR: immune-related objective response rate; (ir-)RECIST: (immune-related) Response Evaluation Criteria in Solid Tumours; NSCLC: non-small cell lung cancer; ORR: objective response rate; OS: overall survival; PD: progressive disease; PD1/PD-L1: programmed cell-death ligand 1; PFS: progression-free survival; POLE: polymerase ε; PRO: patient reported outcome; TILs: tumour-infiltrating lymphocytes. Source : GSK Data on File, GARNET clinical study report.[54]

B.2.3.1.2 Baseline characteristics

Baseline demographics and disease characteristics of the patients included in Cohort A1 of the GARNET trial are presented in Table 7. Two populations from the GARNET trial are relevant to this submission. The intention-to-treat (ITT) population (which is analogous with the safety analysis set) includes all patients that received at least one dose of dostarlimab, and informs the base case cost-effectiveness analysis in this submission. The efficacy population (at the time of interim analysis 2 [IA2]) (N=***) only includes patients who had measurable disease at baseline and who had at least 24 weeks follow-up to allow analysis of the response-related endpoints in GARNET (objective response rate [ORR], best overall response [BOR], disease control rate [DCR]), and is used in a sensitivity analysis. Further details of these analysis sets can be found in Section B.2.3.1. The ITT population represents the base case population in the economic analysis.

The mean age of patients in the ITT population of the GARNET trial was **** years, with the majority of patients being <65 years old (%) at the time of study entry. The majority of patients had an Eastern Cooperative Oncology Group Performance Status (ECOG PS) of 1 (%) and a most recent FIGO stage of IV (%). All of the patients had received at least one prior anti-cancer treatment, with the majority of patients (%) receiving exactly one prior line of anti-cancer therapy.

Please note that both populations summarised below included a very small minority of patients (N=* in the ITT population) who were classified as MMR-unknown, rather than dMMR. Patients with MMR-unknown (MMR-unk) were those whose MMR status was not tested in the trial; however, the expectation is that all patients would be tested for dMMR status in clinical practice.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 43 of 222

It is reasonable to assume that almost all of these patients would have tested positive for dMMR, had they been tested for dMMR, because they tested positive for MSI-H, which is the phenotypic presentation of dMMR.[36] This is evidenced by the fact that results remain similar when data from patients with MMR-unk but MSI-H tumours are pooled with those of subjects with dMMR tumours. Therefore, these populations (dMMR/MSI-H and MMR-unknown/MSI-H) are presented as one throughout the submission.

Table 7: Baseline demographics and disease characteristics of patients in GARNET

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 44 of 222

Footnotes:[a] Includes American Indian or Alaska Native.[b] Includes ‘Not reported’. Abbreviations: BMI: body mass index; ECOG PS: Eastern Cooperative Oncology Group Performance Status; FIGO: International Federation of Gynaecology and Obstetrics; IA2: interim analysis 2; ITT: intention-to-treat; STD: standard deviation. Source : GSK Data on File.[13]

B.2.3.1.3 Statistical analysis and definitions of study groups

The analysis of the primary endpoint ORR analyses in GARNET included summary statistics, including the number of patients (n) and percentage for categorical variables and the number of patients, mean, standard deviation (STD), median, minimum, and maximum for continuous variables. Two-sided exact 95% confidence interval (CIs) based on the Clopper-Pearson method were provided.

The duration of response (DOR) primary endpoint analyses were performed using Kaplan-Meier (KM) methods and summarised by minimum, maximum, 25th, 50th (median), and 75th percentiles with associated 95% CIs, the number and percentage of events, and number and percentage of censored observations.

The null hypothesis that the true response rate is ≤20% (H0: p≤0.2) was tested against a 1-sided alternative of ≥40% (Ha: p≥0.4). With 65 participants treated, Cohort A1 has 92% power to rule out a ≤20% ORR (null hypothesis; expected ORR for conventional therapy) when the true ORR is 40% at the 2.5% type I error rate (1-sided). Based on a recent report that 6 of 9 participants with MSI-H EC achieved a clinical response following treatment with an anti-PD-1 antibody, the activity of dostarlimab in this participant population is expected to negate the necessity for a 2-

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 45 of 222

stage design, and thus, there was no interim analysis in this cohort. Under protocol amendment 5 (10[th] May 2019), the sample size of Cohort A1 was increased to 100 participants, with the potential for up to 165 participants, which allows the lower-limit boundary of the exact 95% CI excluding a response rate of 25% or less and assuming the observed ORR is 35%.

An overview of the statistical analysis sets analysed in GARNET is provided in Table 8.

Table 8: Statistical analysis sets in GARNET

Footnotes:[a ] A smaller number of patients were included in the efficacy population at the first interim analysis (N=72), however, these data are not presented within this submission, in favour of the data for the larger efficacy population at IA2, and the ITT population, where applicable. Abbreviations: BICR: blinded independent central review; DCO: data cut off; IA2: interim analysis 2; irPFS: immune-related progression-free survival; ITT: intention-to-treat; RECIST v1.1: Response Evaluation Criteria in Solid Tumours version 1.1. Source: GSK Data on File.[54]

B.2.3.1.4 Quality assessment

A quality assessment of the GARNET trial was carried out using the Critical Appraisal Skills Program (CASP) Risk for Bias Tool for non-RCTs.[61] This checklist focuses on three broad issues: Are the results of the study valid? (Section A) What are the results? (Section B) Will the results help locally? (Section C). The 12 questions that make up the tool are designed to help the researcher think about these issues systematically.

Overall, due to the lack of randomisation and the single-arm nature of the trial, GARNET was found to have an unclear risk of bias. A summary of the quality assessment is provided below in Table 9.

Table 9: Risk of bias assessment of GARNET trial using the CASP risk of bias tool for non-RCTs

Risk of bias GARNET trial Are the results of the study valid?

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 46 of 222

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 47 of 222

Abbreviations: BICR: blinded independent central review; CASP: Critical Appraisal Skills Programme; CIs: confidence intervals; DCR: disease control rate; dMMR: mismatch repair deficient; DOR: duration of response; EC: endometrial cancer; MSI-H: microsatellite instability-high; ORR: overall response rate; OS: overall survival; PD: progressive disease; PFS: progression-free survival; RCT: randomised controlled trial; RECIST: Response Evaluation Criteria in Solid Tumours. Source: GSK Data on File.[13]

B.2.3.2 Current clinical management (UK RWE study)

To mitigate the impact of the paucity of data identified for comparator therapies in the clinical SLR (see Appendix D), a GSK-initiated RWE study was conducted to describe the characteristics, treatments and outcomes for patients diagnosed with recurrent or advanced EC in the UK.

The study identified a large population of patients (N=***), who were closely aligned with those in GARNET, and for whom detailed data on baseline characteristics, prognostic variables and survival outcomes were available. The UK RWE study provides a real-world representation of current clinical management in this difficult-to-treat patient population where there is a paucity of relevant data in the literature. Accordingly, this UK RWE study serves as the primary comparative efficacy evidence in this submission and informs the base case cost-effectiveness analysis. Full details of the UK RWE study are presented in Section B.2.3.2 and B.2.4.5. To explore the impact of any potential differences in patient populations, an ITC was conducted between GARNET and the UK RWE study for OS, and details of this analysis are presented in Section B.2.7.1.

The methodology of the UK RWE study is presented below.

B.2.3.2.1 Summary of study methodology

Study design

The UK RWE study used routine, linked patient-level UK health data available through the NCRAS, which combines linked data from the Hospital Episode Statistics (HES), SACT, National Radiotherapy Dataset (RTDS), Cancer Outcomes and Services Dataset (COSD) and Office for National Statistics (ONS) mortality data. Data were collected for patients diagnosed between 1st January 2013 and 31st December 2018, with data extraction until 30[th] September 2020.

A brief summary of the eligibility criteria of the study is provided in the sections below, with further details provided in Appendix O.1.

Initial inclusion and exclusion criteria to identify patients with EC

Initially, a series of inclusion and exclusion criteria were defined in order to narrow down the total number of patients available to only adult patients with EC for whom sufficient data were available (date of diagnosis, stage at diagnosis and age at diagnosis). These initial inclusion and exclusion criteria are detailed in Table 10.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 48 of 222

Table 10: Initial inclusion and exclusion criteria of the UK RWE study to identify patients with EC

Footnotes:[a ] Advanced disease was defined as patients who were FIGO Stage III/IV at diagnosis.[b] Probable recurrence was defined as patients who were FIGO Stage I/II and received surgery, systemic anti-cancer therapy or radiation therapy and then had a treatment gap greater than 90 days, followed by treatment with any treatment. This assumption was validated as reasonable by UK clinical expert opinion. Abbreviations: EC: endometrial cancer; FIGO: International Federation of Gynaecology and Obstetrics; RWE: real-world evidence. Source: GSK Data on File.[13]

Inclusion and exclusion criteria to identify patients with recurrent or advanced EC

In addition to the general inclusion and exclusion criteria outlined above, an additional series of inclusion and exclusion criteria were then applied to define a cohort of patients with recurrent or advanced EC, as detailed in Table 11.

Table 11: Inclusion and exclusion criteria of the UK RWE study to identify patients with recurrent or advanced EC

Abbreviations : EC: endometrial cancer; NA: not applicable. Source : GSK Data on File.[13]

Inclusion and exclusion criteria to identify patients for a GARNET-like cohort

The primary population considered in the comparative efficacy analysis (and the base case cost-

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 49 of 222

effectiveness analysis, described in Section B.3.2.1) is a smaller population of patients with recurrent or advanced EC who also fulfilled a further pre-defined set of eligibility criteria designed to identify a cohort of patients that were as closely matched as possible to the population of the GARNET trial. In order to be included in this ‘GARNET-like cohort’, patients additionally had to fulfil all of the inclusion and exclusion criteria listed in Table 12.

Table 12: Inclusion and exclusion criteria of the UK RWE study to identify patients for a GARNET-like cohort

Abbreviations : EC: endometrial cancer; ECOG PS: Eastern Cooperative Oncology Group Performance Status; PD-1: programmed cell death protein 1; PD-L1/L2: Programmed cell death ligand 1/2; RWE: real-world evidence. Source : GSK Data on File[13] .

An ECOG PS of ≤1 was a key inclusion criterion of the GARNET trial. However, patients with an ECOG PS of ‘not recorded (NR)’ were not excluded from the GARNET-like UK RWE cohort, in order to retain a larger sample size of patients, and to allow for a longer follow-up of data. It is likely that only a small minority of patients would have had an ECOG PS >1 if the PS of all patients had been known, given that the number of patients with an ECOG PS >1 only accounted for a small proportion of the overall RWE cohort of patients with recurrent or advanced EC (N=/*** [***%]). Furthermore, the NCRAS dataset does not provide details on why patients were classified with an ECOG PS of NR, and therefore excluding this group of patients could have introduced an unknown bias. However, in order to investigate the impact of not excluding patients with an ECOG PS of NR, a sensitivity analysis was conducted on a restricted cohort of patients, including only patients with a known ECOG PS of 0 or 1. This cohort is referred to as the ‘GARNET-like ECOG PS <1’ cohort; baseline characteristics and efficacy outcomes for this group of patients are detailed in Appendix O.2.

One limitation of the UK RWE study is that the dMMR/MSI-H biomarker is not recorded in the NCRAS database, and therefore, could not be used as an inclusion criterion for the UK RWE GARNET-like cohort. However, the impact of this is likely to be minimal. An SLR conducted by GSK found that there is no evidence that MSI-H or dMMR biomarker status has any prognostic or predictive value for efficacy and survival outcomes (including recurrence, relapse-free survival, PFS and OS) among patients with advanced or recurrent EC receiving non-anti-PD-(L)1 therapy.[62]

Patients with any evidence of receiving anti-PD-(L)1 therapy were excluded from the UK RWE GARNET-like population. As such, the survival outcomes for patients in the UK RWE study receiving current clinical management would not be expected to be significantly different for patients who were classified as dMMR/MSI-H, compared to the overall UK RWE GARNET-like population.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 50 of 222

Study outcomes

Treatment use

The proportion of patients that had previously received surgical resection, systemic anti-cancer therapy or radiation therapy was recorded. The first occurrence of a doublet platinum regimen was identified by the delivery of cisplatin or carboplatin in combination with one other chemotherapy drug, and the distribution of distinct regimens occurring before and after the first doublet platinum regimen was then reported.

Durations of lines of therapy were calculated as the difference in days between the start (the initiation of the earliest selected regimen within the line) and end dates (the last known cycle or administration within a line) of a derived line.

Following the identification of the first doublet platinum therapy administration, an algorithm was applied to capture changes in therapy and breakdown of systemic regimens by derived line of therapy.[13] The algorithm has been applied on other studies and has been found to be generalisable to a range of solid tumours.[14,15] The detailed algorithm can be found in Appendix O.1.

Within the lines of therapy derived for each patient, distinct regimens and drug classes were flagged by line of therapy and patient counts were reported. In line with GARNET, hormone therapy (where identified) did not count towards prior lines of therapy where recorded.

Clinical outcomes

The key clinical outcomes sought in the study are listed below:

• Baseline characteristics and patient demographics : A range of key characteristics were collected, including: age, ethnicity, ECOG PS, stage at diagnosis, histology at diagnosis and Charlson comorbidity score.

• OS from 2L: OS was defined as the time from the initiation of 2L therapy (i.e. index date equal to start date of 2L therapy) until failure (all-cause death). This output was necessarily restricted to patients with derived lines of therapy via data available through the SACT.

• TTNT from 2L: As progression is not recorded within the NCRAS database, time to next therapy (TTNT) was used as a proxy for PFS. TTNT is a common proxy endpoint for PFS and this approach was validated by UK clinical experts. TTNT was defined as the time from the start of line of therapy until failure (the earliest of all-cause death or the start of a new line of treatment). Patients lost to follow-up or still in same line of treatment at the end of the study period were censored. This output was necessarily restricted to patients with derived lines of therapy via data available through SACT.

• Time to treatment discontinuation (TTD) from 2L : Treatment discontinuation was defined as the first of death or the date of any drug administration that is followed by a gap of >90 days.

Patient numbers

A summary of the patient numbers in the UK RWE study is provided in Figure 8. The incident 2L GARNET-like patient cohort (n=***) captures patients with recurrent or advanced EC that are treated with their subsequent line of therapy following their first line of platinum-based doublet

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 51 of 222

chemotherapy. This is also the point in the treatment pathway where patients would first be eligible for treatment with dostarlimab. Thus, this is the cohort of patients relevant to the decision problem.

Figure 8: Patients included in the UK RWE study

==> picture [452 x 103] intentionally omitted <==

Abbreviations: 2L: second-line; EC: endometrial cancer; RWE: real-world evidence.

B.2.3.2.2 Baseline characteristics

Patient demographics and clinical characteristics

Patients in the GARNET-like RWE cohort had a mean age of **** years. The majority of patients had an ECOG PS status of NR at the time of registry diagnosis (****%); ****% and ****% had an ECOG PS status of 0 and 1, respectively. *** of the patients had received exactly one line of prior anti-cancer treatment following a diagnosis of recurrent or advanced EC, as outlined in the UK RWE study inclusion criteria.

Baseline patient demographics and clinical characteristics of patients in the GARNET-like UK RWE study cohort are presented in Table 13.

Table 13: Patient demographics and clinical characteristics in the GARNET-like UK RWE study cohort

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 52 of 222

Carcinosarcoma ******* Clear cell carcinoma ******** Dedifferentiated/Undifferentiated carcinoma ******* Endometrioid ********** Mesonephroma ******* Mixed carcinoma ******** Mucinous ******* Neuroendocrine ******* Non-specific ******* Non-specific carcinoma ******** Sarcoma ******* Serous ********** Squamous ***** FIGO stage at the time of registry diagnosis, n (%)** I ********** II ******** III ********** IV ********** Unknown *** Grade of disease at diagnosis, n (%)** Grade 1 ********** Grade 2 ********** Grade 3 ********** Grade 4 ******* Not assessable ******** Missing ******** Prior anticancer treatment, n (%)** Any prior anti-cancer treatment ********* Prior surgery ******** Number of prior lines of therapy post advanced/recurrent diagnosis, n (%)** 1 ********* 2 ******* 3 ******* ≥4 *******

Footnotes:[a] Includes American Indian or Alaska Native.[b] Includes Not reported. Abbreviations: ECOG PS: Eastern Cooperative Oncology Group Performance Status; FIGO: International Federation of Gynaecology and Obstetrics; ITT: intention-to-treat; RWE: real-world evidence; STD: standard deviation.

Source : GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 53 of 222

Most common chemotherapy regimens

Table 14 details the most common chemotherapy regimens received by patients with recurrent or advanced EC following their initial line of platinum-based chemotherapy derived from the UK RWE study. Overall, patients received a wide range of different regimens, with the most common regimens including carboplatin plus paclitaxel (%), paclitaxel monotherapy (%) and carboplatin plus PLD (****%).[13]

The wide range of different treatment regimens, including comparators outside of those included in the NICE final scope, demonstrates the clear lack of consensus around what which treatments represent the best options for the management of patients with recurrent or advanced EC who have progressed on or after a platinum-containing regimen in the UK.

Table 14: Most common chemotherapy regimens received by patients in the UK RWE study GARNET-like cohort

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 54 of 222

Carboplatin plus PLD plus * paclitaxel Niraparib * Topotecan *

Footnotes: Data for the GARNET-like ECOG ≤1 cohort can be found in Appendix O.2. Abbreviations: EC: endometrial cancer; PLD: pegylated liposomal doxorubicin; RWE: real-world evidence. Source: GSK Data on File.[13]

B.2.4 Clinical effectiveness results from GARNET and the UK RWE study

The following sections present the clinical effectiveness results from GARNET and the UK RWE study. A descriptive comparison of key outcomes between dostarlimab and current clinical management, which includes naïve comparisons between both evidence from the UK RWE study and the literature is also presented.

B.2.4.1 Baseline characteristics (GARNET versus UK RWE study)

The patient demographics and clinical characteristics of the UK RWE study GARNET-like cohort and the GARNET ITT population are presented in Table 15, demonstrating broad similarity between the two patient groups. Patients in the GARNET ITT population were slightly younger than patients in the GARNET-like RWE cohort (**** and **** years respectively).

A greater number of patients were diagnosed with FIGO stage III and IV EC in the GARNET-like UK RWE study cohort (****% and %, respectively) compared to patients in the GARNET ITT population (% and ****%, respectively). It is important to note that stage was recorded at the time of diagnosis in the UK RWE study, while the most recent stage at the time of study entry was recorded for patients in GARNET, which may account for this discrepancy.

More patients were diagnosed with ECOG PS 1 in the GARNET ITT population (% of patients) recorded compared to patients in the GARNET-like UK RWE study cohort (%). It is important to note however that ECOG PS is not directly comparable; ECOG PS was recorded at the time of diagnosis in the UK RWE study, while the most recent ECOG PS at the time of study entry was recorded for patients in GARNET, which may account for this discrepancy.

