==> picture [215 x 22] intentionally omitted <==

Single Technology Appraisal

Pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811]

Committee Papers

==> picture [215 x 22] intentionally omitted <==

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL

Pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811]

Contents:

The following documents are made available to consultees and commentators:

Access the final scope and final stakeholder list on the NICE website.

1. Company submission from MSD

2. Clarification questions and company responses

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

• a. Peaches Womb Cancer Trust

• b. NCRI-ACP-RCP-RCR

4. Evidence Review Group report prepared by PenTAG

5. Evidence Review Group report – factual accuracy check

6. Technical engagement response from MSD

• a. TE response – addendum

7. Technical engagement responses and statements from experts:

• a. Dr Clare Green, Medical Oncologist - clinical expert, nominated by MSD

• b. Helen White - patient expert, nominated by Peaches Womb Cancer Trust

• c. Gracey Remmington Teeling – patient expert, nominated by Peaches Womb Cancer Trust

• d. Dr Susan Lalondrelle, Consultant Clinical Oncologist – clinical expert, nominated by the Royal College of Physicians ( see item 3b )

8. Technical engagement responses from consultees and commentators:

• a. Eisai (commentator)

9. Evidence Review Group critique of company response to technical engagement prepared by PenTAG

Any information supplied to NICE which has been marked as confidential, has been redacted. All personal information has also been redacted.

Confidential

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811]

Document B

Company evidence submission

==> picture [150 x 82] intentionally omitted <==

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 1 of 137

Confidential

Contents

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 2 of 137

Confidential

Tables and figures

Table 1: The decision problem ................................................................................... 7 Table 2: Technology being appraised ........................................................................ 9 Table 3: Clinical effectiveness evidence ................................................................... 15 Table 4: Summary of KEYNOTE-775 trial methodology ........................................... 18 Table 5: Baseline characteristics for KEYNOTE-775 ............................................... 21 Table 6: KEYNOTE-775 trial populations ................................................................. 24 Table 7: Summary of statistical analyses for KEYNOTE-775 ................................... 26 Table 8: Analysis of PFS in the KEYNOTE-775 trial (ITT all-comer population) ...... 29 Table 9: Analysis of OS in the KEYNOTE-775 trial (ITT all-comer population) ........ 30 Table 10: Analysis of best overall response in the KEYNOTE-775 trial (ITT all-comer population) ............................................................................................................... 32 Table 11: Overview of key outcomes from KEYNOTE-775 and KEYNOTE-146 ...... 38 Table 12: Summary of observational data available for platinum rechallenge .......... 41 Table 13: Summary of drug exposure in the KEYNOTE-775 trial (safety all-comer population) ............................................................................................................... 44 Table 14: Adverse event summary for the KEYNOTE-775 trial (safety all-comer population) ............................................................................................................... 45 Table 15: Exposure-adjusted adverse event summary (including multiple occurrences of events) in the KEYNOTE-775 trial (safety all-comer population) ..... 47 Table 16: Patients with AEs by decreasing incidence (incidence ≥10% in one or more treatment groups) in the KEYNOTE-775 trial (safety all-comer population) .... 48 Table 17: Summary of AEs of special interest in the KEYNOTE-775 trial (safety allcomer population) .................................................................................................... 50 Table 18: Participants with AEs of special interest by category in the KEYOTE-775 trial (safety all-comer population) ............................................................................. 51 Table 19: Summary of treatment-related AEs (Incidence ≥5% in one or more arms) in the KEYNOTE-775 trial (safety all-comer population) .......................................... 51 Table 20: Summary of Grade 3 to 5 AEs (Incidence ≥5% in one or more arms) in the KEYNOTE-775 trial (safety all-comer population) .................................................... 53 Table 21: Summary of treatment-related Grade 3 to 5 AEs (Incidence > 5% in one or more arms) in the KEYNOTE-775 trial (safety all-comer population) ....................... 54 Table 22: Summary of SAEs by (Incidence ≥ 1% in one or more arms) in the KEYNOTE-775 trial (safety all-comer population) .................................................... 54 Table 23: End-of-life criteria ..................................................................................... 63 Table 24: Features of the economic analysis ........................................................... 67 Table 25: Summary of OS and PFS models selected for economic analysis (PEM+LEN and TPC) ............................................................................................... 73 Table 26: Summary of TOT models selected for economic analysis (PEM+LEN and TPC) 74 Table 27. Number of events and level of maturity of OS in KN-775 ......................... 75 Table 28: Fit statistics of overall independent two-piece survival extrapolation ........ 79 Table 29: Number of patients at risk, KN-775 .......................................................... 81 Table 30. Number of events and level of maturity of PFS in KN-775 ....................... 84 Table 31: Fit statistics of PFS independent two-piece survival extrapolation ........... 87 Table 32: KN-775 median TOT ................................................................................ 90 Table 33: Fit statistics of TOT extrapolation ............................................................. 90 Table 34: Model 2: Coefficients for HRQoL analyses ............................................... 96

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 3 of 137

Confidential

Table 35: Mean utility values based on time to death included in the economic model 96 Table 36: Model 1: Coefficients for HRQoL analyses ............................................... 97 Table 37: Model 1: Mean health state utility values applied in the economic model (scenario analysis) ................................................................................................... 97 Table 38: HRQoL data from previous NICE appraisals for the treatment of gynaecological cancers ............................................................................................ 98 Table 39: Grade 3+ adverse events occurring in 5% or more patients in either arm in KN-775, and number of episodes per patient included in the economic model ...... 100 Table 40: Adverse event durations ......................................................................... 100 Table 41: Summary of utility values used for cost-effectiveness analysis .............. 102 Table 42: Drug formulation, dose, administration, proportion of doses received and total drug acquisition cost per week, intervention and active comparators ............. 105 Table 43: Administration costs ............................................................................... 108 Table 44: Resource use and costs associated with model health states ................ 110 Table 45: AE costs ................................................................................................. 113 Table 46: Subsequent therapies received by patients in KN-775 in the base case analysis .................................................................................................................. 116 Table 47: Subsequent therapy – drug formulation, dose, administration, proportion of doses received and total drug acquisition cost per week, intervention and active comparators ........................................................................................................... 117 Table 48: Duration of subsequent therapies ........................................................... 119 Table 49: Summary of assumptions of the economic analysis ............................... 120 Table 50: Base-case results – List prices ............................................................... 123 Table 51: Mean probabilistic base case results, pairwise analysis – List prices ..... 124 Table 52: Scenario analysis – PEM+LEN versus TPC – List prices ....................... 127 Table 53: Comparison between reported and modelled median survival ............... 130