Table 15: Patient demographics and clinical characteristics in the GARNET-like UK RWE study cohort and ITT population of the GARNET trial

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 55 of 222

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 56 of 222

Footnotes:[a] Includes American Indian or Alaska Native.[b] Includes Not reported.[c ] For the RWE study this is at registry diagnosis. Abbreviations: ECOG PS: Eastern Cooperative Oncology Group Performance Status; FIGO: International Federation of Gynaecology and Obstetrics; ITT: intention-to-treat; RWE: real-world evidence; STD: standard deviation. Source : GSK Data on File.[13]

B.2.4.2 Objective response rate (ORR)

The primary efficacy endpoint results of GARNET, in terms of ORR and DOR, are available for the efficacy population (N=***) only. According to the study protocol, only patients with at least 24 weeks of tumour assessment were eligible for ORR and DOR analyses; all patients with at least 24 weeks of tumour assessment at the time of IA2 (Data cut off [DCO] 1[st] March 2020) were included in the efficacy population.

Other efficacy endpoints, including PFS and OS, were analysed for both the efficacy population (N=***) and the ITT population (N=129) and are presented from Section B.2.6.2 onwards.

B.2.4.2.1 Dostarlimab (GARNET)

*Dostarlimab demonstrated a robust and clinically meaningful ORR; % of patients () achieved a complete or partial response when treated with dostarlimab, with % of patients () experiencing a complete response (CR).*

In the efficacy population (N=***), the ORR (measured as the proportion of patients who achieved a BOR of CR or partial response [PR] to treatment), was ****% (n=**; 95% CI: **%, %). In total, ** patients (%) achieved a CR, and ** patients (**%) achieved a PR to treatment (Table 16).

A secondary efficacy endpoint in GARNET was DCR, which included patients achieving a BOR of CR or PR as well as patients with a BOR of stable disease (SD). A BOR of SD was observed in ** patients (****%), resulting in a DCR of ****% (n=**; 95% CI: ****%, ****%).

A detailed overview of the BOR to dostarlimab and the ORR in GARNET is presented in Table 16.

Table 16: Summary of the BOR to dostarlimab by RECIST v1.1 in GARNET (efficacy population) (BICR)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 57 of 222

Not done ***** Confirmed ORR by RECIST v1.1 n (%)[a]** ******* 95% CI[b]** ******** Response ongoing[c]** ******* DCR by RECIST v1.1, n (%)** ******* 95% CI[a]** **********

Footnotes:[a] ORR was defined as the percentage of patients with a RECIST v1.1-confirmed CR or PR. DCR was defined as the percentage of patients with a RECIST v1.1-confirmed PR, confirmed CR, or SD. Response assessments were based on BICR.[b] Exact 2-sided 95% CI for the binomial proportion.[c] All responders who have not yet died or progressed (including clinical progression); the denominator for the percentage is the number of responders. Abbreviations: BICR: blinded independent central review; BOR: best overall response; CI: confidence interval; CR: complete response; DCR: disease control rate; dMMR: mismatch repair-deficient; EC: endometrial cancer; MSI-H: microsatellite instability-high; NE: not evaluable; ORR: objective response rate; PD: progressive disease; PR: partial response; RECIST: Response Evaluation Criteria in Solid Tumours; SD: stable disease. Source: GSK Data on File.[13]

B.2.4.2.2 Dostarlimab versus current clinical management

ORR data were not collected in the UK RWE study for current clinical management. However, the response rates observed in GARNET are striking when compared to studies identified from the clinical SLR (detailed in Appendix D). ESMO guidelines highlight that for patients with EC recurring after first-line chemotherapy, only paclitaxel has consistently shown a response rate >20 (less than half the ORR achieved by dostarlimab).[47] Notably, the studies reporting ORRs for paclitaxel, such as Lincoln et al. (2003), which report an ORR for paclitaxel monotherapy of 27.3%, predate the use of paclitaxel as a first-line treatment for patients with recurrent or advanced EC.[49] It is therefore unlikely that similarly high ORRs would be observed for paclitaxel monotherapy in current clinical practice.

Accordingly, in more recent studies, McMeekin et al. (2015)[6] reports an ORR of 15.7% for patients receiving either paclitaxel or doxorubicin monotherapy (N=223); there were no patients that experienced a CR. For patients receiving doxorubicin monotherapy (N=225), the ZoptEC study reported an even lower ORR of 14.1%, with only 2.0% of patients experiencing a CR. The results for the comparator arm of the recently conducted KEYNOTE 775 trial which included doxorubicin or paclitaxel monotherapy showed similar response rates, with an ORR of 14.7% and only 2.6% of patients experiencing a CR.[8-10, 63]

There were only two relevant studies including patients treated with carboplatin plus paclitaxel in the clinical SLR. Both of these were small, retrospective studies: Mazgani et al. (2008) (N=31) and Rubinstein et al. (2019) (N=20), which reported similar ORRs to GARNET, of 38.7% and 50.0%, respectively.[59, 60] However, these studies report only very limited information on patient characteristics, meaning that it is completely unknown whether these patients are similar to those in GARNET. In the absence of more detailed information from these studies, it is not possible to make any robust comparisons between dostarlimab and carboplatin plus paclitaxel with any certainty.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 58 of 222

B.2.4.3 Duration of response (DOR)

B.2.4.3.1 Dostarlimab (GARNET)

**The response to dostarlimab is durable – after a median follow-up of **** months, % of patients () were still experiencing an ongoing response at the IA2 DCO (1[st] March 2020). The probability of maintaining a response until Month 12 and Month 18 was estimated at ****% and ****%, respectively.**

A summary of the DOR in GARNET is presented in Figure 9. At the time of IA2 (DCO 1[st] March 2020), % of patients () treated with dostarlimab were still in response, after a median follow-up of **** months. The median DOR was *********** (range: *** to ***** months, where + indicates response is still ongoing) (Figure 9).

The probability of maintaining a response until Month 6, Month 12 and Month 18 was: ****% (95% CI: ****, ****), ****% (95% CI: ****, ****), and ****% (95% CI: ****, ****), respectively (calculated using KM estimation) (Table 17).

In comparison, only one study identified in the clinical SLR reported the median DOR; patients treated with paclitaxel monotherapy in Lincoln et al. (2003) experienced a median DOR of just 4.2 months.[49]

Figure 9: DOR (from time of first PR or CR) based on RECIST v1.1 in GARNET (efficacy population) (BICR)

==> picture [450 x 300] intentionally omitted <==

Abbreviations: BICR: blinded independent central review; CR: complete response; dMMR: mismatch repair deficient; EC: endometrial cancer; PD: progressive disease; PR: partial response; RECIST v1.1: Response Evaluation Criteria in Solid Tumours version 1.1; SD: stable disease. Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 59 of 222

Table 17: KM analysis of DOR based on BICR in GARNET (efficacy population; patients with objective response)

Patients in the efficacy population with DOR an objective response (N=) DOR status, n (%) Events observed** ****** Censored** ******* Median duration of follow-up (months)** ** DOR (months) Min, Max** ***********[] Quartile (95% CI) 25% ********** 50%* ********* 75%** ********* Duration ≥6 months, n (%)** ******* DOR distribution function (95% CI) Month 6** *************** Month 12** *************** Month 18* *****************

Footnotes:[a] 95% CIs were generated using the method of Brookmeyer and Crowley (1982).[64][b ] A “+” indicates that the patient’s response is ongoing. Abbreviations: BICR: blinded independent central review; CI: confidence interval; dMMR: mismatch repair-deficient; DOR: duration of response; EC: endometrial cancer; KM: Kaplan-Meier; max: maximum; min: minimum; MSI-H: microsatellite instability high; NE: not evaluable; RECIST: Response Evaluation Criteria in Solid Tumours.

Source: GSK Data on File.[13]

B.2.4.3.2 Dostarlimab versus current clinical management

DOR data were not collected in the UK RWE study for current clinical management. The only study in the clinical SLR reporting DOR was Lincoln et al . (2003), which reported a median DOR of 4.2 months for patients receiving paclitaxel monotherapy (N=44).[49]

B.2.4.4 Change in target lesion size

B.2.4.4.1 Dostarlimab (GARNET)

The ******** of patients who experienced a response to treatment with dostarlimab experienced more than a ************* in tumour size, in comparison to baseline.

A waterfall plot of the by-patient maximum percentage change in tumour size by RECIST v1.1 for the efficacy population in GARNET is presented in Figure 10. The maximum percentage change in target lesions from baseline is indicated by bar length. The waterfall plot demonstrates that the majority of patients who experienced a response to treatment with dostarlimab experienced a clinically meaningful **** reduction in tumour size in comparison to baseline.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 60 of 222

Figure 10: Waterfall plot of the maximum percentage change in target lesions compared with baseline measurements based on BICR per RECIST v1.1 in GARNET (efficacy population)

==> picture [452 x 202] intentionally omitted <==

Footnotes: Best change in target lesion size is the maximum reduction from baseline or the minimum increase from baseline in the absence of a reduction. Horizontal reference ranges are defined by -30 for PR. Abbreviations: BICR: blinded independent central review; CR: complete response; dMMR: mismatch repairdeficient; EC: endometrial cancer; NE: not evaluable; PD: disease progression; PR: partial response; RECIST v1.1: Response Evaluation Criteria in Solid Tumours version 1.1; SD: stable disease. Source: GSK Data on File.[54]

B.2.4.4.2 Dostarlimab versus current clinical management

Changes in target lesion size data were not collected in the UK RWE study for current clinical management, or reported in any of the studies included for comparators in the clinical SLR. It is therefore not possible to make any comparisons in terms of changes in target lesion size between dostarlimab and current clinical management.

B.2.4.5 Progression-free survival (PFS)

B.2.4.5.1 Dostarlimab (GARNET)

**Overall, ****% and ****% of patients in the ITT population were progression free at Month 12 and Month 24, respectively. The**

**********************************************************************************************, and the PFS**

curve subsequently plateaued, suggesting the potential for a long-term PFS benefit with dostarlimab.

In the ITT population of GARNET (N=129), ** PFS events were observed, with

*********************************************************************************************************** .

Following this, a clear plateau was observed. At Month 6, **% of patients in GARNET had not experienced disease progression or death. Very few patients then experienced disease progression or death in the next six months, with ****% of patients in GARNET remaining progression-free at Month 12. Patients treated with dostarlimab continued to experience a longterm PFS benefit through to Month 18 and Month 24, with ****% and ****% of patients remaining free of disease progression of death.

Due to the plateauing of the PFS KM curve, when approximately **% of patients had experienced

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 61 of 222

disease progression or death, the median PFS of *** months is highly uncertain, and associated *** **** with very wide CIs (95% CI: , ). As such, the median PFS must be interpreted with caution, and

**************************************************************. This is illustrated in Section B.2.7.1, where rounding the individual PFS estimates for each patient to one decimal place (versus two decimal places) changes the median PFS from *** months to *** months.

However, ****************************************, the overall shape of the KM curve and the presence of a plateau suggests patients who remain progression-free may experience a longterm PFS benefit from dostarlimab treatment through to Month 18 and Month 24, with ****% and ****% of patients remaining free of disease progression of death.

A similar trend was observed in the efficacy population (N=***). The majority of the ** PFS events occurred within the first six months, with ****% of patients remaining progression-free at Month 6. A similar plateau was then observed, with ****% of patients remaining progression-free at Month 12 and ****% at Month 24.

It is also important to note that because dostarlimab is an I-O therapy, the conventional RECIST criteria may not be an adequate measure to monitor response rates and disease progression; successful treatment response following I-O therapies manifests differently, including delayed response, transient tumour enlargement followed by shrinkage, stable size, or initial presence of new lesions followed by stability or response.[65, 66] Immune-related PFS (irPFS; presented in Section B.2.4.7) may therefore suggest that the PFS results presented below, according to the traditional RECIST v1.1, may actually underestimate the true PFS benefit of dostarlimab.

Table 18 presents further details of PFS across both the efficacy and ITT populations, while PFS KM curves are presented in Figure 11.

Table 18: KM analysis of PFS from GARNET (efficacy population and ITT population) (BICR)

Footnotes:[a ] PFS was defined as the time from the first dose of treatment to disease progression or death, whichever occurred first.[b ] Landmark survival estimates at Month 18 and Month 24 were taken from the KM data included in the dostarlimab cost-effectiveness model, and as such, associated confidence intervals are not

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 62 of 222

available.

Abbreviations: BICR: blinded independent central review; CI: confidence interval; EC: endometrial cancer; ITT: intention-to-treat; KM: Kaplan-Meier; NR: not reachable; PFS: progression-free survival. Source: GSK Data on File.[13]

Figure 11: PFS KM curves from GARNET (efficacy population and ITT population) (BICR)

==> picture [440 x 286] intentionally omitted <==

Abbreviations: BICR: Blinded Independent Central Review; ITT: intention-to-treat; KM: Kaplan-Meier; MSI-H: microsatellite instability-high; PFS: progression-free survival. Source: GSK Data on File.[13]

B.2.4.5.2 Dostarlimab versus current clinical management

UK RWE study

**Dostarlimab markedly improves PFS versus current clinical management; only ****% and ****% of a population of GARNET-like patients in the UK RWE study were progression-free at Month 12 and 24, compared with ****% and ****% in the GARNET ITT population, respectively.**

Considering a naïve comparison, patients receiving dostarlimab had a substantially reduced risk of disease progression or death, relative to patients receiving current clinical management based on the UK RWE study (which used TTNT as a proxy for PFS).

In the UK RWE study ‘GARNET-like’ cohort, ****% of patients had not experienced disease progression or death at Month 6, dropping drastically to ****% of patients who remained progression-free at Month 12 (Figure 12). In stark contrast, **% of patients treated with dostarlimab in the GARNET trial were progression-free at *******; notably the PFS curve then plateaued. Very few patients treated with dostarlimab experienced disease progression or death over the next six months, with ****% of patients remaining progression-free at Month 12.

Patients treated with dostarlimab in the GARNET trial then experienced a sustained, long-term PFS benefit through to Month 24, with ****% of patients remaining progression-free. However,

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 63 of 222

there is almost no evidence for a sustained PFS benefit for patients receiving current clinical management in the UK RWE study ‘GARNET-like’ cohort, with only a small minority of patients (****%) remaining progression-free at Month 24.[13]

Notably, these results likely overestimate the true PFS for patients in the UK RWE study ‘GARNET-like’ cohort. The UK RWE study used TTNT as a proxy for PFS, and in reality, it is likely that patients would experience a delay between disease progression and the initiation of their next line of treatment. As such, the true PFS for these patients is likely to be lower than the estimates based on TTNT.[13]

The naïve PFS comparison between GARNET and the UK RWE study is summarised in Table 19.

Figure 12: TTNT as a proxy for PFS for patients with recurrent or advanced EC that have progressed on or after platinum-based chemotherapy receiving current clinical management (UK RWE GARNET-like cohort)

==> picture [433 x 299] intentionally omitted <==

Footnotes: PFS for patients in the GARNET-like ECOG PS ≤1 cohort of the UK RWE study is detailed in Appendix O.2.

Abbreviations: CI: confidence interval; EC: endometrial cancer; PFS: progression-free survival; RWE: real-world evidence; TNTT: time to next treatment. Source: GSK Data on File.[13]

Table 19: Naïve PFS comparison for patients with recurrent or advanced EC that have progressed on or after platinum-based chemotherapy (UK RWE study versus GARNET)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 64 of 222

Footnotes: TTNT was used as a proxy for PFS in the UK RWE study. PFS for patients in the GARNET-like ECOG PS ≤1 cohort of the UK RWE study is detailed in Appendix O.2.[a] Landmark survival estimates at Month 18 and Month 24 for GARNET were taken from the KM data included in the dostarlimab cost-effectiveness model, and as such, associated confidence intervals are not available Abbreviations: CI: confidence interval; EC: endometrial cancer; ITT: intention-to-treat; PFS: progression-free survival; RWE: real-world evidence. Source: GSK Data on File.[13]

Published literature

The sustained benefit in terms of PFS for patients treated with dostarlimab is a critical difference versus published PFS curves for the comparator chemotherapies identified in the clinical SLR. While other studies report comparable or increased median PFS estimates versus GARNET, there is almost no evidence of a plateau or long-term PFS benefit in these studies. Patients treated with doxorubicin monotherapy in the ZoptEC study had a median PFS of 4.7 months (95% CI: 4.1, 6.6), but very few patients were progression-free by Month 24.[8-10] Rubinstein et al. (2019) and Mazgani et al. (2008) report higher median PFS estimates of 10 months (95% CI: 2.0, 47.0) and 8 months (95% CI: 5.0, 13.0), respectively, for patients receiving carboplatin plus paclitaxel.[59, 60] However, 5.2% and 14.7% of patients in these studies were progression-free at Month 24, compared to ****% of patients in GARNET.

B.2.4.6 Overall survival (OS)

B.2.4.6.1 Dostarlimab (GARNET)

*A remarkable % of patients were alive at Month 12 after treatment with dostarlimab. Patients experienced a sustained survival benefit – ************ of the patients in GARNET (%) were still alive by Month 24.*

In the ITT population of GARNET (N=129), ** OS events were observed at the time of the IA2 DCO (1[st] March 2020). Patients treated with dostarlimab experienced a clear OS benefit relative to patients receiving current clinical management in UK clinical practice. At Month 12, a remarkable ****% of patients were still alive. While the OS data are still immature, a clear plateau is observed from approximately ******** (Figure 13), with ****% of patients still alive at Month 18, and ****% of patients still alive at Month 24.

The median OS has ******************** at the time of the IA2 DCO. However, it is reasonable to assume that the true median OS estimate is, at a minimum, equal to the lower confidence bound of **** months.

The efficacy population showed a similar trend: ** OS events were observed, with ****% of patients alive at Month 12, and ****% of patients alive at Month 24. The median OS was also *************** at the time of the IA2 DCO.

Table 20 presents further details of OS across both the efficacy and ITT populations, while OS

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 65 of 222

KM curves are presented in Figure 13.

Table 20: KM analysis of OS from GARNET (efficacy population and ITT population)

Footnotes:[a] OS was defined as the time from the date of the first dose of study treatment to the date of death by any cause.[b] 95% CIs were generated using the method of Brookmeyer and Crowley (1982).[64 c] Landmark survival estimates at Month 18 and Month 24 were taken from the KM data included in the dostarlimab costeffectiveness model, and as such, associated confidence intervals are not available. Abbreviations: CI: confidence interval; KM: Kaplan-Meier; NR: not reached; OS: overall survival. Source: GSK Data on File.[13]

Figure 13: OS KM curves from GARNET (efficacy population and ITT population)

==> picture [444 x 288] intentionally omitted <==

Abbreviations: KM: Kaplan-Meier; ITT: intention-to-treat; OS: overall survival. Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 66 of 222

B.2.4.6.2 Dostarlimab versus current clinical management

UK RWE study

Only one in five (%) GARNET-like patients initiating further chemotherapy in the UK RWE cohort were still alive at Month 24 – ************** the percentage of patients treated with dostarlimab in GARNET who were still alive (%) at Month 24

The naïve OS comparison between GARNET and the UK RWE study demonstrates that patients treated with dostarlimab experienced a marked reduction to the risk of death versus patients treated with current clinical management in the UK.

In the UK RWE study ‘GARNET-like’ cohort, ****% of patients were still alive at Month 12; by Month 24, just ****% were still alive (Figure 14). The comparison with GARNET is clear and conclusive: ****% of patients in GARNET were still alive at Month 12, and ****% of patients were still alive at Month 24, **************** the percentage of patients still alive that received current clinical management, highlighting the dire prognosis faced by patients in current clinical practice.

Figure 14: OS for patients with recurrent or advanced EC that has progressed on or after platinum-based chemotherapy receiving current clinical management (UK RWE GARNETlike cohort)

==> picture [442 x 302] intentionally omitted <==

Footnotes: OS for patients in the GARNET-like ECOG PS ≤1 cohort of the UK RWE study is detailed in Appendix O.2. Abbreviations: CI: confidence interval; EC: endometrial cancer; OS: overall survival; RWE: real-world evidence. Source: GSK Data on File.[13]

The naïve OS comparison between GARNET and the UK RWE study is summarised in Table 21.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 67 of 222

Table 21: Naïve OS comparison for patients with recurrent or advanced EC that has progressed on or after platinum-based chemotherapy (UK RWE study versus GARNET)

Footnotes: OS for patients in the GARNET-like ECOG PS ≤1 cohort of the UK RWE study is detailed in Appendix O.2.[a ] Landmark survival estimates at Month 18 and Month 24 for GARNET were taken from the KM data included in the dostarlimab cost-effectiveness model, and as such, associated confidence intervals are not available. Abbreviations: CI: confidence interval; EC: endometrial cancer; ITT: intention-to-treat; OS: overall survival; RWE: real-world evidence. Source: GSK Data on File.[13]

Published literature

Of the studies identified in the clinical SLR, patients receiving doxorubicin monotherapy in the ZoptEC study had a median OS of 10.8 months (95% CI: 9.8, 12.6), with 23.0% of patients alive at Month 24. Similarly, patients receiving paclitaxel or doxorubicin monotherapy in McMeekin et al. (2015) had a median OS of 12.3 months (95 CI: 10.7, 15.4), with just 29.4% of patients alive at Month 24.[6, 8-10] For patients treated with carboplatin plus paclitaxel, Mazgani et al. (2008) reported a median OS estimate of 15.0 months (95% CI: 9.1–30.4) (patients with an endometrioid histology), with 35.5% of patients alive at Month 24. Rubinstein et al. (2019) reported a median OS of 27.0 months (95% CI: 6.0, 117.0), with 59.5% of patients alive at Month 24.[59, 60]

B.2.4.7 Immune-related endpoints (dostarlimab only)

The primary analyses of response rates and PFS in GARNET were based on the conventional RECIST v1.1 criteria, which measure response by the reduction in tumour size following chemotherapy. However, these criteria are primarily designed to measure disease-progression following traditional cytotoxic chemotherapy. In contrast, successful treatment response following I-O therapies manifests differently, including delayed response, transient tumour enlargement followed by shrinkage, stable size, or initial presence of new lesions followed by stability or response.[65] Consequently, the conventional RECIST criteria may not be an adequate measure in monitoring response to these unique therapies.[65, 66 ]

The modified irRECIST provides an alternative measurement of endpoints that is able to more reliably account for the tumour response to I-O therapies. The primary difference between this measure and the conventional criteria is the introduction of an additional follow-up to confirm or withdraw ‘unconfirmed’ tumour progression after an initial increase in size.[66 ]

Consequently, the immune-related endpoints presented below may reflect a more representative assessment of the true benefit of dostarlimab. It is important to note that irRECIST was evaluated

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 68 of 222

by Investigator Assessment in GARNET, while the primary response rate analyses presented above were evaluated by BICR.

Immune-related progression-free survival (irPFS) data are presented below, while immunerelated response rate endpoints are presented in Appendix N.2.

Immune-related progression-free survival (irPFS)

The irPFS results demonstrate a very similar trend to PFS assessed by conventional RECIST. A proportion of patients experienced disease progression very early (within the

************************** of treatment), but then the survival curve begins to level off and plateau ******* from approximately . Additional details on irPFS are presented in Table 22, and the irPFS KM curve is presented in Figure 15.

In comparison to PFS assessed by conventional RECIST, a greater percentage of patients were alive at Month 6 (%) and Month 12 (%) by irRECIST, compared to the number of patients who were progression free in the efficacy population at the same time points (****% and ****%). By the time the PFS curve completely plateaus, a remarkable ****% of patients were progression-free at Month 24, slightly lower than the ****% of patients who were progression-free by conventional RECIST at the same point. Despite the slight differences in these results, it is clear that regardless of how PFS is assessed, patients experience a clear and sustained benefit following treatment with dostarlimab.

**** *** **** Notably, the median irPFS is much higher, at months (95% CI: , ), versus the median PFS estimate of *** months (95% CI: ***, ****) for the efficacy population per conventional RECIST. This is a drastic difference with similar numbers of PFS events observed between the two populations (** versus **, respectively) and only small differences in the percentages of patients progression-free at Month 6 and Month 12.

Alongside the wide CIs associated with the median PFS estimates, this drastic difference in PFS serves to further highlight that, because of the plateau observed when approximately ****of patients are progression-free, the median PFS is heavily influenced by just one or two patients who experienced a PFS event in this interim data cut across different populations, and therefore median PFS estimates (for PFS per conventional or irRECIST) do not represent a robust summary statistic to estimate the PFS benefit associated with dostarlimab.

Table 22: irPFS per irRECIST KM analysis in GARNET (immune-related efficacy population) (Investigator Assessment)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 69 of 222

Month 9 *************** Month 12** *************** Month 18[c]** ** Month 24[c]** ****

Footnotes:[a] irPFS was defined as the time from the date of the first dose to the earlier date of assessment of irPD event or death by any cause in the absence of disease progression based on the time of irPD event per irRECIST.[b] CIs were generated using the method of Brookmeyer and Crowley (1982).[64 c ] Landmark survival estimates at Month 18 and Month 24 were taken from the KM curve and as such, associated confidence intervals are not available.

Abbreviations: CI: confidence interval; irPFS: immune-related progression-free-survival; irRECIST: immunerelated Response Evaluation Criteria in Solid Tumors; KM: Kaplan-Meier; NR: not reachable. Source: GSK Data on File.[13]

Figure 15: irPFS per irRECIST KM curve in GARNET (immune-related efficacy population) (Investigator Assessment)

==> picture [425 x 272] intentionally omitted <==

Footnotes: Medians presented in months. Abbreviations: CI: confidence interval; dMMR: DNA mismatch repair deficiency; EC: endometrial cancer; irPFS: immune-related progression-free survival; irPFS: immune-related Response Evaluation Criteria in Solid Tumors; KM: Kaplan-Meier; NR: not reported. Source: GSK Data on File.[54]

B.2.4.8 Health-related quality of life (dostarlimab only)

Treatment with dostarlimab preserved patient-reported HRQoL from baseline. Key disease-related symptom subscales, such as pain and fatigue showed a positive trend of improvement throughout the study.

European Quality of Life (EQ) Visual Analogue Scale (VAS)

The EQ-5D-5L is a generic HRQoL questionnaire comprising of five dimensions of health: mobility, self-care, usual activities, pain/discomfort and anxiety/depression.[67] Each dimension has five levels of severity: no problems, slight problems, moderate problems, severe problems and extreme problems.[67] Tariffs are anchored at 1 for full health and 0 for health states

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 70 of 222

considered equivalent to death.[67] A summary of patient responses to each of the EQ-5D-5L subscales at each timepoint in the GARNET trial is presented in Appendix N.3.

The EQ-VAS assesses patients perceived overall health status on the day of scoring on a scale from 0 (worst imaginable health state) to 100 (best imaginable health state). The adjusted mean change from baseline in EQ-VAS is presented in Figure 16. At baseline, the mean EQ-VAS score was **** (STD: ****). Whilst the level of change fluctuated throughout the study duration, a general trend of improvement in EQ-VAS score was observed. At Week 12, the mean EQ-VAS **** *** **** score was (STD: 18.0), demonstrating a mean improvement from baseline of (STD: ). **** **** At Week 18, improvements in EQ-VAS score continued, with a mean score of (STD: ), demonstrating a mean improvement from baseline of *** (STD: ****). At Week 42, the improvement remained the same, with a change from baseline of 4.0 (STD: ****). Whilst a greater improvement in score was observed after end of treatment, the number of subjects at each visit was notably low.

Figure 16: Adjusted Mean Change from Baseline in EQ-VAS (GARNET efficacy population)

==> picture [435 x 218] intentionally omitted <==

Footnotes: Adjusted mean and 95% CI are from mixed model repeated measures (MMRM) with week visit, and Eastern Cooperative Oncology Group (ECOG) status as factors and baseline score as continuous covariate as well as an unstructured covariance structure.

Abbreviations: EOT: end-of-treatment; EQ-VAS: EuroQol-visual analogue scale; PRO: patient reported outcome; SUVF: survival follow-up; WX: week X. Source: GSK Data on File.[13]

European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ-C30)

The EORTC QLQ-C30 is an internationally validated HRQoL questionnaire for cancer. It contains five scales for functioning (physical, social, role, cognitive, and emotional functioning), eight symptom scales (fatigue, nausea/vomiting, pain, dyspnoea, sleep disturbances, appetite loss, constipation, and diarrhoea), financial impact, and an assessment for overall QoL. For the functioning scales and global QOL higher scores indicate better functioning; for the symptom scales, higher scores indicate increased symptom burden.[68]

EORTC QLQ-C30 data were available for ** of the 129 patients in the ITT population of GARNET. The completion rate for the EORTC QLQ-C30 was consistent across domains, ranging from ***% at baseline to ****% at Cycle 7.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 71 of 222

The global health/quality of life (QoL) change over time from Baseline and the mean scores over time demonstrated a clear improvement.

Figure 17: Mean change in Global Health Status/QOL from Baseline (GARNET ITT population)

==> picture [402 x 219] intentionally omitted <==

Abbreviations: ITT: intention-to-treat; QoL: quality of life. Source: GSK Data on File.[13]

Overall, the EORTC QLQ-C30 results showed that patients treated with dostarlimab reported that key disease-related symptom subscales, including pain and fatigue, improved or remained stable over time while on treatment. Both patient-reported pain and fatigue symptoms showed a downwards trend of improvement below baseline (reduced pain and fatigue symptoms), starting at Cycle 2 and Cycle 3, respectively.

Patients also reported an improvement in physical functioning over time while on treatment, with a positive trend of improvement (increased physical functioning) above baseline from Cycle 4, which then remained stable thereafter (Figure 18).

Figure 18: Pain, fatigue and physical functioning mean change from baseline (GARNET ITT population)

==> picture [457 x 158] intentionally omitted <==

Abbreviations : CxDx: Cycle X Day X; ITT: intention-to-treat. Source : GSK Data on File.[13]

For patients who experienced symptomatic adverse events (AEs) included in the symptom

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 72 of 222

scales of the EORTC QLQ-C30, including nausea, vomiting, constipation, diarrhoea, or tiredness, the majority remained stable or had improvement in these symptoms over the treatment course compared to their baseline. Only a minority of patients reported worsening in these AE symptoms including ***% who reported single-category worsening and **% who reported 2- or 3-category worsening.

More detailed analyses of the change in symptomatic AEs in response from baseline are presented in Figure 19.

Figure 19: Symptomatic AE change in response from baseline (GARNET ITT population)

==> picture [422 x 337] intentionally omitted <==

Abbreviations : AE: adverse event; CxDx: Cycle X Day X; ITT: intention-to-treat. Source : GSK Data on File.[13]

B.2.5 Subgroup analysis (dostarlimab only)

Dostarlimab (GARNET)

Subgroup analyses for ORR in the efficacy population are presented in Figure 20. With the exception of the subgroup analysis by ECOG performance status, all categories of the subgroup analyses using this data cutoff date show overlapping 95% CIs with the overall population ORR. With the exception of the *** patients for whom the most recent FIGO stage was unknown, all of the point estimates for ORR were ***** which suggests that the treatment benefit of dostarlimab was observed across all subgroups.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 73 of 222

Figure 20: Forest plot of ORR (CR or PR) and 95% CI by subgroup by RECIST v1.1 in GARNET (efficacy population (N=) (BICR)*

==> picture [469 x 555] intentionally omitted <==

Abbreviations: BOR: best overall response; BICR: blinded independent central review; CI: confidence interval; CPS: combined positive score; CR: complete response; dMMR: ECOG: Eastern Cooperative Oncology Group; FIGO: International Federation of Gynaecology and Obstetrics; IHC: immunohistochemistry; PD: progressed disease; PD-L1: programmed cell death ligand; PR: partial response; sBLA: supplemental Biologics License Application; SD: stable disease.

Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 74 of 222

B.2.6 Meta-analysis

GARNET represents the only trial for dostarlimab. As such, no pooling of trials was undertaken, and this section is not applicable to this submission.

B.2.7 Indirect and mixed treatment comparisons

Two different approaches were employed in order to more closely match the comparative data in this submission (the UK RWE study and the published literature) and the data for dostarlimab in GARNET:

• Dostarlimab versus current clinical management (UK RWE study MAIC versus GARNET): The UK RWE study is included as the primary comparative efficacy evidence in this submission, given the large sample size of the study (N=***), together with the close alignment to patient characteristics in the GARNET trial and the real-world representation of current clinical management in this difficult to treat population. While the patient populations in GARNET and the GARNET-like cohort of the UK RWE study were closely aligned, a MAIC was conducted between GARNET and the UK RWE study for OS in order to investigate the impact of any remaining differences between the two populations. The methodology and results of this MAIC are presented in Section B.2.7.1.

• Supportive comparative evidence for dostarlimab versus individual chemotherapy regimens: Whilst the UK RWE study serves as the primary comparative efficacy evidence in this submission, a series of ITCs have also been conducted, where possible, between dostarlimab and the individual chemotherapy comparators listed in the NICE final scope, based on the published studies identified in the clinical SLR. These comparisons include a series of MAICs between GARNET and trials for carboplatin plus paclitaxel, paclitaxel monotherapy and doxorubicin monotherapy, as well as an inverse probability treatment weighting (IPTW) ITC between GARNET and the ZoptEC trial.[8-10] Given the limitations associated with the data identified in the literature, including the extremely limited data on patient characteristics and prognostic variables, and the resulting uncertainty associated with some of these analyses, they are provided for completeness as supportive comparative efficacy evidence only. The methodology and results of these analyses are presented in Section B.2.7.2.

A summary of the comparative efficacy evidence analyses considered in this submission is presented in Table 5 in Section B.2.2.

B.2.7.1 Dostarlimab versus current clinical management (UK RWE MAIC versus GARNET)

Overview

The methodology of the UK RWE study is detailed in Section B.2.3. The following section provides an overview of the MAIC between the patients in the ITT population of the GARNET trial and patients in the GARNET-like cohort of the UK RWE study. Additional details are presented in Appendix D.5.1.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 75 of 222

Choice of MAICs

Due to the single-arm nature of the GARNET trial, and the UK RWE study, an unanchored MAIC was considered to represent the most appropriate and robust method for indirect comparison, in line with NICE TSD 18.[69]

The MAIC approach was preferred to simulated treatment comparisons (STCs) because MAICs produce marginal treatment effect estimates.[70] MAICs are conducted based on assigning differential weights to IPD available for the intervention (dostarlimab) which is similar to running logistic regression. An important part of any adjusted treatment comparison involves the identification of relevant prognostic variables and treatment effect modifiers. When these weights are applied, the aggregate measures on the modelled prognostic and treatment effect variables equal (or are as close as possible to) the values in the matched aggregate studies. This weighting approach produces a marginal (population level) treatment effect and therefore allows for a population-level indirect treatment comparison, which is an advantage over STCs, which produce only conditional (patient-level) treatment effects.

In addition, the key endpoint for this comparison (OS) is a survival endpoint. For survival endpoints, it is the number of OS events that occur, combined with the overall number of patients (rather than just the overall number of patients alone), that determines the effective sample size (and therefore, degrees of freedom), in any type of regression analysis. Since imbalances in covariates are accounted for using the weighting approach, covariates need not be adjusted for when modelling the outcome, meaning that the degrees of freedom is much closer to the number of patients once the analysis starts, maximising the number of prognostic and treatment effect modifying variables that can be considered, relative to using an STC approach.

MAIC methodology

The primary endpoint analysis considered in the UK RWE study MAIC utilised a Cox proportional hazards model, using weights obtained using the MAIC method, in order to estimate a HR for OS for patients receiving dostarlimab in GARNET versus patients receiving current clinical management in the GARNET-like cohort of the UK RWE study. A Cox proportional hazards model was considered feasible because the definition of OS was considered to be closely matched between GARNET and the UK RWE study.

However, it was not considered feasible to use a Cox proportional hazards model for PFS between GARNET and the UK RWE study, because PFS was not recorded in the UK RWE study (i.e. the NCRAS database does not include any data on progression, remission or recurrence of disease), and while TTNT was considered to be a suitable proxy, the measurement definitions and time-period evaluations associated with TTNT in the RWE population were considered to be too dissimilar to those for PFS in GARNET. Thus, TTNT in the RWE study was descriptively compared to PFS in the GARNET trial based on landmark PFS estimates at various timepoints, using the weighted GARNET IPD (following matching with the RWE population) versus the RWE GARNET-like cohort.

Identification of matching variables

A targeted literature review was conducted in May 2020 to identify a range of prognostic variables typically associated with survival in EC (detailed in Appendix M). The list of prognostic variables was subsequently validated with a panel of clinical experts from the UK, Germany and Canada. The clinical experts indicated that all of the prognostic variables identified would also

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 76 of 222

represent treatment effect modifying variables.

Based on the list of variables identified, the following variables were reported in the UK RWE study and considered for inclusion as matching variables (detailed in the following sections):

• Race/ethnicity (black, others, unknown versus white)

• Age category (≥65 years versus <65 years)

• ECOG PS status at treatment initiation (1 versus 0)

• Histology at initial diagnosis (non-endometrioid, unknown versus endometrioid)

• FIGO stage at initial diagnosis (Stage III/IV versus Stage I/II)

• Grade of disease at diagnosis (Grade 3/4, unknown versus Grade 1/2)

• Number of prior platinum-based therapies (0 or 1 versus ≥2)

• Prior surgery for study indication (yes versus no)

Based on this list, the modification and prognostic value of each potential matching variable was investigated using a Cox proportional hazard model. In this model, the outcome variable (OS or PFS) was modelled as a function of each variable of interest. Cox regression models were fit separately for the GARNET data and the RWE data. Patient characteristics that exhibited association at level of significance p≤0.1 in at least one of the two datasets were considered prognostic.

Based on these results, two scenarios were constructed, based on the prognostic variables identified by clinical expert opinion (Scenario 1) and the matching variables found to be statistically significant based on regression analyses (Scenario 2).

Grade was not found to be statistically significant by the regression analysis, but was identified as a prognostic variable by clinical experts. Unfortunately, grade was challenging to include in the adjustment because a large number of patients in the UK RWE GARNET-like cohort (****%) had an ‘unknown grade’, compared to ***% in the GARNET cohort.