Confidential

Figure 15: OS independent two-piece parametric survival curves for PEM+LEN ..... 79 Figure 16: OS independent two-piece parametric survival curves for TPC .............. 79 Figure 17: PEM+LEN OS hazard function, with breaking point at Week 26 ............. 81 Figure 18: TPC OS hazard function, with breaking point at Week 26 ...................... 81 Figure 19: Selected OS curve fits for PEM+LEN and TPC ....................................... 84 Figure 20: PEM+LEN and TPC – PFS, KM plot ....................................................... 85 Figure 21: PFS independent two-piece parametric survival curves for PEM+LEN ... 87 Figure 22: PFS independent two-piece parametric survival curves for TPC ............ 87 Figure 23: Selected PFS curve fits for PEM+LEN and TPC ..................................... 89 Figure 24: TOT parametric curves for pembrolizumab (PEM+LEN arm) .................. 90 Figure 25: TOT parametric curves for lenvatinib (PEM+LEN arm) ........................... 90 Figure 26: TOT parametric curves for TPC .............................................................. 90 Figure 27: Selected TOT curve fits ........................................................................... 92 Figure 28: PSA scatterplot, PEM+LEN versus TPC – List prices ........................... 124 Figure 29: Cost-effectiveness acceptability curve – List prices .............................. 124 Figure 30: Tornado diagram showing OWSA results, PEM+LEN versus TPC – List price 125

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 5 of 137

Confidential

==> picture [29 x 13] intentionally omitted <==

Decision problem, description of the technology and clinical care pathway

B.1.1. Decision problem

The submission covers the technology’s full marketing authorisation for this indication, as summarised in Table 1.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 6 of 137

Confidential

Table 1: The decision problem

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 7 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 8 of 137

Confidential

B.1.2. Description of the technology being appraised

The summary of product characteristics and the European public assessment report for KEYTRUDA[®] are presented in Appendix C.

The technology, pembrolizumab with lenvatinib (PEM+LEN), is described in Table 2.

Table 2: Technology being appraised

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 9 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 10 of 137

Confidential

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

B.1.3.1. Health condition

Endometrial cancer (EC) is a cancer of the lining of the womb (uterus), known as the endometrium. Accounting for 94% of all cases of uterine cancer (UC)[4] , EC and UC terminology are often used interchangeably. There were 8,081 new cases of UC reported in England for 2019 (ICD10 code C54) with incidence trends observed to be increasing over time.[5] Incidence rates are highest in people aged 75–79, and approximately 85% of cases are diagnosed in people aged ≥ 55 years.[5]

The most common symptom of EC is abnormal bleeding from the vagina, especially in people who have stopped having periods (post-menopausal women).[6] This common and obvious symptom allows EC to be diagnosed early in most cases, with only 16% of people diagnosed with advanced EC where the cancer has spread outside the womb (Stage III) or beyond the womb, bowel or bladder (Stage IV).[7] When diagnosed at its earliest stage (Stage I), over 90% of people with UC will survive their disease for 5 years or more, compared with 15% of people when the disease is diagnosed at the latest stage (Stage IV).[4]

Approximately 13% of all cases of EC recur, with the risk of recurrence higher in people who are diagnosed with later-stage disease, who have Type II histology (endometrioid grade 3 and non-endometrioid carcinomas), who are older, and who have positive progesterone receptor expression.[8, 9] Like advanced disease, recurrent disease is associated with poor prognosis; the 5-year survival rate for people with recurrent disease is 20%, compared with 89% for people without recurrent disease (survival from diagnosis).[8] Patient health-related quality of life (HRQL) is also reported to decrease with disease recurrence, with an associated increase in anxiety and depression and more threatening illness perceptions reported after diagnosis of relapse.[10]

People with advanced or recurrent EC who have disease progression on or following prior treatment with a platinum-containing therapy are unlikely to live beyond a year with current treatment options (see Section B.2.13.4).[2, 11]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 11 of 137

Confidential

B.1.3.2. Clinical pathway of care

The typical clinical pathway of care for people with EC in England, based on the British Gynaecological Cancer Society (BGCS) Guidelines[1] and validated by practicing clinicians is depicted in Figure 1. Detailed recommendations from BGCS Uterine Cancer Guideline Recommendations for Practice 2021 are provided in Appendix L.

Figure 1: Clinical pathway of care for people with endometrial cancer in England

==> picture [452 x 237] intentionally omitted <==

Key: EC: endometrial cancer.

Notes: *hormone therapy is only used when all other treatment options have been exhausted or further chemotherapy cannot be tolerated.

Source : Adapted from BGCS treatment guidelines.[1]

Little progress has been made in EC management over the past decade, and there is no standard of care beyond platinum-based chemotherapy (carboplatin and paclitaxel) for advanced or recurrent EC, as acknowledged in the BGCS Guidelines.[1, 2, 12]

Platinum rechallenge is most likely used to treat people with advanced or recurrent EC who have only received platinum-based chemotherapy in the (neo)adjuvant setting.[1] There is no consensus on a standard treatment for people with advanced or recurrent EC who have received platinum-based chemotherapy in the systemic

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 12 of 137

Confidential

setting, with limited efficacious treatments available for this patient population, and no treatment option demonstrating superiority over another in a robust clinical trial setting.[12]

Platinum rechallenge may be considered in people who relapse more than six months after first-line platinum-based chemotherapy.[1] Alternative treatment options are currently limited to non-platinum chemotherapy (paclitaxel or doxorubicin monotherapy) or hormone therapy, neither of which have proven efficacy and in practice are associated with low effect expectations and tolerability concerns.[2] Hormone therapy is specifically reserved for use only when all other treatment options have been exhausted or further lines of chemotherapy cannot be tolerated, and even then hormone therapy has a palliative intent rather than an expectation of clinical response.[2]

Pembrolizumab with lenvatinib (PEM+LEN) is intended as a new treatment option for adults with advanced or recurrent EC who have disease progression on or following prior treatment with a platinum-containing therapy in any setting, and who are not candidates for curative surgery or radiation, aligning with its marketing authorization (Table 2). The most common position of use for PEM+LEN is expected to be after platinum-based chemotherapy in the systemic setting, aligning with the current evidence base for PEM+LEN (see Document A, Section A.6). Based on real-world evidence (see Section B.2.9), approximately ***** of people with advanced or recurrent EC previously treated with systemic therapy in the UK receive paclitaxel or doxorubicin monotherapy in current practice with the remaining ***** a mixture of platinum-based chemotherapy, such as carboplatin monotherapy or carboplatin and paclitaxel (data on file).