Clinical experts confirmed that grade information was unlikely to be missing from the UK RWE study at random, and therefore, the potential for an underlying difference in this ****% of patients relative to the remaining proportion of patients in the UK RWE GARNET-like cohort means that the use of grade as a matching variable is associated with substantial uncertainty. If grade is included in the analysis, it results in a requirement to heavily up-weight the few patients in the GARNET cohort with an unknown grade (as shown in Figure 21), which drastically reduces the ESS to N=** (i.e., a **% reduction in the ESS compared to the original sample of N=129), leading to unreliable results.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 77 of 222

Figure 21: Histogram of the weights assigned to patients in the GARNET ITT population in the MAIC versus the UK RWE GARNET-like population using the matching variables in Scenario 1 in addition to grade

==> picture [201 x 163] intentionally omitted <==

Abbreviations : ITT: intention-to-treat; MAIC: matching adjusted indirect comparison; RWE: real-world evidence; UK: United Kingdom.

As a result, grade was excluded from the list of matching variables considered in Scenario 1, which included histology (non-endometrioid and unknown versus endometrioid) and the number of lines of prior platinum-based therapy in the advanced/recurrent setting (0 or 1 versus ≥2) as matching variables, based on clinical expert opinion. Scenario 1 was considered robust, with a relatively large sample size of N=** (a reduction of **% compared to the original sample of N=129), more than twice the effective sample size compared to also including grade as a matching variable.

Scenario 2 considered the matching variables identified as statistically significant by regression analysis (further details of the regression analysis are presented Appendix D.5.1). For the UK RWE GARNET-like cohort, these variables consisted of: race/ethnicity (black, others, unknown versus white), stage at diagnosis (Stage III/IV versus Stage I/II), histology (non-endometrioid, unknown versus endometrioid) and prior surgery (yes versus no). The ESS in Scenario 2 was highly comparable to Scenario 1; Scenario 2 resulted in an ESS of N=** (a **% reduction compared to the original sample of N=129).

The two scenarios and the matching variables considered in each, based on clinical expert opinion (Scenario 1) and the regression analysis (Scenario 2), are detailed in Table 84. Further details on prognostic matching, including the regression analyses, characteristics of the matched populations for each scenario, histograms of the weighting of the GARNET cohorts for each scenario, the process for determination of treatment effect modifiers and assessments of proportional hazard can be found in Appendix D.5.1.

Table 23: Scenarios considered in the UK RWE study MAICs versus GARNET

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 78 of 222

Footnotes:[a] For scenarios including histology as a matching variable, one patient with an “unknown” histology was removed from the GARNET cohort in order to achieve balance.[b] For scenarios including the number of prior platinum-based therapies, patients with 0 or ≥2 prior platinum-based therapies from the GARNET cohort were removed in order to achieve balance.

Abbreviations : MAIC: matching-adjusted indirect comparison; RWE: real-world evidence.

Results

Overall survival

The results of both matching scenarios showed that the median OS, as well as the percentage of patients alive at Month 6, 12 and 18, was greater for patients treated with dostarlimab versus patients treated with current clinical management. The HRs for OS (dostarlimab versus current clinical management) showed that dostarlimab statistically significantly reduced the risk of death versus current clinical management in both scenarios.

Scenario 1 and Scenario 2 both found that a slightly improved OS, compared to the unadjusted GARNET ITT population. At Month 12, **% of patients were alive in the unadjusted ITT population, compared to **% of patients in scenario 1, and **% of patients in scenario 2. The results of the unadjusted population and scenario 1 were similar at Month 18, with **% and **% of patients still alive, while scenario 2 found that **% of patients were still alive at Month 18. The OS HR between dostarlimab and current clinical management is **** (95% CI: ****, ****) when **** **** **** including the unadjusted GARNET ITT population, compared to (95% CI: , ) in Scenario 1 and **** (95% CI: ****, ****) in Scenario 2.

The similarity of the OS results between the unadjusted GARNET ITT population and both matched scenarios indicate that there were minimal differences between the GARNET ITT population and the UK RWE GARNET-like population, and suggests that any differences between the two populations could mean that the OS benefit of dostarlimab is slightly underestimated in the unadjusted comparisons presented previously in this submission.

OS for the UK RWE GARNET-like cohort and the GARNET ITT population before and after matching (Scenario 1 and 2) is presented in Table 24. KM curves for PFS prior to adjustment of the GARNET population are presented in Figure 22, and KM curves for PFS including the adjusted GARNET populations are presented in Figure 23 (Scenario 1), and Figure 24 (Scenario 2), respectively.

Table 24: OS for patients in the GARNET ITT population (before and after matching) and the UK RWE GARNET-like cohort

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 79 of 222

==> picture [452 x 107] intentionally omitted <==

----- Start of picture text -----
OS rate at 18 **** **** ************************** ***************************
months (95% CI) ************ ************
Hazard ratio for
OS (95% CI) for
dostarlimab ** **** *************************** ***************************
versus current ************
clinical
management
----- End of picture text -----

Abbreviations: CI: confidence interval; ESS: effective sample size; ITT: intention-to-treat; NA: not applicable; NR: not reached; OS: overall survival; RWE: real-world evidence.

Figure 22: OS KM curves – dostarlimab (unadjusted GARNET ITT population, N=129) versus current clinical management (UK RWE GARNET-like cohort, N=)*

==> picture [336 x 252] intentionally omitted <==

Abbreviations: ITT: intention-to-treat; KM: Kaplan-Meier; OS: overall survival; RWE: real-world evidence.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 80 of 222

Figure 23: OS KM curves – dostarlimab (adjusted GARNET population, Scenario 1, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=***)**

==> picture [338 x 230] intentionally omitted <==

Abbreviations: ESS: effective sample size; ITT: intention-to-treat; KM: Kaplan-Meier; OS: overall survival; RWE: real world evidence.

Figure 24: OS KM curves – dostarlimab (adjusted GARNET population, Scenario 2, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=***)**

==> picture [336 x 231] intentionally omitted <==

Abbreviations : ESS: effective sample size; ITT: intention-to-treat; KM: Kaplan-Meier; OS: overall survival; RWE: real world evidence.

Progression-free survival

Comparison of PFS for the adjusted GARNET cohorts in Scenario 1 and Scenario 2 suggested that, once the GARNET ITT population and the GARNET-like RWE population are more closely matched, dostarlimab provides a slightly greater PFS benefit versus current clinical management, when compared with the unadjusted comparison between the two. However, the generally similar PFS results indicate that the two populations were closely matched prior to adjustment, providing confidence in the unadjusted comparisons described in Section B.2.4.5.2.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 81 of 222

The median PFS for patients receiving dostarlimab increases in Scenario 1 and Scenario 2, to **** *** ** **** *** ** months (95% CI: , ) and months (95% CI: , ), respectively, compared to the median PFS of ****months (95% CI: ***, **) in the GARNET ITT population.

The landmark estimates show that the percentage of patients treated with dostarlimab who are progression-free are largely similar before and after adjustment. At Month 12, **% of patients treated with dostarlimab were progression-free in the unadjusted population, compared to **% in Scenario 1 and **% in Scenario 2. By Month 18, **% of patients were progression-free in the unadjusted population, compared to **% in Scenario 1 and **% in Scenario 2.

PFS for the UK RWE GARNET-like cohort and the GARNET ITT population before and after matching (Scenario 1 and 2) is presented in Table 25. KM curves for PFS prior to adjustment of the GARNET population are presented in Figure 25, and KM curves for PFS including the adjusted GARNET populations are presented in Figure 26 (Scenario 1), and Figure 27 (Scenario 2), respectively.

The median PFS of *** (95% CI: ***, **) months for the unadjusted GARNET ITT population in *** *** **** Table 25 is different to the median PFS for the GARNET ITT population of (95% CI: , ) months presented in Section B.2.4.5, due to rounding of individual patient PFS times to one decimal place for the MAIC versus the UK RWE study, but were not rounded in the PFS analysis of the GARNET ITT cohort presented in Section B.2.4.5.

Due to the plateauing of the PFS curve for patients treated with dostarlimab when approximately ****of patients are progression-free,

**************************************************************************** , ************************************************************* (as previously detailed in Section B.2.4.7) *************************************************************************** . Rounding of individual patient PFS estimates results in the KM curve staying above **% until *** months, versus *** months when the individual patient PFS estimates are not rounded.

Table 25: PFS for patients in the GARNET ITT population (before and after matching) and the UK RWE GARNET-like cohort

==> picture [452 x 217] intentionally omitted <==

----- Start of picture text -----
UK RWE GARNET Adjusted GARNET Adjusted GARNET
GARNET- ITT population (Scenario population (Scenario 2)
like population 1)
cohort prior to
(N=**) [a] matching
(N=129)
ESS *** *** **
Median PFS, *** *** []
*********************** **************************
months (95% **********
CI)
PFS rate at 6 **** ****
*************
months (95% ************ ************ ************
CI)
PFS rate at **** ****
12 months ************ ************ ***************************** ***************************
(95% CI)
----- End of picture text -----

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 82 of 222

PFS rate at **** **** *** 18 months** ************ ************ **************************** ********** (95% CI)**

Footnotes:[a ] The results presented for the UK RWE GARNET-like cohort use TTNT as a proxy for PFS, as PFS was not recorded in the NCRAS database.[b] The median PFS for GARNET presented here is different to the median PFS preented for the GARNET ITT population in Section B.2.4.5 due to rounding of individual patient PFS estimates in the analysis presented above, which means the KM curve stays above 50% until *** months. Abbreviations: CI: confidence interval; ESS: effective sample size; ITT: intention-to-treat; NCRAS: National Cancer Registry Analysis System; NR: not reached; PFS: progression free survival; RWE: real-world evidence; TTNT: time to next treatment.

Source: GSK Data on File.[13]

Figure 25: PFS KM curves – dostarlimab (unadjusted GARNET ITT population, N=) versus current clinical management (UK RWE GARNET-like cohort, N=)

==> picture [384 x 286] intentionally omitted <==

Abbreviations : ITT: intention-to-treat; KM: Kaplan-Meier; PFS: progression-free survival; RWE: real-world evidence.

Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 83 of 222

Figure 26: PFS KM curves – dostarlimab (adjusted GARNET population, Scenario 1, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=***)**

==> picture [361 x 243] intentionally omitted <==

Abbreviations : ESS: effective sample size; ITT: intention-to-treat; KM: Kaplan-Meier; PFS: progression-free survival; RWE: real world evidence.

Source: GSK Data on File.[13]

Figure 27: PFS KM curves – dostarlimab (adjusted GARNET population, Scenario 2, ESS N=) versus current clinical management (UK RWE GARNET-like cohort, N=***)**

==> picture [379 x 252] intentionally omitted <==

Abbreviations: ESS: effective sample size; ITT: intention-to-treat; KM: Kaplan-Meier; PFS: progression-free survival; RWE: real-world evidence. Source: GSK Data on File.[13]

Additional results

Additional results, including the patient characteristics of the unadjusted and adjusted populations, the histograms of weightings for each adjusted population, and the assessment of

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 84 of 222

proportional hazards, are presented in Appendix D.5.1. Furthermore, the results of the MAIC between GARNET and the UK RWE GARNET-like ECOG PS ≤1 population are presented as a sensitivity analysis in Appendix D.5.1.

Conclusions

The results of the UK RWE ITC present clear supportive evidence that patients treated with dostarlimab experience a significant and substantially decreased risk of disease progression or death compared to patients receiving current clinical management for the treatment of recurrent or advanced EC that has progressed on or following prior treatment with a platinum-containing regimen in the UK.

The similarity of the PFS and OS results for the unadjusted GARNET ITT cohort and the adjusted GARNET cohorts in Scenario 1 (matching based on the most important prognostic variables according to UK clinical expert feedback, excluding grade) and Scenario 2 (matching based on prognostic variables as identified by regression analyses) suggests that the two populations are closely matched, with minimal differences with respect to key prognostic variables. Moreover, the results show that the remaining differences between the GARNET ITT population and the UK RWE GARNET-like cohort may actually result in an underestimation of the true PFS and OS benefit that dostarlimab provides versus current clinical management; the **** **** unadjusted OS HR between dostarlimab and current clinical management was (95% CI: , ****), compared to adjusted OS HRs of **** (95% CI: ****, ****) in Scenario 1 and **** (95% CI: **** **** , ) in Scenario 2, respectively.

Based on the strengths of the UK RWE study, and the results of the above comparison, the unadjusted UK RWE GARNET-like cohort is used in the base case-cost effectiveness analysis, while scenarios are considered based on the adjusted scenarios presented in this section (as described in B.3.8.3).

B.2.7.2 Supportive comparative evidence for dostarlimab versus individual chemotherapy regimens

B.2.7.2.1 ITC between dostarlimab (GARNET) and doxorubicin (ZoptEC study)

Overview

As detailed above and in Appendix D, during the data extraction phase of the clinical SLR, it became apparent that there was a paucity of data for the comparator chemotherapy studies, which would impact the robustness of any ITCs. As such, GSK contacted the corresponding authors of each of the relevant chemotherapy studies identified from the SLR in order to request further data.

Following this, Aeterna Zentaris, the sponsoring company of the ZoptEC study provided IPD for the ZoptEC study.[8-10] Using this IPD, an adjusted comparison of OS between dostarlimab and doxorubicin monotherapy was conducted. The OS comparison was performed using a Cox proportional hazards model with stabilised inverse probability of treatment weighting (IPTW) to estimate an HR for OS between dostarlimab and doxorubicin. The IPTW approach minimises the standardised differences between the baseline characteristics of two populations, and allows for two separate populations to be compared with as little bias as possible. This approach is aligned with NICE TSD 17.[71]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 85 of 222

It was not possible to use IPTW to estimate a HR for PFS between dostarlimab and doxorubicin, due to differences in the definition of PFS and the timepoints of tumour assessments between GARNET and ZoptEC.[8-10] PFS was defined from the date of the first dose of dostarlimab in GARNET, but defined as the time elapsed from randomisation in ZoptEC.[8-10] Patients were assessed every six weeks for disease progression starting from Week 12 in GARNET; conversely, patients were re-evaluated for response every nine weeks in ZoptEC.[8-10] A summary of the PFS definitions in both studies is presented in Appendix D.5.2. Due to the differences in PFS between the two studies, a descriptive-only KM analysis was conducted to compare PFS between GARNET and ZoptEC.[8-10]

Whilst the results of this ITC provide a comparison versus only one of the relevant chemotherapy comparators (doxorubicin monotherapy), and the UK RWE study detailed in Section B.2.7.1 represents the primary comparative efficacy evidence in this submission, this ITC was nevertheless still conducted to provide an alternative analysis for consideration within a scenario analysis in the economic analysis (see Section B.3.8.3).

A summary of the methodology is provided below; the full methodology can be found in Appendix D.5.2.

Application of exclusion criteria to GARNET and ZoptEC

Prior to the statistical analysis, it was necessary to consider the inclusion/exclusion criteria and baseline characteristics of the GARNET and ZoptEC studies, and to apply a series of additional exclusion criteria to each trial in order to match the two populations as closely as possible.

Patients were excluded from the ZoptEC trial if they had a follow-up greater than 36 months, or if they did not have an ECOG PS score of 0 or 1.[8-10] Patients were excluded from the GARNET trial if they had previously received more than one prior platinum-based therapy.

The exclusion of these patients reduced the patient populations to N=** patients in GARNET and N=*** patients in ZoptEC.[8-10] These populations are known as the main analysis sets used throughout this ITC. A summary of the exclusion criteria applied and the patients excluded at each step is detailed in Appendix D.5.2.

IPTW ITC methodology

As detailed in Section B.2.7.1, a targeted literature review was conducted in May 2020 to identify a range of prognostic variables typically associated with survival in EC (detailed in Appendix M). The list of prognostic variables was subsequently validated with a panel of clinical experts from the UK, Germany and Canada. The clinical experts indicated that all of the prognostic variables identified would also represent treatment effect modifying variables.

Of the prognostic variables identified, the following variables were reported in the ZoptEC study and used for estimating stabilised-IPTW in this analysis:[8-10]

• Age (<65 years versus >65 years)

• Race (non-white versus white)

• ECOG PS score (0 or 1 versus 2)

• Histology (endometrioid versus non-endometrioid)

• Most recent FIGO stage at Baseline (Stage I/II versus Stage III/IV)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 86 of 222

• Prior surgery (no versus yes)

While grade of disease was also reported in ZoptEC, grade could not be used for estimating stabilised IPTW because it causes a violation of the positivity assumption, as detailed in Appendix D.5.2.

IPTW ITC results

Overall survival

The results of the IPTW ITC versus doxorubicin monotherapy based on the ZoptEC study are presented in B.2.7.2, and the adjusted KM curves are presented in Figure 28 and summarised in Table 27.[8-10] The comparison showed that treatment with dostarlimab resulted in a significant and marked reduction in the risk of death versus doxorubicin monotherapy. Patients treated with dostarlimab were **% less likely to die at any given timepoint compared to patients receiving ***** ***** ***** ******* doxorubicin monotherapy (HR: ; 95% CI: ( , ); p ).

Table 26: Results for the safety analysis data set on OS with adjusting stabilised-IPTW

Abbreviations : CI: confidence interval; HR: hazard ratio; IPTW: inverse probability treatment weighting; OS: overall survival; PH: proportional hazards. Source: GSK Data on File.[13]

Figure 28: OS KM curves – dostarlimab (adjusted GARNET main analysis set, N=) versus doxorubicin monotherapy (adjusted ZoptEC main analysis set, N=**** following adjustments based on IPTW[8-10]**

==> picture [359 x 270] intentionally omitted <==

Footnotes : The number at risk with IPTW adjustment may differ slightly from the total sample size. This is because the number at risk has been weighted by IPTW. The IPTW weighted number at risk may not be an

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 87 of 222

integer and in the KM plots the weighted IPTW number at risk has been rounded to the nearest integer value. Abbreviations: KM: Kaplan-Meier; OS: overall survival; TRT: treatment. Source: GSK Data on File.[13]

Table 27: Summary of OS with adjusting stabilised-IPTW for the main analysis data set

Abbreviations : CI: confidence interval; IPTW: inverse probability treatment weighting; NR: not reached; OS: overall survival. Source: GSK Data on File.[13]

Additional results including the patient characteristics for both populations before and after matching, checking of effect modifiers and assessments of proportional hazards and unmeasured confounding, are presented in Appendix D.5.2.

Overall survival (sensitivity analysis)

In the primary IPTW analysis, ** patients were removed from the analysis across GARNET and ZoptEC to ensure comparability between the two populations.[8-10] A sensitivity analysis was conducted where ** of these ** patients, who were excluded in the main analysis (detailed in *** Appendix D.5.2) are included, resulting in a total sample size of N= , with N=129 patients in GARNET and N=*** patients in ZoptEC (it was still necessary to exclude *** patient in ZoptEC because they did not have a baseline ECOG PS, so could not be included).[8-10]

The results of the sensitivity analysis showed that patients treated with dostarlimab were **% less likely to die at any given timepoint compared to patients receiving doxorubicin monotherapy (HR: *****; 95% CI: *****, ; p<*). The similarity between the two point estimates in the main analysis (HR: ; 95% CI: (, ); p**). and the sensitivity analysis (HR: *****; 95% CI: *****, ; p<*). The provides confidence that treatment with dostarlimab results in approximately a **% reduced risk of death compared to doxorubicin monotherapy.