B.1.4. Equality considerations

No equality issues are foreseen.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 13 of 137

Confidential

==> picture [29 x 13] intentionally omitted <==

Clinical effectiveness

Study identification

• A systematic literature review (SLR) identified two studies that provide efficacy and safety evidence for PEM+LEN in adult patients with advanced or recurrent endometrial cancer (EC) who have disease progression on or following prior treatment with a platinum containing therapy in any setting and who are not candidates for curative surgery or radiation:

  • Single-arm Phase Ib/II study, KEYNOTE-146 (NCT02501096)

  • Phase III randomised controlled trial (RCT), KEYNOTE-775 (NCT03517449)

• KEYNOTE-146 demonstrated that PEM+LEN provided a clinically meaningful objective response rate (ORR) (38.0%; 95% CI: 28.8, 47.8) in patients with previously treated EC

• KEYNOTE-775 is the subsequent confirmatory study for the results observed in KEYNOTE-146, and forms the main clinical evidence base for this submission

Efficacy

• The dual primary endpoints of progression free survival (PFS) and overall survival (OS) were met in KEYNOTE-775:[13]

  • PEM+LEN offers significant and clinically meaningful improvements in PFS (median PFS: 7.2 vs. 3.8 months), OS (median OS: 18.3 vs. 11.4 months), and overall response rate (ORR) (31.9% vs. 14.7%) compared with TPC in an all-comer population

  • Treatment with PEM+LEN also offers significant and clinically meaningful improvements in duration of response (DOR) (14.4 vs. 5.7 months) compared with TPC in an all-comer population

  • Across all patients with advanced EC, no substantial differences were observed in health-related quality of life (HRQoL)

• KEYNOTE-775 trial is the first randomised, controlled Phase III study evaluating a novel combination therapy in the previously treated advanced EC setting that has demonstrated positive results for both primary endpoints of OS and PFS across a broad range of participants, demonstrating the synergistic molecular effects of immuneoncology (IO) and tyrosine kinase inhibitor (TKI) therapy

Safety

• The KEYNOTE-775 trial demonstrates a safety profile consistent with the known safety profile of PEM+LEN, with similar overall incidence of treatment-emergent adverse events (TEAEs) and treatment-related AEs (TRAEs) across both arms[13]

B.2.1. Identification and selection of relevant studies

See Appendix D for full details of the systematic literature review (SLR) process and methods used to identify and select the clinical evidence relevant to the technology being appraised. Two studies were identified that provide efficacy and safety evidence for PEM+LEN in people with advanced or recurrent EC.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 14 of 137

Confidential

B.2.2. List of relevant clinical effectiveness evidence

Details of the PEM+LEN clinical effectiveness evidence are provided in Table 3.

The pivotal trial providing evidence of the clinical benefits of PEM+LEN is KEYNOTE-775: a Phase III trial comparing the efficacy and safety of PEM+LEN versus treatment of physician's choice (TPC: doxorubicin or paclitaxel monotherapy) in adults with advanced or recurrent EC. This pivotal trial informs the economic model presented in Section B.3.

Supportive evidence is provided by the earlier KEYNOTE-146 trial: a Phase Ib/II designed to (i) determine the maximum tolerated dose for lenvatinib in combination with 200 mg intravenous (IV) pembrolizumab every 3 weeks, and (ii) evaluate the efficacy and safety of PEM+LEN in people with solid tumours, including EC. This trial provides almost 3 years of median follow-up compared with the approximate 1 year median follow-up of KEYNOTE-775 and is therefore used to validate model extrapolations.

Table 3: Clinical effectiveness evidence

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 15 of 137

Confidential

Full details of the pivotal trial (KEYNOTE-775) are provided in Sections B.2.2 to B.2.6 of this submission. Relevant outcomes of the supportive trial (KEYNOTE-146) are provided in Section B.2.6; details of the methods and population are provided in Appendix O and safety data are provided in Appendix F.

B.2.3. Summary of methodology of the relevant clinical effectiveness evidence

A summary of the trial methodology for KEYNOTE-775 is presented in Table 4.

KEYNOTE-775 is an ongoing Phase III, multi-centre, randomised, open-label study investigating the efficacy and safety of treatment with PEM+LEN compared with TPC.[14] . TPC consisted of either doxorubicin or paclitaxel monotherapy, which were chosen to reflect standard clinical practice.

The primary objective was to demonstrate that treatment with PEM+LEN is superior to TPC in improving PFS and OS.[3] The study is being conducted at 167 centres in 21 countries, with nine sites in the UK (Table 4). The first patient was enrolled on 11 June 2018, and the estimated study completion date is 25 January 2023.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 16 of 137

Confidential

Both proficient mismatch repair (pMMR) patients and deficient mismatch repair (dMMR) patients were eligible for the trial.[3] Approximately 15% of patients with dMMR were planned to be enrolled, to align with the expected prevalence of dMMR EC in this treatment setting.

B.2.3.1. Trial design

Eligible patients were randomly assigned in a 1:1 ratio to receive one of the following treatment arms:

• Arm A: pembrolizumab 200 mg administered via IV every 3 weeks (Q3W) up to 35 cycles, plus lenvatinib 20 mg every day (QD)

• Arm B: TPC consisting of either doxorubicin 60 mg/m2 Q3W up to a maximum cumulative dose of 500 mg/m2 or paclitaxel 80 mg/m2 every week (QW) on a 28 day cycle, 3 weeks on and 1 week off[3]

The KEYNOTE-775 trial design is presented in Figure 2. The trial consisted of three phases: screening, treatment and efficacy follow up.[3, 14] The screening period occurred for a duration of 28 days, during which informed consent was obtained and protocol eligibility was established according to the pre-determined inclusion and exclusion criteria. Eligible patients went on to complete the treatment period, which lasted from the time of randomisation until the completion of the end of treatment (EOT) visit. Patients received study treatment as continuous 21-day cycles (for patients treated with PEM+LEN and doxorubicin as the TPC choice), or continuous 28-day cycles (for patients receiving weekly paclitaxel as the TPC choice). The efficacy follow-up period is measured from the day after the EOT visit, and will continue for the duration of each patient’s lifetime, or until the data cut-off date for the primary OS analysis if the participant is still alive.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 17 of 137

Confidential

Figure 2: KEYNOTE-775 trial design

==> picture [475 x 127] intentionally omitted <==

Key: TPC, treatment by physician’s choice.