Full details for this sensitivity analysis are presented in Appendix D.5.2.

Progression-free survival

As outlined previously, it was not possible to conduct an IPTW ITC for PFS, due to differences in the definitions of PFS and the timepoints of tumour assessments between GARNET and ZoptEC, as detailed in Section B.2.7.2.1 and Appendix D.5.2.[8-10] PFS was defined from the date of the first dose of dostarlimab in GARNET, but defined as the time elapsed from randomisation in ZoptEC.[8-10] Patients were assessed every six weeks for disease progression starting from Week 12 in GARNET; conversely, patients were re-evaluated for response every nine weeks in ZoptEC.[8-10] A summary of the PFS definitions in both studies is presented in Appendix D.5.2.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 88 of 222

However, a supportive comparative analysis was conducted, investigating PFS for dostarlimab versus doxorubicin once the additional exclusion criteria outlined previously had been applied to match the populations of GARNET and ZoptEC more closely. The KM curves for PFS for dostarlimab (GARNET main analysis set, N=) and doxorubicin (ZoptEC main analysis set, N=*) are presented in B.2.7.2 below, and a summary of the PFS from the two studies is presented in B.2.7.2.

The results showed that dostarlimab provided a significant PFS benefit for patients compared to doxorubicin monotherapy after initial adjustment between the two studies. A clear plateau in the dostarlimab PFS curve can be observed in Figure 29 when approximately **% of patients were progression-free. Conversely, there was no such plateau was observed for patients treated with doxorubicin monotherapy; almost all of the patients treated with doxorubicin had experienced disease progression or death prior to Month 24.

Figure 29: PFS KM curves – dostarlimab (GARNET main analysis set, N=) versus doxorubicin (ZoptEC main analysis set, N=***)[8-10]**

==> picture [339 x 255] intentionally omitted <==

Abbreviations: KM: Kaplan-Meier; PFS: progression-free survival; TRT: treatment. Source: GSK Data on File.[13]

Table 28: Summary of PFS for the main analysis data set

Footnote: This analysis does not consider the differences in tumour assessment schedules between GARNET and ZoptEC detailed in Section B.2.7.2.1 and Appendix D.5.2.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 89 of 222

Abbreviations : CI: confidence interval; IPTW: inverse probability treatment weighting; NR: not reached; PFS: progression-free survival. Source: GSK Data on File.[13]

B.2.7.2.2 MAICs versus carboplatin plus paclitaxel, paclitaxel monotherapy and doxorubicin monotherapy

Overview

As IPD were available for the ZoptEC study identified in the clinical SLR, a separate ITC was conducted between GARNET and ZoptEC, as described above.[8-10] The remaining relevant studies identified in the clinical SLR were reviewed for inclusion in a series of MAICs between dostarlimab and the individual chemotherapy comparators listed in the NICE final scope. Given the significant limitations associated with the MAICs presented below, the results of these MAICs provide supportive comparative efficacy evidence for this submission only.

Feasibility assessment

Of the 13 studies included in the clinical SLR, only studies including the individual chemotherapy regimens listed in the NICE final scope were considered for inclusion within the MAIC feasibility assessment:

• Re-challenge with carboplatin plus paclitaxel

• Paclitaxel monotherapy

• Doxorubicin monotherapy

• Carboplatin monotherapy

An overview of the trials identified in the clinical SLR and the details of the trials that were included and excluded from the MAICs is presented in Appendix D.5.3.

As part of the feasibility assessment, eight studies included in the clinical SLR were excluded from the series of MAICs. Three studies were excluded because the study regimen was not listed as a relevant comparator in the NICE final scope, and two studies were excluded because they did not report KM curves for OS. Other reasons for exclusion were that the study considered a population that was considered too different to GARNET with respect to race for a comparison to be feasible (N=1), and that the study reported data by platinum-free interval which was not reported as a covariate in other studies (N=1). Finally, GSK were able to obtain IPD for the ZoptEC study, and so the ZoptEC study was excluded from the series of MAICs, but a separate ITC versus ZoptEC was considered as described above in Section B.B.2.7.2.1, in line with NICE TSD 17 and 18.[8-10, 69, 71]

MAIC methodology

A summary of the MAIC methodology is provided below. The detailed methodology underlying the MAICs, including the programming code used, is presented in Appendix D.5.3.

Choice of MAICs

MAICs were chosen as the most appropriate and robust method for the indirect comparisons with the studies identified in the published literature, in line with NICE TSD 18[69] and the reasons outlined in Section B.2.7.1, considering the single-arm nature of the GARNET trial and four of the five trials included in the series of MAICs, and the consideration of survival outcomes (PFS and

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 90 of 222

OS).

Matching of the study populations

The first step in the conduct of the MAICs involved the removal of patients from the base GARNET dataset that would not have met the matched study inclusion/exclusion criteria and baseline characteristics range criteria in each of the comparator studies (where data identification was feasible). A summary of the patients removed from the base GARNET population for each of the MAICs is presented in Appendix D.5.3.

Further to this, the unanchored MAIC methodology was applied to the reduced set of GARNET patients, in line with the approach described in NICE TSD 18.[69] The process considers two inputs: the mean values of the prognostic/treatment effect modifying variables in the nonGARNET study, and the individual matching data from GARNET. The methodology then utilises a method of moments analytical technique (as listed in NICE TSD 18[69] ) to produce individual patient specific weights (for GARNET patients), that when applied, produce weighted prognostic means that approximately equal those inputted from the aggregate study.

As previously detailed, a targeted literature review was conducted in May 2020 to identify a range of prognostic variables typically associated with survival in EC (detailed in Appendix M). The list of prognostic variables was subsequently validated with a panel of clinical experts from the UK, Germany and Canada. The clinical experts indicated that all of the prognostic variables identified would also represent treatment effect modifying variables.

Based on the list of variables identified, the following variables were reported in the published studies included in the series of MAICs, and were included as matching variables:

• Age

• Race

• Number of prior anti-cancer treatments

• Histology (endometrioid type I only versus others)

• Prior surgery for the study indication (yes versus no)

• ECOG PS score (before comparator treatment start date)

• Most recent FIGO stage (before comparator treatment start date)

Grade was also identified as an important prognostic variable by clinical experts, however none of the studies identified in the clinical SLR reported sufficient data on grade (the only study to report any information on grade was Makker et al. (2013), which reported extremely limited data). The MAIC weights were then applied to the GARNET PFS/OS data using an appropriate modelling process to produce contrast estimates against the comparator in the comparator study. The KM curves of the comparator studies were digitised (using Engauge Digitizer 10.2 software[72] ), and the algorithm detailed in Guyot et al. (2012) was utilised to produce pseudoindividual patient data for PFS/OS from the aggregate trial data.[73] A weighted Cox regression was then applied to this dataset, combining the pseudo-IPD with the GARNET data (no covariates other than treatment). A weight of one was assigned to each patient from the pseudoIPD of the comparator study; the MAIC-calculated weight was assigned to each patient in GARNET.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 91 of 222

Full details on the MAIC methodology, including details of the assessment of proportional hazards and quantitative bias analyses, are provided in Appendix D.5.3.

MAIC results

A summary of the MAIC results are presented in this section; additional results, as well as quantitative bias analyses and proportional hazard assessments associated with each of the MAICs are presented in Appendix D.5.3.

Study characteristics

A total of five studies were included in the series of MAICs in addition to GARNET (B.2.7.2). McMeekin et al. (2015)[6] was a Phase III, open-label trial which compared one group of patients that received either paclitaxel monotherapy or doxorubicin monotherapy to a group of patients receiving ixabepilone.[6] Of the remaining four trials, three were retrospective studies investigating carboplatin plus paclitaxel (Rubinstein et al. [2019]) and doxorubicin monotherapy (Julius et al. [2013] and Makker et al. [2013]).[7, 11, 59] The final study, Mazgani et al. (2008), was a cohort study investigating carboplatin plus paclitaxel.[60]

Patient characteristics were poorly reported across the comparator studies. Of the most important prognostic variables identified above, none of the comparator studies reported information on the number of prior anti-cancer treatments, prior surgery for the study indication, grade or MMR/MSI molecular profile type. In particular, the paucity of data on prior anti-cancer treatments is a key limitation, as this was included alongside histology, grade and ECOG PS status as the most important prognostic variables based on clinical expert opinion.

Prognostic variables were particularly limited in the Julius et al. (2019), Rubinstein et al. (2019) and Mazgani et al. (2008) studies.[7, 59, 60] None of these studies reported any information on ECOG PS, while only Makker et al. (2013) reported extremely limited data on grade. Data on histology, race and age were also poorly reported (detailed in B.2.7.2). Given the extremely limited data available, as well as small sample sizes, the MAICs versus these studies must be interpreted with particular caution, given the substantial uncertainty and unknown potential for bias.

Of the two studies reporting more comprehensive details on prognostic variables, it is clear that patients in McMeekin et al. (2015)[6] are most closely aligned with those is GARNET. Notably, both GARNET and McMeekin et al. (2015) excluded patients with an ECOG PS ≥ 1, although GARNET included slightly more patients with PS 1 versus McMeekin et al. (2015)[6] (**% versus 33% of patients with an ECOG PS of 1).

However, the inclusion of 12% of patients with an ECOG PS of 2 represents a key limitation of the MAICs versus Makker et al. (2013).[11] There were ** patients with an ECOG PS of 2 in GARNET, and therefore the IPD of GARNET cannot be adjusted to account for this imbalance between the two studies, likely resulting in a slight bias in favour of dostarlimab.

A summary of the patient characteristics reported in each study, as well as additional details about each of the studies is presented in Appendix D.5.3.

Progression-free survival

PFS data were only available from three of the five studies that were included from the MAIC feasibility assessment, allowing MAICs to be conducted versus doxorubicin monotherapy and

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 92 of 222

carboplatin plus paclitaxel only.

An overview of the MAIC results for PFS is provided in Table 29. PFS KM curves for dostarlimab (pre- and post-MAIC weighting) versus each of the comparators are provided in Figure 30 (dostarlimab versus doxorubicin monotherapy) as well as Figure 31 and Figure 32 (dostarlimab versus carboplatin plus paclitaxel), respectively.

Compared with doxorubicin monotherapy, results from the MAIC versus Makker et al. (2013) demonstrate that dostarlimab significantly reduces the risk of disease progression versus doxorubicin monotherapy.[11] Of note, patients treated with dostarlimab were more than five times less likely to experience disease progression or death (HR: *****; 95% CI: *****, ; p*) versus patients treated with doxorubicin monotherapy based on evidence derived from Makker et al. (2013).[11] Even when considering the upper-most CIs, patients treated with dostarlimab still ***** experienced more than a three times reduced risk of disease progression or death (HR: ). These results clearly indicate that dostarlimab provides a PFS benefit versus doxorubicin monotherapy, even if the magnitude of the PFS benefit is uncertain, given the inherent limitations with Makker et al. (2013).[11]

The results of the PFS MAICs versus Rubinstein et al. (2019) and Mazgani et al. (2008) are more uncertain, and indicate that no significant differences in PFS between patients treated with dostarlimab and patients treated with carboplatin plus paclitaxel.[59, 60]

However, in addition to the wide CIs which introduce uncertainty, there are a number of limitations associated with these MAICs, including the limited sample sizes, paucity of patient characteristic data, while Mazgani et al. (2008) based tumour assessment on RECIST v1, compared to GARNET which used RECIST v1.1.[60] As such, it is difficult to draw any meaningful conclusions about the PFS benefit of dostarlimab relative to carboplatin plus paclitaxel based on the results of these MAICs.

Table 29: PFS hazard ratios derived from the MAICs

Footnotes:[a ] Patients in the McMeekin et al. (2015)[6] study received either paclitaxel or doxorubicin. Clinical expert opinion indicated that the efficacy between the two treatments is likely to be similar, and therefore it is appropriate to consider this as one combined arm.[b ] McMeekin et al. (2015) and Julius et al. (2013) did not report PFS KM curves, and therefore MAICs could not be conducted for PFS.[6, 7] Abbreviations: CI: confidence interval; HR: hazard ratio; MAIC: matching-adjusted indirect comparison; NA: not applicable; PLD: pegylated liposomal doxorubicin. Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 93 of 222

Figure 30: MAIC PFS KM curves for dostarlimab based on GARNET versus doxorubicin based on Makker et al. (2013)[11]

==> picture [452 x 180] intentionally omitted <==

Abbreviations: KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; PFS: progression-free survival. Source: GSK Data on File.[13]

Figure 31: MAIC PFS KM curves for dostarlimab based on GARNET versus carboplatin plus paclitaxel based on Rubinstein et al. (2019)[59]

==> picture [454 x 181] intentionally omitted <==

Footnotes: It is likely that the MAIC between dostarlimab and carboplatin plus paclitaxel based on Rubinstein et al. (2019) violates the proportional hazards assumption (further detailed in Appendix D.5.3), although it is difficult to say this definitively due to the small number of patients include in Rubinstein et al. (2019).[59] Nonetheless, there are no viable alternatives to the proportional hazards assumption, as Rubinstein et al. (2019) does not report sufficient data to be able to estimate time varying hazard ratios.[59] Abbreviations: KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; PFS: progression-free survival. Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 94 of 222

Figure 32: MAIC PFS KM curves for dostarlimab based on GARNET versus carboplatin plus paclitaxel based on Mazgani et al. (2008)[60]

==> picture [452 x 180] intentionally omitted <==

Abbreviations : KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; PFS: progression-free survival.

Source : GSK Data on File.[13]

Overall survival

The primary results of the MAICs between dostarlimab and individual studies of the relevant chemotherapy comparators listed in the NICE final scope are presented in Table 30.

The KM plots of OS for the pre-MAIC unweighted and post-MAIC weighted dostarlimab cohorts and the comparator in each of the MAICs are presented in Figure 33 to Figure 37.

It is clear that patients treated with dostarlimab in the GARNET trial experience a statistically significant and clinically meaningful reduction in the risk of death versus both paclitaxel and doxorubicin monotherapy; significant improvements were observed in the MAICs versus all three relevant studies.

Based on the MAIC results, patients treated with dostarlimab were approximately three times less likely to die at any given timepoint compared to patients receiving paclitaxel or doxorubicin ***** ***** ***** ***** based on McMeekin et al. (2015) (HR: ; 95% CI: , ; p< ).[6] An even greater reduction in the risk of death for patients treated with dostarlimab was seen for the other two MAICs; patients treated with dostarlimab were at least five times less likely to die at any given timepoint versus patients treated with doxorubicin in Makker et al. (2013) (HR: ***; 95% CI: *****, ; p<) and Julius et al. (2013) (HR: ****; 95% CI: *****, ; p<).[7, 11] The point estimates versus Makker et al. (2013) and Julius et al. (2013) must be interpreted with caution, given the limitations associated with the MAICs; although the magnitude of the benefits observed, and their similarity to the more robust MAIC versus McMeekin et al. (2015) indicates it is likely that dostarlimab provides a significant and substantial OS benefit versus doxorubicin monotherapy and paclitaxel monotherapy.[6, 7, 11]

The HRs from the OS MAICs versus Rubinstein et al. (2019) and Mazgani et al. (2008) indicate that there are no significant differences between dostarlimab and carboplatin plus paclitaxel with respect to OS.[59, 60] However, it is important to note that the MAIC versus Rubinstein et al. (2019) violates the proportional hazards assumption, meaning the HR is associated with uncertainty; the sample size was too small (N=**) to estimate time varying hazards.[59] The KM curves presented in Figure 36 and Figure 37 suggest that dostarlimab may provide an increased long-term OS

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 95 of 222

benefit versus carboplatin plus paclitaxel, with evidence of a OS plateau for dostarlimab when approximately % of patients are still alive, while the corresponding plateaus for carboplatin plus paclitaxel are associated with much lower proportions of patients alive (% and **%, respectively).

However, considering the wide CIs, limited sample sizes (N=20 and N=31 for Rubinstein et al. [2019] and Mazgani et al. [2008], respectively) and the paucity of data on prognostic variables, it is extremely difficult to draw any meaningful conclusions about the relative benefit of dostarlimab versus carboplatin plus paclitaxel.[59, 60] The discordance in the median OS estimates in the two published studies adds further uncertainty, with estimates of 27.0 months (95% CI: 6.0, 117.0) for Rubinstein et al. (2019) and 15.0 months (95% CI: 9.1, 30.4) for Mazgani et al. (2008), respectively.[59, 60] It is clear the populations of the two studies are not homogenous, although without further details on patient characteristics and prognostic variables, it is difficult to determine if either of these populations are closely matched to the population in GARNET after the matching process.

Table 30: OS for dostarlimab versus chemotherapy comparators based on MAICs

Abbreviations: CI: confidence interval; HR: hazard ratio; MAIC: matching-adjusted indirect comparison; PLD: pegylated liposomal doxorubicin. Source: GSK Data on File.[13]

Figure 33: MAIC OS KM curves for dostarlimab versus paclitaxel or doxorubicin[a] based on McMeekin et al. (2015)[6]

==> picture [454 x 181] intentionally omitted <==

Footnote: Patients in McMeekin et al. ( 2015)[6] received either paclitaxel or doxorubicin. Abbreviations: KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; OS: overall survival. Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 96 of 222

Figure 34: MAIC OS KM curves for dostarlimab versus doxorubicin based on Makker et al. (2013)[11]

==> picture [452 x 180] intentionally omitted <==

Abbreviations: KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; OS: overall survival. Source: GSK Data on File.[13]

Figure 35: MAIC OS KM curves for dostarlimab versus PLD based on Julius et al. (2013)[7]

==> picture [454 x 180] intentionally omitted <==

Abbreviations: KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; OS: overall survival; PLD: pegylated liposomal doxorubicin. Source: GSK Data on File.[13]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 97 of 222

Figure 36: MAIC OS KM curves for dostarlimab versus carboplatin plus paclitaxel based on Rubinstein et al. (2019)[59]

==> picture [454 x 181] intentionally omitted <==

Footnotes: It is likely that the MAIC between dostarlimab and carboplatin plus paclitaxel based on Rubinstein et al. (2019) violates the proportional hazards assumption (further detailed in Appendix D.5.3), although it is difficult to say this definitively due to the small number of patients include in Rubinstein et al. (2019).[59] Nonetheless, there are no viable alternatives to the proportional hazards assumption, as Rubinstein et al. (2019) does not report sufficient data to be able to estimate time varying hazard ratios.[59] Abbreviations: KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; OS: overall survival. Source: GSK Data on File.[13]

Figure 37: MAIC OS KM curves for dostarlimab versus carboplatin plus paclitaxel based on Mazgani et al. (2008)[60]

==> picture [452 x 180] intentionally omitted <==

Abbreviations: KM: Kaplan-Meier; MAIC: matching-adjusted indirect comparison; OS: overall survival. Source: GSK Data on File.[13]

B.2.7.2.3 Strengths, limitations and conclusions of the supportive comparative efficacy evidence

ITC between dostarlimab (GARNET) and doxorubicin (ZoptEC)

Strengths and limitations of the ITC versus ZoptEC (Section B.2.7.2.1)

One of the key strengths of the comparison versus ZoptEC (relative to the MAICs versus the published studies), is that IPD was available for the ZoptEC trial.[8-10] This provided far more detailed data on patient characteristics and key prognostic factors, as well as allowing patients to be removed from both cohorts, rather than just GARNET, in order to minimise the heterogeneity between the two study populations.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 98 of 222

Despite this, some data were still missing for patients in ZoptEC, including the number of lines of prior anti-cancer treatment, a key prognostic variable identified by clinical experts, which does limit the robustness of the comparison. However, it is important to reiterate that far more data for the patients in ZoptEC were available relative to the studies included in the MAICs, and as such, the resulting comparison in this analysis is more robust, and associated with much less uncertainty because the two populations can be more closely aligned with respect to key prognostic variables, relative to the MAICs (and particularly the MAICs versus studies other than McMeekin et al. [2015]).[6, 8-10]

Similar to McMeekin et al. (2015), ZoptEC provides a far greater sample size, including N=233 patients treated with doxorubicin monotherapy, which reduces the uncertainty associated with the comparison.[6, 8-10] This sample size is more than ten times the number of patients included in the MAICs versus Mazgani et al. (2008), Rubinstein et al. (2019) and Makker et al. (2008), representing another key strength of this analysis.[11, 59, 60]

Strengths and limitations of the MAICs (Section B.2.7.2.2)

The comparison between GARNET and McMeekin et al. (2015)[6] is the strongest comparison within this series of MAICs, and provides a reasonably robust comparison between dostarlimab and doxorubicin or paclitaxel monotherapy. McMeekin et al. (2015)[6] is the only RCT of the five comparator studies, and reports the most detailed inclusion/exclusion criteria and published patient characteristics and prognostic data. The inclusion/exclusion criteria are closely aligned with those of GARNET, and notably, patients with and ECOG PS of 2 were excluded from both studies.