Table 4: Summary of KEYNOTE-775 trial methodology

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 18 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 19 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 20 of 137

Confidential

efficacy endpoints (see Table 7) and were based on the following baseline demographic and disease characteristics:

• Age (<65, ≥65 years)

• Race (White, Asian, other)

• Region (Region 1, Region 2)

• MMR status (pMMR, dMMR)

• ECOG status (0, 1)

• Prior history of pelvic radiation (yes, no)

• Histology (endometrioid, non-endometrioid)

• Prior lines of therapy (1, 2, ≥3)

Key: AE: adverse event; AUC, area under the curve; BICR: Blinded Independent Central Review; CBR: clinical benefit rate; CR: complete response; CSR: clinical study report; DCR : disease control rate; dMMR: deficient mismatch repair; DOR: duration of response; ECOG : Eastern Cooperative Oncology Group; EORTC, European Organisation for the Research and Treatment of Cancer; HRQL: health-related quality of life; LVEF: Left ventricular ejection fraction; MMR: mismatch repair; ORR: overall response rate; OS: overall survival; PD: progressive disease; PEM+LEN, pembrolizumab with lenvatinib; PFS: progression free survival; PFS2: progression free survival on next line therapy; pMMR: proficient mismatch repair; PR: partial response; RECIST: Response Evaluation Criteria in Solid Tumours; SAE: serious adverse event; SD: standard deviation; TTR: time to response; VAS, visual analogue scale.

Notes:[a] , There was no restriction regarding prior hormonal therapy;[b] , These endpoints have not been presented as part of this submission but are available in the CSR. Source: KEYNOTE-775 Clinical Study Report.[3]

B.2.3.2. Baseline characteristics

The baseline characteristics for patients evaluated in the KEYNOTE-775 trial are presented in Table 5.

The baseline demographic and disease characteristics are representative of a patient population with advanced EC. The two treatment groups were generally well balanced for all baseline characteristics. In the intention-to-treat (ITT) population, most patients were white, non-Hispanic, had a median age of 65 years, and had an ECOG performance status of 0.[3] The most frequently reported metastatic site at baseline per investigator assessment were lymph node, intra-abdominal, lung and liver. Multiple histopathological subtypes were presented in both treatment groups, with the majority of carcinomas classified as endometrioid, though carcinoma nonendometrioid subtypes were also well represented.

Table 5: Baseline characteristics for KEYNOTE-775

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 21 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 22 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 23 of 137

Confidential

B.2.4. Statistical analysis and definition of study groups in the relevant clinical effectiveness evidence

B.2.4.1. Trial populations

Definitions of the key study populations analysed and the patient numbers included in each analysis set in the KEYNOTE-775 clinical trial are presented in Table 6.

Efficacy analyses were based on the ITT population, which included all patients in the treatment arm to which they were randomly assigned, regardless of whether they received treatment.[13] Patients were analysed in the treatment group to which they were randomised. No patients were excluded from the analyses.

Safety analyses were based on the all participants as treated (APaT) population, which included all randomly assigned patients who received at least 1 dose of study treatment.[13] Two populations were included in the ITT and APaT analyses: all-comer patients and pMMR patients. Given the anticipated license and positioning for PEM+LEN, only data on the all-comers population are presented within this submission.

For patient reported outcomes, HRQL analyses were based on the HRQL Full Analysis Set (FAS) population, defined as all randomised patients who had at least one HRQL assessment available for the specific endpoint and had received at least one dose of study intervention.[13] Patients were analysed in the treatment group to which they were randomised.

Table 6: KEYNOTE-775 trial populations

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 24 of 137

Confidential

specific endpoint and had received at least one dose of study intervention

Key: EORTC, European Organisation for the Research and Treatment of Cancer; FAS, full analysis set; HRQL, health-related quality of life; ITT: Intention-to-treat. Source: KEYNOTE-775 Clinical Study Report (Tables 10.3; 10.4).[13]

B.2.4.2. Statistical analyses

The trial hypotheses and statistical analysis methods are summarised in Table 7.

Two sets of hypotheses were tested; primary statistical analyses were completed for ‘pMMR participants’ first, then for the entire study population (‘all-comer participants’).[14]

This approach was taken into account for the improved efficacy of pembrolizumab in people with dMMR tumours.[14] dMMR tumours are known to harbour hundreds to thousands of somatic mutations; tumours that have a large number of somatic mutations have been shown to be more susceptible to PD-1/PD-L1 inhibition.

Through sample size powering and hypothesis testing of participants with pMMR tumours, potential positive bias from the improved efficacy of pembrolizumab in people with dMMR tumours is avoided.

Aligning with the marketing authorisation and population of relevance to the decision problem addressed in this submission, efficacy and safety data are provided for the all-comer population in Sections B.2.6, B.2.7 and B.2.9.3. Data for the pMMR population are provided in Appendix M. dMMR subgroup analyses are also provided in Appendix M.

Two interim analyses were planned along with the final analysis, as detailed in Table 7. The data presented in Section B.2.6 are from interim analysis 1 (IA1), providing final efficacy analysis for PFS and the first interim efficacy analysis for OS. The database cut-off date for IA1 was 26 October 2020.[14] Due to primary objectives being met in IA1, the requirement of interim analysis 2 (IA2) was removed through a protocol amendment.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 25 of 137

Confidential

Table 7: Summary of statistical analyses for KEYNOTE-775

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 26 of 137

Confidential

B.2.4.3. Patient disposition

Patient disposition by treatment group for the all-comer population is presented in in Appendix D. A total of 827 patients were randomised (39 randomised in the UK) in a 1:1 ratio to receive either PEM+LEN (n=411) or TPC (n=416).[11] 794 patients received at least one dose of study intervention. As of the latest data cut-off, 93 patients (24%) completed therapy due to reaching the maximum cumulative doses of doxorubicin or paclitaxel in the TPC group. The proportion of patients still receiving study medication was substantially higher in the PEM+LEN arm compared with the TPC arm. The proportion of all patients who discontinued treatment was similar between the two treatment arms.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 27 of 137

Confidential

B.2.5. Quality assessment of the relevant clinical effectiveness evidence

The Cochrane Collaboration’s Risk of bias tool (version 2)[15] was used to assess risk of bias in KEYNOTE-775, as presented in Appendix D.3.

Overall, this study is judged to be at risk of bias in one domain due to its open-label design, with all other domains judged to be at a low risk of bias.