Additionally, the detailed patient characteristics reported by McMeekin et al. (2015)[6] allowed for the IPD in the GARNET trial to be adjusted at a greater length, in order to align the data as closely as possible with McMeekin et al. (2015)[6] and to minimise any heterogeneity between the two studies before the comparison was conducted.

The robustness of the MAIC between GARNET and McMeekin et al. (2015)[6] was supported by a novel quasi-validation analysis, which showed a high level of agreement between the published and the MAIC-calculated endpoint estimates, with no indication of bias (this validation is detailed further in Appendix D.5.3). As a result, the comparison between dostarlimab based on GARNET and paclitaxel or doxorubicin based on McMeekin et al. (2015)[6] can be considered to be reasonably robust, and is associated with minimal uncertainty.

However, the MAICs between GARNET and the remaining four comparator studies must be interpreted with much more caution, and are associated with substantial uncertainty. All of these are retrospective, single-arm studies, and the lack of intra-trial randomisation means that the MAICs cannot account for any prognostic variable imbalances that are not reported, introducing an unknown level of bias. This is a concern given the paucity of reported patient characteristics and prognostic variables for these studies. Clinical expert opinion stated that the number of lines of anti-cancer treatment, histology and ECOG PS were the three most important prognostic variables that should be considered when conducting the series of MAICs. A key limitation is therefore that none of the studies (including McMeekin et al. [2015][6] ) reported the number of lines of prior anti-cancer therapy, the most important prognostic variable.

Nevertheless, McMeekin et al. (2015) and Makker et al. (2013) did both report data on histology and ECOG PS, which experts indicated were the second and third most important prognostic

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 99 of 222

variables that should be considered.[6, 11] However, ECOG PS data was not reported by Rubinstein et al. (2019), Mazgani et al. (2008) or Julius et al. (2013), and Julius et al. (2013) did not report any information on histology.[7, 59, 60] As such, the MAICs versus Julius et al. (2013), Rubinstein et al. (2019) and Mazgani et al. (2008) must be interpreted with particular caution, because they may be influenced by unknown levels of bias and the GARNET population cannot be matched with respect to multiple key prognostic variables.[59, 60]

The discordance in the PFS and OS results reported in Rubinstein et al. (2019) and Mazgani et al. (2008) highlights the concern resulting from the limited information of prognostic variables.[59, ] 60 The clear difference in the median OS estimates of 27.0 months (95% CI: 6.0, 117.0) for Rubinstein et al. (2019) and 15.0 months (95% CI: 9.1, 30.4) for Mazgani et al. (2008) indicates that there is heterogeneity between these two populations, although the limited patient characteristics means it is difficult to know if either of these patient populations can be closely matched to the GARNET trial as part of the matching process.[59, 60]

The sample sizes of these studies represents a further limitation. Aside from McMeekin et al. (2015)[6] , the largest of the other four studies, Julius et al. (2013), only included 41 patients in the relevant study population, while the relevant populations of Rubinstein et al. (2019), Mazgani et al. (2008) and Makker et al. (2013) included 20, 19 and 17 patients, respectively.[7, 11, 59, 60] These extremely small sample sizes mean that all of the resulting MAICs, and point estimates (HRs), are highly unreliable and must be interpreted with caution. The small sample sizes result in a further limitation for the MAICs versus Rubinstein et al. (2019); the proportional hazards assumption was violated for both PFS and OS meaning that the HRs are associated with substantial uncertainty, although, due to the small sample size, it is not possible to account for this by estimating time-varying HRs.[59]

Conclusions

The results of the IPTW ITC versus doxorubicin monotherapy based on the ZoptEC study provide clear evidence of the marked survival benefit of dostarlimab compared to doxorubicin monotherapy.[8-10] The comparison estimates that patients treated with dostarlimab were % less likely to die at any given timepoint compared to patients receiving doxorubicin monotherapy (HR: ; 95% CI: (, ); p).The results of the MAIC versus McMeekin et al. (2015)[6] suggest that dostarlimab provides an even greater OS benefit, indicating that patients were ****************** less likely to die at any given timepoint compared to patients receiving paclitaxel ***** ***** ***** ***** or doxorubicin monotherapy (HR: ; 95% CI: , ; p< ).

These two comparisons represent the most robust analyses, and the similarity of the results suggests that it is reasonable to conclude the true magnitude of the OS benefit that dostarlimab provides relative to doxorubicin monotherapy lies approximately around a %–% reduced risk – of death (HR: ) based on the OS HRs for dostarlimab versus doxorubicin monotherapy in ZoptEC, and dostarlimab versus doxorubicin or paclitaxel monotherapy in McMeekin et al. (2015).[6, 8-10] Clinical experts indicated it is reasonable to assume that the efficacy of paclitaxel monotherapy is approximately equal to doxorubicin monotherapy, and so the OS benefit for dostarlimab versus paclitaxel is likely to be similar.

It is more difficult to draw any meaningful conclusions versus the remaining four MAICs, due to the substantial limitations and associated uncertainty. The MAICs versus Makker et al . (2013) and Julius et al. (2013) both suggest that dostarlimab provides an even greater OS benefit versus doxorubicin monotherapy, compared to the results observed in the MAICs versus ZoptEC

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 100 of 222

and McMeekin et al. (2015).[6-10] The exact magnitude of the benefits observed in the MAICs versus Makker et al. (2013) and Julius et al. (2013) are highly uncertain, but the magnitude of these risk reductions favour dostarlimab to such a degree that it would be extremely unlikely that any prognostic imbalances that could not be accounted for would eliminate this advantage completely.[7, 11] Therefore, it can be concluded that these MAICs provide additional supportive evidence for an OS benefit for dostarlimab versus doxorubicin monotherapy, despite the associated uncertainty.

Similarly, the MAICs versus Rubinstein et al. (2019) and Mazgani et al. (2008) are associated with substantial uncertainty.[59, 60] These MAICs indicate that there are no significant differences between dostarlimab and carboplatin plus paclitaxel with respect to PFS or OS, although, the limitations noted previously mean that the analyses are insufficiently robust to draw conclusions with any certainty.

The results of the ITCs versus the published literature presented in this section were considered within economic scenario analyses, in order to provide supportive evidence to the base case economic analysis comparison versus the UK RWE study, as detailed in Section B.3.8.3.

B.2.7.3 Comparative efficacy evidence versus hormone therapy

Hormone therapy was not fully captured within the UK RWE study, because this study only included treatments captured within the SACT dataset, which primarily focusses on treatments provided in secondary care. Patients receiving hormone therapy dispensed in primary care or community pharmacies would therefore not have been captured in this analysis.

Hormone therapies were also not included in the original clinical SLR, because they were not considered to be relevant comparators to dostarlimab in this submission at the time the review of conducted. However, hormone therapy has since been included in the NICE final scope for this submission, and UK clinical expert opinion sought by GSK has since suggested that hormone therapies, specifically medroxyprogesterone acetate and letrozole, may also be treatment options for a patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after platinum-based chemotherapy. These patients may be treated with hormone therapy despite the fact that BGCS guidelines highlight that there is no evidence that hormone therapy confers any survival benefit in the post-platinum setting. [14] In order to identify any relevant published evidence for hormone therapies in this indication, a targeted literature review (detailed in Appendix L) was conducted. The review followed the same eligibility criteria of the clinical SLR but was limited to searches of PubMed, which comprises more than 30 million citations from MEDLINE, life science journal and online books.

The targeted literature review did not identify any studies that provided evidence for hormone therapies in the correct population relevant to this submission’s decision problem: i.e. patients with recurrent or advanced EC who have progressed on or after platinum-based chemotherapy. Consequently, in order to attempt a possible direct comparison between dostarlimab and hormone therapy, the studies that were excluded at the full-text review stage of the targeted literature review were re-evaluated using a relaxed set of eligibility criteria to try and identify any published PFS and OS data.

The process of identifying potential studies for inclusion, and a summary of the six studies that were included as part of this re-evaluation, is provided in Appendix L.5. The patients in these six studies did not represent a GARNET-like population, but the possibility of using these studies as

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 101 of 222

a potentially relevant proxy to inform a comparison between dostarlimab and hormone therapy was explored.

However, UK clinical experts strongly indicated that they did not consider these studies to represent a plausible proxy for hormone therapy for patients with recurrent or advanced EC who have progressed on or after platinum-based chemotherapy. The substantial heterogeneity between the patient populations in these studies and the patient population in GARNET meant that the median survival outcomes reported in these studies were not considered to be reflective of the survival outcomes that would be associated with hormone therapy in the post-platinum setting.

Based on the recommendations from UK clinical experts, it was concluded that any comparisons between GARNET and these studies would be associated with too much uncertainty to be meaningful. Ultimately, it was therefore not possible to conduct any clinical efficacy comparison between dostarlimab and hormone therapy in this submission. A scenario analysis was conducted assuming that hormone therapy as equal to the efficacy of current clinical management (Section B.3.8.3).

B.2.8 Adverse reactions

Safety profile of dostarlimab

• In GARNET, dostarlimab was shown to be well-tolerated, and associated with a manageable adverse event (AE) profile – treatment related treatment-emergent adverse events (TEAEs) were generally low grade (only % of patients reported any Grade ≥3 treatment-related TEAE), and discontinuation as a result of treatment-related AEs was low (%). • The most frequent treatment-related TEAEs were diarrhoea (%), asthenia (%), fatigue (%) and nausea (%). In total, ** patients experienced a ≥ Grade 3 treatmentrelated TEAE; the most frequently observed events were anaemia in * patients (%) and lipase increased in * patients (%). • Treatment-related serious TEAEs were experienced by ** patients (%). Colitis was the only treatment-related serious TEAE reported in more than one patient ( patients [**%]). No deaths were associated with dostarlimab. Dostarlimab versus current clinical management • Cytotoxic chemotherapy regimens are associated with debilitating side effects and a substantial burden of toxicity to patients. In the ZoptEC study, 96.4% of patients receiving doxorubicin monotherapy study reported a treatment-related TEAE (compared to ****% in GARNET), and % of patients reported any Grade >3 TEAE in ZoptEC (versus % in GARNET).[8-10] • The variety in types and frequency of treatment-related TEAEs also highlight the increased burden of toxicity of chemotherapy relative to dostarlimab. The main differences observed between dostarlimab and doxorubicin monotherapy in the ZoptEC study include anaemia (% versus 40.6% in GARNET and ZoptEC, respectively), fatigue (% versus 39.8%) and nausea (***% versus 50.2%).[8-10] • Alopecia and neutropenia were not reported as frequently observed TEAEs experienced by **% of patients in the GARNET study, while 34.9% and 50.6% of patients in ZoptEC experienced alopecia and neutropenia, respectively.[8-10] Summary

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 102 of 222

• Overall, the safety data suggest that the AE profile of dostarlimab is aligned with other currently licensed anti-PD-L1 therapies, and no unexpected AE signals were identified.

• Alongside the potential efficacy benefit that dostarlimab may provide, it is clear that dostarlimab is associated with a reduced burden of toxicity versus cytotoxic chemotherapy regimens that are currently used in UK clinical practice.

In the GARNET study, the safety profile of dostarlimab was evaluated based on reported AEs, which were captured as a secondary endpoint. Safety data are presented for the safety analysis set which was defined as all patients who received any amount of dostarlimab regardless of follow-up time at the time of IA2 (DCO 1[st] March 2020) (Table 8) (N=129). As the safety analysis set represents the same patient population as the ITT population, it is hereafter referred to as the ITT population.

Overall, dostarlimab was shown to be well-tolerated. The majority of treatment-related TEAEs were of low grade; only ****% of patients reported any Grade ≥3 treatment-related TEAE. Discontinuation as a result of treatment-related AEs was low (***%). Overall, the safety data suggest that the AE profile of dostarlimab is aligned with other currently licensed anti-PD-L1 I-O therapies, and with no unexpected AE signals identified at the time of IA2 (DCO 1[st] March 2020).

B.2.8.1 Treatment exposure

A total of 129 patients had received at least one dose of dostarlimab and were included in the safety analysis at the time of IA2 (DCO: 1[st] March 2020). The overall median treatment duration was **** weeks (range: –** weeks), and the mean was **** weeks (STD: *****).

In total, ** patients (%) were exposed to dostarlimab monotherapy for at least 24 weeks, whereas ** patients (%) and ** patients (****%) were exposed for at least 48 and 72 weeks respectively. The median treatment dose intensity of ***% indicates that the majority of patients in GARNET received treatment as planned, without delays or interruptions.

A detailed overview on the duration of treatment with dostarlimab is provided in Table 31.

Table 31: Duration of treatment with dostarlimab (ITT population)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 103 of 222

Median IQR** ******** Min, Max** ******** Dose intensity Median dose intensity** ****

Abbreviations: IQR: interquartile range; ITT: intention to treat; STD: standard deviation. Source: GSK Data on File.[54]

B.2.8.2 TEAEs

TEAEs were defined per protocol as any AE or serious adverse event (SAE) with onset beginning at the day of first administration of study treatment, throughout the treatment period until 90 days after the EOT visit (or until the start of alternate anticancer therapy, whichever occurred earlier), or any event that was present at baseline but worsened in intensity or was subsequently considered treatment-related by the Investigator through the end of the study.

Most patients in the ITT population (N=129) experienced at least 1 TEAE (****%; Table 32). Treatment-related TEAEs were reported in ****% of patients. The majority of TEAEs were not severe or serious and did not require treatment interruption or discontinuation. TEAEs leading to death were reported in * patients (***%); none of these TEAEs were assessed by the Investigator as related to study treatment or considered to be an immune-related adverse event (irAE).

A summary of TEAEs in GARNET is presented in Table 32.

Table 32: Overall summary of TEAEs (ITT population)

Footnotes: For each category, participants were included only once, even if they experienced multiple events in that category. TEAEs are new AEs that began, or any pre-existing condition that worsened in severity, after at least 1 dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anticancer therapy, whichever occurred earlier). AE severity was graded using NCI CTCAE v4.03. Abbreviations: AE: adverse event; CTCAE: Common Terminology Criteria for Adverse Events; EOT: end-oftreatment; irAE: immune-related adverse event; ITT: intention to treat; NCI: National Cancer Institute; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[13]

****** ********* ******* The most frequently reported TEAEs (≥20%) with dostarlimab were , , , ******* ******** , and . These common TEAEs were Grade 1 or Grade 2 in severity in most patients for whom the TEAEs were reported, with the exception of *******, for which ** patients (****%) had Grade 3 events. A summary of common TEAEs (≥10% of patients) is presented in Table 33.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 104 of 222

Table 33: Common TEAEs by system organ class and preferred term (≥10% of patients) (ITT population)

Footnotes: AEs were coded using MedDRA version 23.0. For each preferred term, a patient was included only once, even if they experienced multiple events in that preferred term. TEAEs are new AEs that began, or any preexisting condition that worsened in severity, after at least one dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anti-cancer therapy, whichever occurred earlier).

Abbreviations: AE: adverse event; EOT: end-of-treatment; ITT, intention-to-treat; MedDRA: Medical Dictionary for Regulatory Activities; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[13]

Grade ≥3 TEAEs were reported in ** patients (****%). The Grade ≥3 TEAEs with the highest ******* ************** incidence (≥5%) were and . A summary of Grade 3 or greater TEAEs occurring in more than three patients is presented in Table 34.

Table 34: Grade 3 or greater TEAEs occurring in ≥3 patients (ITT population)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 105 of 222

Grade 3 Anaemia ******* Grade 3 Abdominal pain** ***** Hyponatraemia** ******* Grade 3 ******* Grade 4 ***** Grade 3 Acute kidney injury** ***** Grade 3 Back pain** ***** Pulmonary embolism** ******* Grade 3 ******* Grade 4 ***** Sepsis** ******* Grade 4 ******* Grade 5 ***** Grade 3 Alanine aminotransferase increased** ***** Grade 3 Diarrhoea** ***** Grade 3 Hypertension** ***** Grade 3 Lipase increased** ***** Pneumonia** ******* Grade 3 ******* Grade 5 ***** Grade 3 Urinary tract infection** *******

Footnotes: AEs were coded using MedDRA version 23.0. AE severity was graded using NCI CTCAE v4.03. For each preferred term, a patient was included only once, even if they experienced multiple events in that preferred term. TEAEs are new AEs that began, or any pre-existing condition that worsened in severity, after at least one dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anti-cancer therapy, whichever occurred earlier). Abbreviations: AE: adverse event; CTCAE: Common Terminology Criteria for Adverse Events; EOT: end-oftreatment; ITT: intention-to-treat; MedDRA: Medical Dictionary for Regulatory Activities; NCI: National Cancer Institute; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[54]

The only Grade 4 TEAE reported in >1% of patients was ****** (* patients [***%]). Grade 4 TEAEs of *******, ***********, ************************************, ***********************************************, *************, and ****************** were reported in * patient (***%) each. Grade 5 TEAEs of **********, ****************, *********, ******, and ***** were reported in * patient (***%) each. None of the Grade 5 TEAEs were considered related to study treatment by the Investigator.

In total, ** patients experienced TEAEs leading to study treatment interruption; the most frequently reported TEAEs leading to treatment interruptions (>2% of patients) were ******* and ********* ** **** . A total of patients ( %) experienced TEAEs that led to discontinuation of study treatment. ********************************** and *********************** were the only TEAEs reported in >1 patient that led to permanent discontinuation of study treatment (* patients [***%] each, all events were Grade 3).

It is important to note that Grade ≥3 irAEs of alanine aminotransferase increase, aspartate aminotransferase increased, and pneumonitis required permanent discontinuation of study treatment per protocol.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 106 of 222

B.2.8.3 Treatment-related TEAEs

A total of ** of 129 patients (****%) experienced a treatment-related TEAE. Treatment-related TEAEs experienced by patients treated with dostarlimab were generally low grade, and characteristic of anti-PD-1 therapy. A summary of treatment-related TEAEs is presented in Table 37.

Table 35: Overall summary of treatment-related TEAEs (ITT population)

Footnotes: For each category, participants were included only once, even if they experienced multiple events in that category. TEAEs are new AEs that began, or any pre-existing condition that worsened in severity, after at least 1 dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anticancer therapy, whichever occurred earlier). AE severity was graded using NCI CTCAE v4.03.

Abbreviations: AE: adverse event; CTCAE: Common Terminology Criteria for Adverse Events; EOT: end-oftreatment; irAE: immune-related adverse event; ITT: intention-to-treat; MedDRA: Medical Dictionary for Regulatory Activities; NCI: National Cancer Institute; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[54]

The most frequently reported treatment-related TEAEs (occurring in ≥5% of patients) were diarrhoea (%), asthenia (%), fatigue (%), nausea (%), pruritus (%), arthralgia (%), hypothyroidism (%), anaemia (%) and rash (***%). A summary of treatment-related TEAEs occurring in ≥5% of patients is presented in Table 36.

Table 36: Treatment-related TEAEs experienced by ≥5% of patients (ITT population)

Footnotes: AEs were coded using MedDRA version 23.0. For each preferred term, a patient was included only once, even if they experienced multiple events in that preferred term. TEAEs are new AEs that began, or any preexisting condition that worsened in severity, after at least one dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anti-cancer therapy, whichever occurred earlier).

Abbreviations: AE: adverse event; EOT: end-of-treatment; MedDRA: Medical Dictionary for Regulatory Activities; ITT: intention to treat; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[54]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 107 of 222

Treatment-related Grade ≥3 TEAEs were experienced by ** patients (%). Anaemia ( patients [%]) and lipase increased (* patients [***%]) were the only treatment-related Grade ≥3 TEAEs reported in >2 patients. Table 37 shows related Grade 3 and above treatment-related TEAEs that occurred in 2 or more patients.

Table 37: Treatment-related[a] Grade ≥3 TEAEs occurring in ≥2 patients (ITT population)

Footnotes: AEs were coded using MedDRA version 23.0. AE severity was graded using NCI CTCAE v4.03. For each preferred term, a patient was included only once, even if they experienced multiple events in that preferred term. TEAEs are new AEs that began, or any pre-existing condition that worsened in severity, after at least one dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anti-cancer therapy, whichever occurred earlier). Treatment-related TEAEs refer to any TEAE assessed by the Investigator as related to study treatment (“Related, “Possibly Related” or missing). Events are summarised according to the maximum CTCAE grade experienced by the patient for that event. Abbreviations: AE: adverse event; ALT: alanine aminotransferase; CTCAE: Common Terminology Criteria for Adverse Events; EOT: end-of-treatment; ITT: intention to treat; MedDRA: Medical Dictionary for Regulatory Activities; NCI: National Cancer Institute; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[54]

B.2.8.4 Serious TEAEs

Serious TEAEs were experienced by ** patients (****%) in the ITT population. The most ************** ******************* ****** frequently reported serious TEAEs (>2%) were , , , ****************** ******* *********************** , and . Table 38 presents details of serious TEAEs experienced by >1% of patients.

Table 38: Serious TEAEs by system organ class and preferred term experienced by >1% of patients (ITT population)

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 108 of 222

Pyelonephritis ***** General disorders and administration site conditions** ******* Pyrexia ***** General physical health deterioration** ******* Pain ***** Respiratory, thoracic, and mediastinal disorders** ******* Pulmonary embolism ***** Renal and urinary disorders** ******* Acute kidney injury ***** Neoplasms benign, malignant and unspecified (including** ***** cysts and polyps)** Tumour pain *******

Footnotes: AEs were coded using MedDRA version 23.0. For each preferred term, a patient was included only once, even if they experienced multiple events in that preferred term. TEAEs are new AEs that began, or any preexisting condition that worsened in severity, after at least one dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anticancer therapy, whichever occurred earlier). SAEs between the first dose date and 90 days after the EOT Visit are summarised. Abbreviations: EOT: end-of-treatment; ITT: intention-to-treat; MedDRA: Medical Dictionary for Regulatory Activities; TEAE: treatment-emergent adverse event. Source: GSK- Data on File.[13]

B.2.8.5 Treatment-related serious TEAEs

Treatment-related serious TEAEs were experienced by ** patients (%). ******* was the only treatment-related serious TEAE reported in >1 patient ( patients [*%]). A list of treatmentrelated serious TEAEs is presented in Table 39.

Table 39: Treatment-related serious TEAEs (ITT population)

Footnotes: AEs were coded using MedDRA version 23.0. For each preferred term, a patient was included only once, even if they experienced multiple events in that preferred term. TEAEs are new AEs that began, or any preexisting condition that worsened in severity, after at least one dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anticancer therapy, whichever occurred earlier). Serious TEAEs between the first dose date and 90 days after last dose date are

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 109 of 222

summarised. Treatment-related TEAEs refer to any TEAE assessed by the Investigator as related to study treatment (“Related,” “Possibly Related,” or missing).

Abbreviations: AE: adverse event; CTCAE: Common Terminology Criteria for Adverse Events; EOT: end-oftreatment; ITT: intention-to-treat; MedDRA: Medical Dictionary for Regulatory Activities; NCI: National Cancer Institute; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[54]

B.2.8.6 Deaths

A total of ** patients (****%) died while in the study. Overall, the most common primary reason for death was disease progression.

Of the ** patients with who died while in the study, * patients had a treatment-unrelated TEAE ********** **************** ********* ****** ***** leading to death. All TEAEs leading to death ( , , , , and ) were reported in *** patient each. Of all **** TEAEs leading to death, *** each were classed as System Organ Classes (SOCs) of ‘***************************’, ‘***********’, ‘********’, and

‘*********************’, respectively. None of the TEAEs that led to death were assessed by the Investigator as related to dostarlimab or considered to be an irAE. A summary of the deaths in GARNET is presented in Table 40.

Table 40: Summary of deaths in GARNET (ITT population)

Footnotes:[a] If the last cycle of treatment was ≤4 cycles, the duration of treatment was from first dose to last dose of dostarlimab +21 days; otherwise, it was from first dose to last dose of dostarlimab +42 days. [b] Within 90 days after the EOT Visit or until the first follow-up anticancer therapy, whichever occurred earlier.[c] After the EOT Visit +90 days or until the first follow-up anticancer therapy, whichever occurred earlier. Abbreviations: AE: adverse event; EOT: end-of-treatment; ITT: intention-to-treat. Source: GSK Data on File.[54]

B.2.8.7 Comparative safety

Unfortunately, no AE information is available in the NCRAS data set, therefore safety data were not captured in the UK RWE study. Consequently, naïve comparisons of the safety of dostarlimab versus current clinical management from studies identified in the clinical SLR are discussed below.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 110 of 222

It is important to note that comparisons of safety and AEs between studies must be interpreted with caution, because trial designs and the protocol for AEs reporting and classification may be different between studies.

Of the relevant studies identified in the clinical SLR for individual comparator chemotherapy regimens, only a few reported any details on the safety profile of the chemotherapies listed in the NICE final scope. Studies reporting safety and tolerability information included Lincoln et al. (2003) (paclitaxel monotherapy), McMeekin et al. (2015)[6] (paclitaxel or doxorubicin monotherapy) and the ZoptEC study (doxorubicin monotherapy).[8-10]

Despite the paucity of AE data, it is apparent that patients treated with dostarlimab experience a reduced burden of toxicity relative to patients receiving chemotherapy. Overall, ****% of patients receiving dostarlimab in GARNET experienced any treatment-related TEAEs; in comparison, 90% of patients receiving paclitaxel or doxorubicin monotherapy in McMeekin et al. (2015)[6] , and almost all of the patients receiving doxorubicin monotherapy in ZoptEC (96.4%) experienced a treatment-related TEAE.[8-10] Similarly, only ****% of patients experienced a Grade≥3 treatmentrelated TEAE in GARNET, compared to over half of the patients receiving paclitaxel monotherapy in Lincoln et al. (2003) (58.3%).

A summary of the treatment-related TEAE data presented for patients receiving dostarlimab in GARNET and patients in the chemotherapy studies in presented in Table 41, though it should be noted that these comparisons are naïve only.

Table 41: Summary of treatment-related TEAEs in dostarlimab versus chemotherapy studies identified in the clinical SLR (naïve comparisons only)

Abbreviations: ITT: intention-to-treat; NR: not reported; SAE: serious adverse event; SLR: systematic literature review; TEAE: treatment-emergent adverse event.

There was clear variety in the types and frequency of treatment-related TEAEs experienced by patients receiving dostarlimab versus patients receiving doxorubicin monotherapy in ZoptEC.[8-10] The main differences that were observed between dostarlimab and doxorubicin monotherapy include anaemia (% versus 40.6% in GARNET and ZoptEC, respectively), fatigue (*% versus %) and nausea (% versus 50.2%), highlighting the increased burden of toxicity of chemotherapy relative to dostarlimab.[8-10] Moreover, alopecia and neutropenia were ************ as frequently observed TEAEs experienced by **% of patients in the GARNET study, while ****% and ****% of patients in ZoptEC experienced alopecia and neutropenia, respectively.[8-10]

A summary of treatment-related TEAEs experienced in the GARNET and ZoptEC trials is

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 111 of 222

provided in Table 42 , though it should be noted that these comparisons are naïve only.[8-10]

Table 42: Treatment-related TEAEs experienced patients in GARNET and ZoptEC (naïve comparison only)[8-10]

Footnotes:[a] AEs in GARNET were coded using MedDRA version 23.0. For each preferred term, a patient was included only once, even if they experienced multiple events in that preferred term. TEAEs are new AEs that began, or any pre-existing condition that worsened in severity, after at least one dose of study treatment was administered and throughout the treatment period until 90 days after the EOT Visit (or until the start of alternate anti-cancer therapy, whichever occurred earlier).[b] Each AE and SAE term reported in ZoptEC was mapped to a preferred term using theMedDRA dictionary. The investigator classified the severity of AEs using the NCI CTCAE v4.03 and will assess the relationship of each event to study treatment.

Abbreviations : AE: adverse event; CTCAE: Common Terminology Criteria for Adverse Events; EOT: end of treatment; ITT: intention-to-treat; MedDRA: Medical Dictionary for Regulatory Activities; NCI: National Cancer Institute; NR: not reported; SAE: serious adverse event; TEAE: treatment-emergent adverse event. Source: GSK Data on File.[13] ZoptEC.[8-10]

As highlighted previously, no relevant studies in the patient population of interest were identified for hormone therapy in the targeted literature review. Feedback from clinical experts indicated that the PARAGON study may provide information on the AE profile of hormone therapy and data from this study suggest hormone therapy is associated with a mild safety profile; the only Grade ≥3 AE reported in ≥5% of patients was fatigue.[74]

B.2.9 Ongoing studies

The GARNET trial is still ongoing, with the next data cut expected in early 2022. Dostarlimab is also currently being investigated as a first-line treatment in combination with carboplatin plus paclitaxel for patients with recurrent or advanced EC in the Phase III randomised RUBY trial (NCT03981796).[75] The study has an estimated primary completion date of October 11, 2021.[55]

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 112 of 222

B.2.10 Innovation

Dostarlimab is a treatment option for patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after a platinum-containing regimen. This patient population, which equates to approximately 124 women each year in England, reflects a small, well-defined proportion of the total EC population, and represents the patients with the highest unmet need.

Currently, there is no definitive standard of care for these patients, who are left feeling abandoned and facing a bleak prognosis, with extremely limited and inadequate treatments options based on unclear and inconsistent treatment guidelines (as a result of the dearth of adequate data in this area). Many patients will receive further lines of chemotherapy, although by this stage, EC is considered to be a chemotherapy-resistant disease.[12] A number of patients may alternatively receive hormone therapy, despite the fact that there are no robust published data to support the efficacy of hormone therapy in this post-platinum setting.[14] No new treatment options have been licensed in the UK for this patient population for decades and there remains a critical unmet need for a new addition to the treatment armamentarium with a novel mechanism of action.

Dostarlimab brings innovation to the treatment paradigm for patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after a platinum-containing regimen. It is the first I- O therapy to receive a licence in this indication and has an entirely novel and distinct mechanism of action to the treatment options currently available for these patients. As a humanised, monoclonal antibody, dostarlimab binds with high affinity and specificity to PD-1, a cell surface receptor expressed on activated T-cells.[19] By inhibiting the binding of PD-1 to PD-L1 and PD-L2, dostarlimab blocks the PD-1 signalling pathway and subsequent immune evasion resulting in an increased anti-tumour immune response.

Like other I-O therapies, the mechanism of action of dostarlimab enables a patient’s own immune system to mount an anti-tumour response. This novel mechanism of action has allowed other I-O therapies to revolutionise the management of other cancers, including colorectal and lung cancer and melanoma, where I-O therapies have demonstrating clinically meaningful responses and significantly improved the prognosis for many patients. Most notably, I-O therapies have been shown to result in extended treatment benefits and long-term remission even after treatment discontinuation.[52]

Illustrating the unmet need, it is noted that nivolumab, which is not licensed in this setting, is currently available via the CDF for patients with metastatic or locally advanced dMMR/MSI-H EC through a COVID-19 response programme due to the belief in it having tumour agnostic properties. Nevertheless, there is no available clinical evidence to support the use of nivolumab monotherapy for advanced and recurrent endometrial cancer, nor is a license in this indication being explored to our knowledge.[16] The introduction of a licensed treatment option, such as dostarlimab, would be preferred by UK clinical experts, given the availability of regulatoryapproved data when making a prescribing decision. The use of nivolumab highlights the exceptional unmet clinical need and the limited options available for patients in this setting.

dMMR/MSI-H EC is a subtype of EC that comprises approximately 23% of all ECs and represents a subgroup where dostarlimab and PD-1/PD-L1 inhibition is most effective.[18, 23] dMMR/MSI-H EC is highly immunogenic, and exhibits more tumour-specific neoantigens, which results in increased T-cells, including tumour-infiltrating lymphocytes, and compensatory upregulation of immune checkpoints.[23] This combination of increased mutation load, T-cells and

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 113 of 222

PD-1/PD-L1 expression means that dMMR/MSI-H EC represents an ideal target for dostarlimab and PD-1/PD-L1 inhibition and the efficacy of dostarlimab in this patient population has been realised in the pivotal GARNET trial.[23]

The fact that dostarlimab uses a patient’s own immune system to mount an anti-tumour response also means that AEs are more likely to be on-target and dostarlimab is associated with a reduced burden of toxicity when compared to the indiscriminate cytotoxic effects of chemotherapy.[16] Chemotherapy is associated with a number of harmful and debiliating side effects, and has a substantial detrimental impact on patients with EC.[16 ]

The dosing schedule of dostarlimab is six-weekly versus chemotherapies which have a three to four-weekly dosing schedule.[76, 77] As such, feedback from a UK clinical expert was that the greater gap between each IV administration of dostarlimab may also substantially improve patient convenience and adherence, by reducing the number of required hospital visits.[16]

Finally, it is important to note how during the recent NICE scoping workshop, there was significant excitement about dostarlimab from both the patient group representatives and the clinical experts. GSK have received the same excitement from advisory boards and other insight seeking activities, which clearly and conclusively highlights how dostarlimab would represent a critical addition to the treatment armamentarium for EC, providing hope to patients who currently feel abandoned and who face an extremely dire prognosis with almost no chance of receiving effective treatment.

B.2.11 Interpretation of clinical effectiveness and safety evidence

B.2.11.1 Principal findings from the clinical evidence base

Overall survival

In the GARNET trial, dostarlimab demonstrated a remarkable OS benefit when compared to current clinical management in the UK. Naïve comparisons of the OS results in GARNET versus those in the UK RWE study and the published literature demonstrate that the introduction of dostarlimab would represent a clinically meaningful step-change in the management of patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or after platinum-based chemotherapy.

After 12 months of treatment with dostarlimab, ****% of patients were still alive, and ****% of patients were progression-free. Most notably, a clear and sustained survival benefit was observed; by Month 24, ****% of patients were still alive, and ****% of patients were still progression-free.

These results paint current clinical management in harrowing light, with the UK RWE study showing that just ****% of patients were alive at Month 12. Unlike GARNET, the UK RWE study showed no evidence of long-term survival benefits for patients; by Month 24, just ****% of patients were still alive and ****% of patients were progression free – ************** the percentage of patients who were alive and progression-free in GARNET, respectively.

The marked and sustained OS benefit associated with dostarlimab clearly and conclusively highlights how dostarlimab would represent a critical addition to the treatment armamentarium, providing hope to patients with recurrent or advanced dMMR/MSI-H EC that has progressed on

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 114 of 222

or after platinum-based chemotherapy who currently feel abandoned and who currently face an extremely dire prognosis with almost no chance of receiving effective treatment. The PFS and response rates detailed below only further serve to highlight the clinically significant efficacy that dostarlimab would provide, relative to current clinical management in the UK.

Progression-free survival

In GARNET, ****% and ****% of patients in the ITT population had not experienced disease progression or death at Month 12 and Month 24, respectively. The majority of PFS events occurred in the first six months, and the PFS curve subsequent plateaued when approximately **% of patients had experienced disease progression of death.

This means that the median PFS of *** months is highly uncertain, and associated with very wide *** **** CIs (95% CI: , ). This is illustrated in Section B.2.7.1,

********************** from *** months to *** months. **************************************************************************************************************

**************** , which is line with the long-term benefit observed from other I-O treatments, suggesting that patients who remain progression-free initially may experience a long-term PFS benefit from dostarlimab treatment through to Month 18 and Month 24, with ****% and ****% of patients remaining free of disease progression or death, respectively.

In comparison, patients in the UK RWE study treated with current clinical management faced a far worse prognosis. Just ****% of patients were progression-free at Month 12, and this dropped to just ****% of patients at Month 24 – similar to OS, this was ************** the percentage of patients who were progression-free at the same timepoint following treatment with dostarlimab.[13] Accordingly, there was very limited evidence of any long-term PFS benefit associated with current clinical management.

Response rates

Dostarlimab demonstrated a clinically meaningful and robust ORR of ****% (n=**; 95% CI: **, ), with ** patients (%) achieved a CR, and ** patients (**%) achieving a PR. The ******** of patients who experienced a response to treatment with dostarlimab experienced *************** reduction in tumour size versus baseline. Notably, these responses were durable – after a median follow-up of **** months, the median DOR *******************, and patients who experienced a response had a ****% (*****) and ****% (*****) of experiencing an ongoing response at Month 12 and Month 18, respectively.