B.2.6. Clinical effectiveness results of the relevant trials

B.2.6.1. KEYNOTE-775

The median duration of follow-up at the time of IA1 (defined as the time from randomisation to date of death or database cut-off date) was 12.2 months (range: 0.3–26.9) in the PEM+LEN arm and 10.7 months (range: 0.3–26.3) in the TPC arm, and for all patients was 11.4 months (range: 0.3–26.9).[11, 13]

As described previously, data are presented for the all-comer population throughout this section. Key results observed in the dMMR and pMMR populations are presented in Appendix M.

B.2.6.1.1. Progression-free survival (primary endpoint)

Treatment with PEM+LEN is associated with significant improvement in progression free survival

The median PFS was significantly improved with PEM+LEN compared with TPC; 7.2 and 3.8 months respectively, with a HR of 0.56 (95% CI: 0.47, 0.66; p< 0.0001), crossing the pre-specified boundary for statistical significance at IA1 of ≤0.0005.[11]

PFS rates were also higher in the PEM+LEN group compared with the TPC group at 6 (53.5% vs. 34.3%), 12 (31.2% vs. 13.2%), 18 (25.0% vs. 7.6%) and 24 months (20.9% vs. 3.8%), as shown in Table 8.[13]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 28 of 137

Confidential

Table 8: Analysis of PFS in the KEYNOTE-775 trial (ITT all-comer population)

The Kaplan–Meier curves for PFS clearly separated, starting at around the time of the first protocol scheduled imaging assessment (i.e., 8 weeks), and remained consistently separated throughout the duration of the evaluation period (Figure 3).[11]

Figure 3: Kaplan–Meier estimates of PFS in KEYNOTE-775 trial (ITT all-comer population)

==> picture [452 x 200] intentionally omitted <==

Key: ITT, intention-to-treat; PFS, progression-free survival; TPC: treatment of physician’s choice. Source: Makker at al. 2022.[11]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 29 of 137

Confidential

B.2.6.1.2. Overall survival (primary endpoint) PEM+LEN is superior to TPC with respect to overall survival in patients with advanced EC who have failed prior therapy

Median OS was significantly longer in the PEM+LEN group compared with the TPC group; 18.3 and 11.4 months respectively, with a HR of 0.62 (95% CI: 0.51, 0.75; p< 0.0001) at IA1, crossing the pre-specified boundary for statistical significance at IA1 of ≤0.0064.[11]

The OS rates were higher in the PEM+LEN group compared with the TPC group at 6 (82.4% vs. 75.4%), 12 (62.5% vs. 47.9%), 18 (50.9% vs. 28.6%), and 24 months (42.0% vs. 21.4%), as shown in Table 9.[13]

Table 9: Analysis of OS in the KEYNOTE-775 trial (ITT all-comer population)

The KM curves for OS clearly separated, starting at around 3 months and remained consistently separated throughout the duration of the evaluation period (Figure 4).[11] Overall, treatment with PEM+LEN led to OS improvements that were statistically significant and clinically meaningful, with the OS HR crossing the pre-specified boundary for statistical significance at IA1.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 30 of 137

Confidential

When interpreting survival data, it is important to acknowledge that nearly half of the patients (48%) in the TPC arm received subsequent anticancer therapy, compared with 28% of patients in the PEM+LEN arm (Table 46).[11] Of the patients in the TPC arm, *****% received subsequent treatment with PD1/PDL1 regimens that are not currently available in the UK which likely dilutes the real-world survival benefit PEM+LEN offers to patients.[13] This is further discussed in Sections B.2.13 and B.3.5.4.

It should also be acknowledged that OS data from the IA1 dataset are subject to heavy censoring, but the study continues to mature with the final analysis database lock estimated in ***** (events dependent).

Figure 4: Kaplan–Meier estimates of OS in the KEYNOTE-775 trial (ITT allcomer population)

==> picture [452 x 203] intentionally omitted <==

Key: ITT, intention-to-treat; OS, overall survival; TPC: treatment of physician’s choice. Source: Makker et al. 2022.[11]

B.2.6.1.3. Objective response rate (secondary endpoint)

Treatment with PEM+LEN led to statistically significant and clinically meaningful improvements in objective response rate (ORR)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 31 of 137

Confidential

ORR (per RECIST 1.1 by BICR) was 31.9% for the PEM+LEN group compared with 14.7% for the TPC group, with an estimated difference of 17.2% (95% CI: 11.5, 22.9%; p<0.0001).[11] The proportion of patients who achieved a CR was higher in the PEM+LEN group compared with the TPC group, as shown in Table 10.

Table 10: Analysis of best overall response in the KEYNOTE-775 trial (ITT allcomer population)

B.2.6.1.3.1. Duration of response and tumour reduction (exploratory endpoints)

Among patients achieving a response, the median DOR was 14.4 months (range: 1.6, 23.7) in the PEM+LEN group compared to 5.7 months (range: 0.0, 24.2) for the TPC group (Appendix N).[11] A greater proportion of patients receiving PEM+LEN

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 32 of 137

Confidential

experienced a reduction in tumour size than those receiving TPC (Figure 5, Figure

6).

Figure 5: Waterfall plot of best percentage change from baseline for target lesions in the PEM+LEN arm of the KEYNOTE-775 trial (ITT all-comer

population)

==> picture [465 x 272] intentionally omitted <==

Source: Makker et al. 2022.[11]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 33 of 137

Confidential

Figure 6: Waterfall plot of best percentage change from baseline for target lesions in the TPC arm of the KEYNOTE-775 trial (ITT all-comer population)

==> picture [477 x 279] intentionally omitted <==

Source: Makker et al. 2022.[11]

B.2.6.1.4. Exploratory efficacy outcomes

A summary of outcomes from the exploratory endpoints in the KEYNOTE-775 trial including DOR, TTR, DCR, CBR and PFS2 is provided in Appendix N.

B.2.6.1.5. Health-related quality of life HRQL was generally similar between both treatment groups

Patient reported outcome (PRO) scores were generally similar between the PEM+LEN group and the TPC group throughout the study[16] , demonstrating that treatment with PEM+LEN has no adverse impact on patients HRQL compared with current treatments available for patients with advanced EC. Patients are therefore able to benefit from the improved responses and chances of survival without compromising their quality of life.

A summary of results for the global health status/quality of life (GHS/QoL) score of EORTC QLQ-30, the EORTC QLQ-C30 physical functioning score and EORTC QLQ EN24 urological symptoms are presented in Appendix N. EQ-5D-5L visual analogue

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 34 of 137

Confidential

scale (VAS) is presented below and is used in the cost effectiveness model for this submission.

Over 12 weeks of follow-up, the EQ-5D-5L VAS scores decreased in both the PEM+LEN group (-4.44; 95% CI: -6.43, -2.46) and the TPC group (-6.79; 95% CI: - 8.98, -4.60).[13] There was no significant difference between groups; difference in least squares (LS) mean change from baseline at Week 12 was 2.35 (95% CI: −0.44, 5.14; *****).[13, 16] Empirical mean change from baseline and 95% CI for EQ-5D-5L VAS over 45 weeks for PEM+LEN and TPC groups are displayed graphically in Figure 7. A slight decrease in mean change was seen in both treatment groups over time.

Figure 7: Empirical mean change from baseline and 95% CI for the EQ-5D VAS score over time in KEYNOTE-775 (FAS all-comer population)

==> picture [484 x 122] intentionally omitted <==

==> picture [484 x 122] intentionally omitted <==

==> picture [484 x 123] intentionally omitted <==

Key: CI, confidence interval; FAS, full analysis set; TPC, treatment of physician’s choice; VAS, visual analogue scale. Source: Lorusso et al. 2021.[16]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 35 of 137

Confidential

B.2.6.2. KEYNOTE-146

Data are presented for the pre-treated EC participants of relevance to the decision problem being addressed in this submission. PFS and OS data are presented here with ORR and DoR data presented in Appendix O.

The median duration of follow-up at the time of the most recent analysis (database cut-off 18 August 2020) presented below was 34.7 months (95% CI: 30.9, 41.2).[17]

B.2.6.2.1. Progression-free survival

The median PFS was 7.4 months (95% CI: 5.2, 8.7) for pre-treated EC participants (Figure 8).[17] These data are consistent with PFS observed for the PEM+LEN group in the most recent analysis of the KEYNOTE-775 trial (see Section Error! Reference source not found. ).

Figure 8: Kaplan–Meier estimates of PFS in KEYNOTE-146 trial (pre-treated EC

population)

==> picture [484 x 100] intentionally omitted <==

==> picture [484 x 100] intentionally omitted <==

Key: CI: confidence interval; dMMR: mismatch-repair deficient; EC: endometrial carcinoma; EC 2L+: endometrial cancer second-line (or greater) treatment; MSI-H: microsatellite instability-high; MSS: microsatellite stable; NE: non estimable; PFS, progression-free survival; pMMR: mismatch-repair proficient; RECIST, response evaluation criteria in solid tumours. Notes: PFS assessed by the investigator per immune-related RECIST Source: Makker et al. 2021.[17]

B.2.6.2.2. Overall survival

The median OS was 17.7 months (95% CI: 15.5, 25.8) for pre-treated EC

participants (Figure 9).[17] These data are consistent with OS observed for the

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 36 of 137

Confidential

PEM+LEN group in the most recent analysis of the KEYNOTE-775 trial (see Section

Error! Reference source not found. ).

The longer-term follow-up in KEYNOTE-146 demonstrates that the 5-year survival rate for patients treated with PEM+LEN is approximately 30% (Figure 9).[17]

Figure 9: Kaplan–Meier plot of OS in the KEYNOTE-146 trial (pre-treated EC

population)

==> picture [484 x 105] intentionally omitted <==

==> picture [484 x 104] intentionally omitted <==

Key: CI: confidence interval; dMMR: mismatch-repair deficient; EC: endometrial carcinoma; EC 2L+: endometrial cancer second-line (or greater) treatment; MSI-H: microsatellite instability-high; MSS: microsatellite stable; NE: non estimable; pMMR: mismatch-repair proficient. Source: Makker et al. 2021.[17]

B.2.7. Subgroup analysis

The treatment benefit observed in PFS and OS for PEM+LEN compared with TPC in all-comer patients was consistent across all major subgroups, including by MMR status, histology and prior lines of therapy (Appendix E).[11]

B.2.8. Meta-analysis

KEYNOTE-146 and KEYNOTE-775 are heterogenous in terms of study design and population and formal meta-analyses have therefore not been conducted. A qualitative overview of key outcomes from both trials is provided in Table 11.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 37 of 137

Confidential

Table 11: Overview of key outcomes from KEYNOTE-775 and KEYNOTE-146

B.2.9. Indirect and mixed treatment comparisons

Two SLRs, one identifying interventional and one identifying observational evidence, were conducted to identify relevant published clinical evidence of pharmacological treatments for advanced or recurrent EC, in line with the population investigated in the KEYNOTE-775 trial.

Full details of the SLR search strategy, study selection process and results are presented in Appendix D for interventional and observational evidence.

B.2.9.1. Interventional evidence

A total of 34 trials were included in the interventional evidence base (including KEYNOTE-775 and KEYNOTE-146) SLR. However, the majority of treatments evaluated were not of direct relevance to UK clinical practice. For the comparators of relevance, interventional evidence was limited to three RCTs with a control arm of paclitaxel or doxorubicin and six single-arm trials of paclitaxel or doxorubicin (see Table 7 of Appendix D). As the KEYNOTE-775 trial provides head-to-head data for PEM+LEN versus paclitaxel or doxorubicin, any ITC would not provide additional informative data. Nonetheless, feasibility of an indirect treatment comparison (ITC) was considered, details of which are provided in Appendix D. The conclusion of this assessment was that ITC was not feasible and that the KEYNOTE-775 trial provides the only robust data for the comparison of PEM+LEN versus chemotherapy for the treatment of patients with advanced or recurrent EC who have disease progression after one prior systemic platinum-based chemotherapy regimen.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 38 of 137

Confidential

B.2.9.2. Observational evidence

A total of six studies were included in the observational evidence base SLR, all of which were retrospective cohort / claims database studies. Four studies provided data on platinum rechallenge, however there were several concerns with the quality and comparability of these data; the remaining two provided data on doxorubicin and therefore were not considered further.

A summary of the platinum rechallenge studies is presented in Table 12.

Two studies reported on platinum rechallenge in recurrent EC patients in Japan who had already received systemic platinum-based chemotherapy (aligning to the treatment positioning investigated in KEYNOTE-775). Both studies reported a significant relationship between treatment-free interval and effectiveness of platinum rechallenge[18, 19] , as reflected in the BGCS Guidelines (see Section B.1.3.2). Patients who received platinum rechallenge within six months of their first-line platinum chemotherapy failed to respond and had a median OS of <12 months (Table 12). Patients who received platinum rechallenge later than six months from their first-line platinum chemotherapy had a better response and a median OS ranging from 14.8 months to 43.0 months depending on the treatment-free interval (Table 12).