These response rates to dostarlimab are striking in comparison to current clinical management. ESMO guidelines highlight that for patients with EC recurring after first-line chemotherapy, only paclitaxel has consistently shown a response rate >20%, less than half the ORR achieved by dostarlimab. In recently conducted RCTs, McMeekin et al. (2015)[6] reports an ORR of 15.7% for patients receiving either paclitaxel or doxorubicin monotherapy (N=223); no patients experienced a CR. For patients receiving doxorubicin monotherapy (N=225), the ZoptEC study reported an even lower ORR of 14.1%, with only 2.0% of patients experiencing a CR.[8-10]

HRQoL

Treatment with dostarlimab preserved patient-reported HRQoL from baseline, as measured by both the EQ-VAS and EORTC QLQ-C30. The results of the EQ-VAS showed an

***********************************************************, while results of the EORTC QLQ-C30

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 115 of 222

showed that patient-reported pain, fatigue symptoms and physical functioning showed a

***************************** above baseline starting at Cycle 2, 3 and 4, respectively. For patients that did experience symptomatic AEs, the majority remained that these AEs remained stable of improved of the course of their treatment relative to baseline.

Safety and tolerability

Overall, dostarlimab was shown to be well-tolerated. The majority of treatment-related TEAEs were of low grade; only % of patients reported any Grade ≥3 treatment-related TEAE. Discontinuation as a result of treatment-related AEs was low (%). The most frequent treatment-related TEAEs were diarrhoea (%), asthenia (%), fatigue (13.2%) and nausea (12.4%). In total, ** patients experienced a ≥ Grade 3 treatment-related TEAE; the most frequently observed events were anaemia in * patients (%) and lipase increased in * patients (*%).

Overall, the safety data suggest that the AE profile of dostarlimab is aligned with other currently licensed anti-PD-L1 I-O therapies, and with no unexpected AE signals identified at the time of IA2 (DCO 1[st] March 2020).

Patients treated with dostarlimab experience a reduced burden of toxicity when compared with cytotoxic chemotherapy. Overall, ****% of patients receiving dostarlimab in GARNET experienced any treatment-related TEAE; in comparison, 90% of patients receiving paclitaxel or doxorubicin monotherapy in McMeekin et al. (2015)[6] , and almost all of the patients receiving doxorubicin monotherapy in ZoptEC (96.4%) experienced a treatment-related TEAE.[8-10]

Comparative efficacy

The results of the adjusted comparisons between dostarlimab in GARNET and current clinical management in the UK RWE study supported the findings of the unadjusted comparisons, indicating that patients treated with dostarlimab experience significantly increased OS compared to patients treated with current clinical management. The results of the two scenarios with the largest ESSs produced adjusted OS HRs between dostarlimab and clinical management of **** **** **** **** **** **** (95% CI: , ) (scenario 1A) and (95% CI: , ) (scenario 2), respectively, **** **** **** compared to an unadjusted OS HR of (95% CI: , ).[13] These results suggest that the UK RWE GARNET-like cohort was closely matched to the GARNET ITT population, and any remaining differences between the two populations may mean that the true OS benefit associated with dostarlimab is slightly underestimated within the unadjusted comparisons.

These findings in terms of an OS benefit for dostarlimab versus comparator therapies were supported by ITCs versus two published RCTs, ZoptEC and McMeekin et al. (2015), which demonstrated clear evidence that dostarlimab provided a marked OS benefit compared to doxorubicin monotherapy.[6, 8-10] Patients treated with dostarlimab with % less likely to die compared to doxorubicin monotherapy in ZoptEC (HR: ; 95% CI: (, ); p) and **% less likely to die at any given timepoint compared to patients receiving paclitaxel or ***** ***** ***** ***** doxorubicin monotherapy in McMeekin et al. (2015)[6] (HR: ; 95% CI: , ; p< ).[8-10]

Four additional MAICs were conducted against other published studies, although these MAICs are associated with substantially increased uncertainty relative to the comparisons described above. These MAICs were conducted versus retrospective, single-arm studies, and the lack of intra-trial randomisation means that the MAICs cannot account for any prognostic variable imbalances that were not reported, introducing an unknown level of bias and representing a key

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 116 of 222

concern, given the paucity of patient characteristics that were reported for these studies. Of the three key prognostic variables identified by clinicians, the number of lines of prior anti-cancer therapy was not reported by any of these four studies, and ECOG PS was only reported by Makker et al. (2013).[11] The small sample sizes of these studies (N=17, N=41, N=20 and N=31) creates additional uncertainty. As such, it is difficult to draw any conclusions with any certainty versus the remaining four MAICs.

B.2.11.2 Strengths of the clinical evidence base

GARNET

The clinical evidence presented as part of the submission has been derived from a comprehensive SLR that was conducted according to the principles of systematic reviewing published in the Cochrane Handbook.

GARNET, a single-arm, open label trial was the only clinical trial identified for dostarlimab from the SLR. This trial represents the key evidence for efficacy and safety for dostarlimab, and is the largest dataset evaluating the anti-PD-1 in recurrent or advanced EC to date. GARNET was conducted as a single-arm trial, yet despite the single-arm nature of the trial, the CASP risk for bias tool determined that exposure in the GARNET trial was accurately measured through validated, objective measurements including ORR, DOR, DCR, PFS, which minimised bias.

The results of GARNET are relevant to the decision problem specified in the NICE final scope, which proposes the use of dostarlimab for patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or after platinum-based chemotherapy. The external validity of GARNET is supported by the following:

• Population : All of the patients in Cohort A1 of the GARNET trial were confirmed to have received prior platinum-based chemotherapy, and all patients were confirmed to have either dMMR/MSI-H or MMR-unk/MSI-H EC. Of the extremely few patients classed as MMR-unk it is highly likely these patients were also dMMR, and this discrepancy results from the fact they were not tested for dMMR status as part of GARNET. The results of GARNET thus provide supportive evidence for the use of dostarlimab in the patient population specified in the decision problem. Furthermore, patients were enrolled in nine UK trial sites, thus increasing the generalisability to the UK recurrent or advanced EC population.

• Intervention: Dostarlimab was evaluated in line with its licensed indication, as a treatment option for patients with recurrent or advanced dMMR/MSI-H endometrial cancer that has progressed on or after platinum-based chemotherapy.

• Comparators: As a result of the lack of standard of care and absence of clear treatment guidelines in this indication, the primary comparative efficacy analysis in this submission compares dostarlimab to current clinical management in the UK. This consists of aggregate data for a cohort of GARNET-like patients identified in a UK RWE study receiving a range of the most commonly utilised chemotherapy regimens in UK clinical practice. These include the individual chemotherapy regimens listed in the NICE final scope, as well as a range of other chemotherapy regimens that are used in UK clinical practice.

• Whilst the UK RWE study serves as the primary comparative efficacy evidence in this submission, a series of ITCs were also conducted, where possible, between dostarlimab and the individual chemotherapy comparators listed in the NICE final scope, based on published studies identified in the clinical SLR.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 117 of 222

• Therefore, despite a paucity of data in the published literature, this submission presents a range of comparative efficacy evidence versus a range of relevant comparators that would be received by patients with recurrent or advanced dMMR/MSI-H EC who have progressed on or after platinum-based chemotherapy in UK clinical practice, including the relevant individual comparators listed in the NICE final scope

• Outcomes: The efficacy and safety profile of dostarlimab was demonstrated in a welldefined, homogenous patient population, considering a wide range of outcomes. This included all of the outcomes outlined in the scope that are relevant to clinicians and to patients (ORR, PFS, OS, adverse events, HRQoL).

Comparative efficacy evidence

The principal limitation of the evidence base supporting this submission is the lack of head-tohead evidence for the comparative efficacy between dostarlimab and the relevant comparators to this submission. There was also a distinct paucity of comparator data identified in the clinical SLR; most studies in the relevant patient population were observational studies, where patient characteristics and KM survival data were poorly reported, limiting the quality, and therefore increasing the uncertainty, of any ITCs.

To mitigate the impact of the paucity of data in the literature for the comparators to this submission, GSK conducted a UK RWE study that included a cohort of GARNET-like patients identified via the NCRAS who received a range of chemotherapy regimens that represent current clinical management for patients with recurrent or advanced EC who have progressed on or after platinum-based chemotherapy, as detailed in Section B.2.7.1.

The UK RWE study provides robust comparative evidence. The NCRAS database collects quality-assured data with complete coverage of all patients diagnosed with cancer in England, meaning that this cohort of *** GARNET like patients are wholly representative of patients in UK clinical practice. These patients were followed-up between 1[st] January 2013 and 30[th] September 2020, representing another key strength, particularly for the evaluation of survival endpoints for these patients.

The large sample size of the UK RWE study is a particular strength; the UK RWE study included almost ten times the number of patients in GARNET and four times the number of patients in the relevant arms of the ZoptEC and McMeekin et al. (2015), the largest relevant comparator studies identified in the clinical SLR.[6, 8-10] Most importantly, detailed patient characteristics and prognostic variable data were available for all of these patients, allowing the wider population of patients initially identified to be narrowed down to a smaller population of patients that closely matched those in GARNET.

Furthermore, these GARNET-like patients included in the UK RWE study received a wide range of different treatment options reflecting the different options that might be used in clinical practice. Given the clear lack of standard of care treatments for patients with recurrent or advanced EC who have progressed on or after platinum-based chemotherapy, a comparison with this cohort of patients, the UK RWE GARNET-like cohort, provides a truer reflection of the outcomes that these patients would experience in UK clinical practice, versus the limited data available in the published literature. Furthermore, four of the top five regimens identified in the RWE study align with the regimens listed in the NICE final scope.

Given the generalisability of this RWE cohort to patients with recurrent or advanced EC who

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 118 of 222

have progressed on or after platinum-based chemotherapy in UK clinical practice, the UK RWE study provides the primary comparative efficacy analysis in this submission, and based on the strengths of the RWE study, the results of the adjusted comparison between dostarlimab and current clinical management inform the base case economic analysis.

Moreover, a comparison of the survival outcomes reported in the UK RWE study with those reported in the published literature supports the generalisability of the UK RWE study to the **** *** **** patient population in GARNET. The median OS estimate of months (95% CI: , ), from the UK RWE study is similar to median OS estimates from the two relevant RCTs presented in the literature: ZoptEC (10.8 months [95% CI: 9.8, 12.6]) and McMeekin et al. (2015) (12.3 months [95% CI: 10.7, 15.4]).[8-10][6, 13] Similarly, the PFS estimate of *** months (95% CI: ***, ***) in the UK RWE study is higher than median PFS estimates in ZoptEC (4.7 months [95% CI: 4.1, 6.6]) and McMeekin et al. (2015) (4.0 months [95% CI: 2.7-4.3]), suggesting that the UK RWE study, and the use of TTNT as a proxy for PFS, may result in an overestimation of PFS relative to the published literature.[6, 8-10]

The comparability of the UK RWE GARNET-like cohort to patients in GARNET was further supported by the results of the adjusted comparison between the two studies, which suggested that the two populations were closely matched with minor differences in the unadjusted and adjusted OS HRs. The comparison suggested that any differences between the two populations may actually result in a slight underestimation of the OS benefit of dostarlimab relative to current clinical management.

The availability of IPD from the ZoptEC study, and the subsequent IPTW ITC that could be conducted between dostarlimab and doxorubicin monotherapy, represents another key strength of the comparative efficacy evidence supporting this submission.[8-10] The IPD meant that it was possible to match the populations of GARNET and ZoptEC with respect to a number of key prognostic variables, unlike almost all of the other published studies identified in the clinical SLR.[8-10] The results of the adjusted analysis showed that dostarlimab resulted in an approximate OS benefit of HR: ; 95% CI: (, ); p**) versus doxorubicin monotherapy, a comparable magnitude of benefit to those observed in the unadjusted (HR: ****) and adjusted comparisons (HR: **** and ****, respectively) between dostarlimab and current clinical management in the UK RWE study. The results of the MAIC versus McMeekin et al. (2015), the only other RCT that could be included in the comparative efficacy analyses, and the only other trial providing detailed information on inclusion/exclusion criteria and prognostic variables, found a similarly comparable OS HR of ***** 95% CI: *****, ; p<) between dostarlimab and doxorubicin or paclitaxel monotherapy.[6]

The clear concordance between the OS benefits observed for dostarlimab versus current clinical management in the UK RWE study, versus doxorubicin monotherapy in the ZoptEC study, and versus doxorubicin or paclitaxel monotherapy in the McMeekin et al. (2015) study is a key strength of the comparative efficacy analyses.[6, 8-10] These three sources of evidence represent the three most robust sources of comparative efficacy evidence available for this submission, and the fact that they all provide comparable estimates for the OS benefit of dostarlimab versus current clinical management provides additional certainty to the conclusion that treatment with dostarlimab results in a marked improvement in OS relative to current clinical management.

B.2.11.3 Limitations of the clinical evidence base

As outlined previously, the principal limitation of the clinical evidence base supporting this

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 119 of 222

submission is the lack of head-to-head evidence for the comparative efficacy between dostarlimab and the relevant comparators to this submission. There was also a distinct paucity of comparator data identified in the clinical SLR; most studies in the relevant patient population were observational studies, and patient characteristics and KM survival data were poorly reported, limiting the quality, and therefore increasing the uncertainty, of any ITCs.

While it is reasonable to state that the UK RWE study somewhat mitigates this limitation and provides robust comparative evidence, it is associated with limitations. While OS data were available, the UK RWE study did not include direct PFS data, because the NCRAS database does not include data collection for progression, remission or recurrence of disease. As such, it was necessary to use proxy measures for PFS and disease recurrence. Whilst this measure has been validated by clinical experts, it is likely that the TTNT estimate used represents a conservative estimate because it is likely that patients would experience a delay between disease progression and the initiation of their next line of treatment, the use of proxy does introduce a level of uncertainty.

The differences between PFS in GARNET and TTNT in the UK RWE study precluded the use of a Cox proportional hazards model to conduct an adjusted ITC for PFS, creating additional uncertainty.

The limited comparative PFS analysis represents an overarching limitation to this submission; while alternative robust comparative efficacy evidence is available for OS, it is extremely limited for PFS. It was not possible to conduct an IPTW ITC for PFS based on the ZoptEC study, due to the differences in tumour assessment timepoints between GARNET and ZoptEC, while McMeekin et al. (2015)[6] did not report a PFS KM curve.[8-10]

PFS KM curves were reported by Makker et al. (2013), Rubinstein et al. (2019) and Mazgani et al. (2008), although all of these are small (N≤31), single-arm retrospective studies which reported extremely limited data on patient characteristics and prognostic variables.[11, 59, 60] As a result, while MAICs were conducted versus these studies, the results are associated with an unknown level of bias and a high degree of uncertainty, due to possible imbalances between GARNET and the comparator studies that could not be matched due to the limited patient characteristics reported.

The results of the unadjusted comparison between PFS in GARNET and TTNT in the UK RWE study, and the results of the adjusted comparison between PFS in GARNET and ZoptEC using the main analysis sets indicate that dostarlimab provides a clear PFS benefit versus current clinical management and doxorubicin monotherapy, respectively.[8-10] However, it is difficult to draw any further conclusions on the PFS benefit of dostarlimab versus current clinical management with any uncertainty, given the substantial limitations associated with the other four MAICs.

The substantial limitations associated with Rubinstein et al. (2019) and Mazgani et al. (2008) introduce a second limitation, meaning that there is no robust comparative efficacy evidence between dostarlimab and carboplatin plus paclitaxel, except for the UK RWE study.[13, 59, 60] It is reasonable to suggest that this represents a minor limitation, given that carboplatin plus paclitaxel was the most frequently received regimen by patients in the UK RWE study (****%) and therefore carboplatin plus paclitaxel is strongly represented in the efficacy outcomes for current clinical management in the UK RWE study.[59, 60] Nevertheless, the paucity of data for carboplatin plus paclitaxel in the published literature and the substantial uncertainty associated

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 120 of 222

with the MAICs versus Rubinstein et al. (2019) and Mazgani et al. (2008) represent a limitation of the comparative efficacy evidence supporting this submission.[59, 60]

Another minor limitation of the comparative efficacy evidence in this submission is that it was not possible to match the populations of patients in any of the comparative efficacy evidence sources with respect to dMMR/MSI-H status, a key inclusion criterion in the GARNET trial. Nevertheless, the impact of this is likely to be minimal, as there is no evidence that MSI-H or dMMR biomarker status has any prognostic or predictive value efficacy and survival outcomes (including recurrence, relapse-free survival, PFS and OS) among patients with advanced or recurrent EC receiving non-anti-PD-(L)1 therapy.[62]

Finally, the UK RWE study only included treatments captured within the SACT dataset; within this dataset drugs which are delivered ‘outside’ an oncology environment (e.g. in surgical clinics or in primary care) are often poorly recorded. Patients receiving hormone therapy dispensed in primary care or community pharmacies would therefore have been poorly captured in this analysis, with a previous study estimating more than 80% of endocrine therapies captured in an alternative NHSE dataset (Cancer Waiting Times) had not been captured in SACT.[78] Therefore, the UK RWE study likely underestimates the true usage of hormone therapy in UK clinical practice in patients with recurrent or advanced EC who have progressed on or after platinumbased chemotherapy.[79] As such, while the UK RWE study provides comparative efficacy versus patients receiving current clinical management consisting of chemotherapy regimens in the UK, it does not provide any comparative evidence between dostarlimab and hormone therapy.

The lack of comparative efficacy evidence versus hormone therapy represents a limitation of the submission because it was also not considered feasible to conduct a comparison between dostarlimab and hormone therapy based on the published literature. No directly relevant studies including hormone therapy were identified in the literature including a patient population closely matched to patients in GARNET. A number of studies in patients outside the post-platinum setting were identified, although clinical expert feedback strongly indicated that the survival outcomes reported in these studies would not be reflective of the survival outcomes that would be associated with hormone therapy in the post-platinum setting. Ultimately, it was therefore not possible to conduct a comparison between dostarlimab and hormone therapy.

B.2.11.4 Conclusion

The marked and sustained OS benefit demonstrated between dostarlimab and current clinical management in the UK RWE study, and supported by the results of an adjusted comparison versus the UK RWE study, an IPTW ITC versus ZoptEC and a MAIC versus McMeekin et al. (2015) conclusively highlight that dostarlimab would represent a critical addition to the EC treatment armamentarium.[6, 8-10] Dostarlimab would provide hope to patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or after platinum-based chemotherapy who currently feel abandoned and who currently face an extremely dire prognosis with almost no chance of receiving effective treatment. Similar improvements in PFS and response rates for dostarlimab versus current clinical management further serve to highlight the clinically significant efficacy that dostarlimab would provide, relative to current clinical management in the UK.

B.2.11.5 End-of-life criteria

The evidence that dostarlimab meets the end-of-life criteria, as outlined by NICE, are detailed in Table 43 below. Based on UK RWE survival estimates with current clinical management,

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 121 of 222

together with naïve comparisons versus survival estimates from the literature identified via a clinical SLR for chemotherapy regimens, it is evident that survival for patients with recurrent or advanced dMMR/MSI-H EC that has progressed on or after platinum-based chemotherapy is less than 24 months with currently available treatments, and that dostarlimab provides an extension to current life expectancy of more than three months.

Company evidence submission template for dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency [ID3802] ©GlaxoSmithKline (2021). All rights reserved Page 122 of 222

Table 43: End-of-life criteria