Of the other two studies, one reported on platinum rechallenge in recurrent or metastatic EC patients in the US who had received platinum-based chemotherapy in the adjuvant setting, and the other reported on platinum chemotherapy use in advanced or recurrent EC, a small proportion of whom had received prior chemotherapy.[20, 21] Both studies had small patient numbers and reported highly varied and uncertain outcomes. Patients who received platinum rechallenge after adjuvant platinum-based chemotherapy had a low response rate of 10% but a median OS of 27.0 months (95% CI: 6, 117) (Table 12).

Formal ITC using these observational data would be inappropriate for multiple reasons and subject to a high degree of uncertainty. Marked differences in study design and associated data quality, study timing and patient populations are observed compared with the KEYNOTE-775 trial. Naïve comparison of outcomes for platinum rechallenge in those patients relapsing within 12 months of receiving firstline platinum chemotherapy show reasonable alignment to the TPC arm of

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 39 of 137

Confidential

KEYNOTE-775. This represents the patient group which made up the majority of participants in the KEYNOTE-775 trial and would be expected in clinical practice. A highly uncertain ITC using these data would therefore not add value to the highquality comparative data available from the KEYNOTE-775 Phase III RCT.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 40 of 137

Confidential

Table 12: Summary of observational data available for platinum rechallenge

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 41 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 42 of 137

Confidential

B.2.9.3. ECHO: Real-world evidence

Endometrial Cancer Health Outcomes – Europe (ECHO EU5) is a retrospective, multicentre chart review study evaluating treatment patterns and clinical outcomes in advanced or recurrent EC patients previously treated with systemic therapy that has been commissioned by MSD (data on file). Data from the UK cohort of the ECHO study were available, as described below. The study was subsequently used for validating the survival outcomes in the comparator arm for in this appraisal (see Section B.3.3.3)[22]

EC-treating oncologists (medical oncologist or gynaecologic-oncologist) were screened and recruited from the publicly available data. In the UK, ***** physicians that consented to participate in the study provided data for ***** eligible patients. Physicians provided data obtained from medical records of adults diagnosed with advanced or recurrent EC

between 1 July 2016 – 31 December 2018, and who had disease progression after a prior systemic therapy during 1 July 2016 – 30 June 2019. Other inclusion criteria were ≥18 years of age at the time of advanced or recurrent EC diagnosis; must not be a suitable candidate for curative-intent surgery; must not have participated in any EC-related clinical trial during treatment; and must have known medical history from the date of advanced or recurrent EC diagnosis. Patients were excluded if they had any prior malignancy active within the previous 3 years, except for locally curable cancers that have been cured.

All data was collected from diagnosis of advanced or recurrent EC until last available follow up. Key outcomes included patient demographics, clinical and treatment characteristics, and clinical outcomes. Descriptive analysis was conducted to report mean, standard deviation, median and range for all continuous variables, and count and frequency for categorical variable. Kaplan-Meier analyses were performed to estimate time to event outcomes such as real-world PFS and OS. Patients were censored at the last follow up and estimated probabilities of the events were provided at pre-determined time intervals. Real-world response to second-line therapy was abstracted as reported by the physician from patients’ medical records. All statistical analyses were conducted using SAS version 9.4.

Treatment pattern and outcome data from ECHO UK are summarised below:

• ***** were used to treat the majority of patients (*****), with the remaining ***** of patients receiving a mixture of ***** (data re-weighted to exclude investigative treatments).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 43 of 137

Confidential

*****  Median OS from the start of second line of systemic therapy was *****.

This survival outcome is closely aligned to that observed with the TPC arm of the KEYNOTE-775 trial, supporting the applicability of this trial to outcomes expected with current care in clinical practice. These data have been used to inform and validate economic scenario analyses (see Section B.3.8.3).

B.2.10. Adverse reactions

B.2.10.1. KEYNOTE-775

As described previously, data are presented for the all-comer population throughout this section.

B.2.10.1.1. Treatment exposure

The median duration of exposure to at least 1 drug was more than double for the PEM+LEN group compared with the TPC group. More patients in the PEM+LEN group had a duration of exposure of ≥6 months, ≥12 months and ≥18 months compared with the TPC group (Table 13).[11]

Table 13: Summary of drug exposure in the KEYNOTE-775 trial (safety all-comer population)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 44 of 137 of 137137

44 of 137 of 137137

Confidential

B.2.10.1.2. Summary of adverse events

The safety data from the KEYNOTE-775 trial showed that the incidences of Grade 3 to 5 AEs, Grade 3 to 5 drug-related AEs, SAEs and drug-related SAEs, were higher for treatment with PEM+LEN group compared with TPC.[11] The incidence of drug-related deaths was similar in the two groups. A summary of the adverse events in the KEYNOTE775 trial is provided in Table 14.

Table 14: Adverse event summary for the KEYNOTE-775 trial (safety all-comer population)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 45 of 137 of 137137

45 of 137 of 137137

Confidential

The median duration of exposure was over twice as long for PEM+LEN compared with TPC.[11] When adjusted for exposure, the incidences of Grade 3 to 5 AEs and death were lower for the PEM+LEN arm compared with the TPC arm, and that of SAEs was similar between the two groups (Table 15).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 46 of 137 of 137137

46 of 137 of 137137

Confidential

Table 15: Exposure-adjusted adverse event summary (including multiple

occurrences of events) in the KEYNOTE-775 trial (safety all-comer population)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 47 of 137 of 137137

47 of 137 of 137137

Confidential

B.2.10.1.3. Adverse events

The types, incidence, and severity of AEs in the PEM+LEN group were generally consistent with the known safety profiles of pembrolizumab monotherapy or lenvatinib monotherapy.[11] These safety profiles are generally well managed with supportive medications and dose modifications.

The types, incidence, and severity of AEs in the TPC group were consistent with the known safety profile of doxorubicin or paclitaxel.[11] The most frequently reported AEs (incidence ≥30%) were hypertension, hypothyroidism, diarrhoea, nausea, decreased appetite, vomiting, weight decreased, fatigue, and arthralgia for the PEM+LEN arm. For the TPC arm, the most frequently reported AEs were anaemia, nausea, neutropenia, and alopecia.

All the most frequently reported AEs are common or very common AEs associated with pembrolizumab and Lenvatinib as monotherapies.[3]

A summary of the most frequently reported adverse events in the KEYNOTE-775 trial is presented in Table 16.

Table 16: Patients with AEs by decreasing incidence (incidence ≥10% in one or more treatment groups) in the KEYNOTE-775 trial (safety all-comer population)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 48 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 49 of 137 of 137137

49 of 137 of 137137

Confidential

Exposure-adjusted rates of most AEs were either lower for PEM+LEN compared with TPC or similar between the two groups (Appendix N).

A summary AEs of special interest in the KEYNOTE-775 trial is presented in Table 17; participants with AEs of special interest by category are summarised in Table 18.

Table 17: Summary of AEs of special interest in the KEYNOTE-775 trial (safety allcomer population)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 50 of 137

Confidential

Table 18: Participants with AEs of special interest by category in the KEYOTE-775 trial (safety all-comer population)

B.2.10.1.4. Treatment-related adverse events

The incidence of treatment-related AEs in the PEM+LEN group was similar compared with the TPC group (Table 19).[11] The most frequently reported treatment-related AEs (incidence ≥30%) were hypertension, hypothyroidism, diarrhoea, nausea, and decreased appetite for the PEM+LEN group. For the TPC group, the most frequently reported treatment-related AEs were nausea, anaemia, neutropenia and alopecia.

Table 19: Summary of treatment-related AEs (Incidence ≥5% in one or more arms) in

the KEYNOTE-775 trial (safety all-comer population)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 51 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 52 of 137 of 137137

52 of 137 of 137137

Confidential

B.2.10.1.5. Grade 3 to 5 adverse events

The incidence of Grade 3 to 5 AEs in the PEM+LEN group was higher compared with the TPC group (88.9% vs 72.7%).[11] The most frequently reported Grade 3 to 5 AEs (incidence ≥5%) were hypertension, weight decreased, decreased appetite, diarrhoea, lipase increased, anaemia, asthenia, proteinuria, fatigue, and hypokalaemia for the PEM+LEN group. For the TPC group, the most frequently reported Grade 3 to 5 AEs were

neutropenia, neutrophil count decreased, anaemia, white blood cell count decreased, and febrile neutropenia (Table 20).

Table 20: Summary of Grade 3 to 5 AEs (Incidence ≥5% in one or more arms) in the

KEYNOTE-775 trial (safety all-comer population)

B.2.10.1.6. Grade 3 to 5 treatment-related adverse events

The incidence of treatment-related Grade 3 to 5 AEs in the PEM+LEN group was higher compared with the TPC group (77.8% vs 59.0%).[11] The most frequently reported

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 53 of 137 of 137137

53 of 137 of 137137

Confidential

treatment-related Grade 3 to 5 AEs (incidence ≥5%) were hypertension, weight decreased, decreased appetite, and diarrhoea in the PEM+LEN group. For the TPC group, the most frequently reported treatment-related Grade 3 to 5 AEs were neutropenia, neutrophil count decreased, anaemia, white blood cell count decreased, leukopenia, and febrile neutropenia (Table 21).

Table 21: Summary of treatment-related Grade 3 to 5 AEs (Incidence > 5% in one or

more arms) in the KEYNOTE-775 trial (safety all-comer population)

B.2.10.1.7. Serious adverse events

The incidence of SAEs in the PEM+LEN group was higher compared with the TPC group (52.7% vs 30.4%).[13] The most frequently reported SAEs (incidence ≥1%) were hypertension, urinary tract infection (UTI), diarrhoea, decreased appetite, vomiting, acute kidney injury, pyrexia, cholecystitis, colitis, pneumonia, death, dehydration, intestinal obstruction, sepsis, abdominal pain, ileus, and pulmonary embolism for the PEM+LEN group (Table 22). For the TPC group, the most frequently reported SAEs were febrile neutropenia, anaemia, neutropenia, pulmonary embolism, and sepsis.

Table 22: Summary of SAEs by (Incidence ≥ 1% in one or more arms) in the

KEYNOTE-775 trial (safety all-comer population)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 54 of 137

Confidential

B.2.10.1.8. Treatment-related serious adverse events

The incidence of treatment-related SAEs was higher in the PEM+LEN group (33.3%) compared with the TPC group (14.2%).[11] The most frequently reported treatment-related SAEs (incidence ≥1%) were hypertension, colitis, decreased appetite, vomiting, diarrhoea, pyrexia, and acute kidney injury for the PEM+LEN group.[13] For the TPC group, the most frequently reported treatment-related SAEs were febrile neutropenia, neutropenia, and anaemia.[13]

B.2.10.1.9. Deaths

The incidence of AEs resulting in death in the PEM+LEN group was similar compared with the TPC group. When adjusted by exposure, the incidence of AEs resulting in death was lower in the PEM+LEN group than in the TPC group.[11]

Of 23 (5.7%) patients in the PEM+LEN group who experienced AEs resulting in death, 6 deaths (1.5%) were considered related to study intervention by the investigator.[11] One death due to multiorgan dysfunction syndrome was considered related to both lenvatinib and pembrolizumab. One death each due to cerebrovascular accident, right ventricular dysfunction, myelodysplastic syndrome, and death were considered related to lenvatinib, and 1 death due to colitis was considered related to pembrolizumab.[13] The preferred term

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 55 of 137

Confidential

“death” was reported in situations where limited information on the cause of death was available, or where the investigator could not assign a specific AE term.

Of 19 (4.9%) patients in the TPC group who experienced AEs resulting in death, eight deaths (2.1%) were considered related to study intervention by the investigator.[11] These events were all considered related to doxorubicin: two events of pneumonia, and one event each of aspiration, pulmonary embolism, cardiogenic shock, toxic cardiomyopathy, cardiac failure, and sepsis.[13]

B.2.10.1.10. Discontinuation and interruption of treatment

The incidence of patients with AEs that led to discontinuation of any study intervention was higher in the PEM+LEN group compared with the TPC group; 14% of patients discontinued combined treatment of PEM+LEN, with discontinuation of lenvatinib (30.8%) higher than for pembrolizumab (18.7%).[11]

In the PEM+LEN group, the AEs that led to discontinuation of both drugs were generally consistent with the known safety profile of pembrolizumab and lenvatinib, both in combination and as monotherapies.[11] In the TPC group, the AEs that led to discontinuation were consistent with the known safety profile of doxorubicin or paclitaxel.

The incidence of patients with treatment-related AEs that led to discontinuation of any study intervention was higher in the PEM+LEN group compared with the TPC group; 4.9%*****of patients in the PEM+LEN group discontinued treatment due to treatmentrelated AEs, with discontinuation of lenvatinib (22.7%) higher than for pembrolizumab (9.9%).[11]

The incidence of patients with AEs that led to interruption of any study intervention was higher in the PEM+LEN group compared with the TPC group.[11] 30.8% of patients in the PEM+LEN group experienced interruption, with interruption of lenvatinib (58.6%) higher than for pembrolizumab (50.0%).

The most common AEs resulting in treatment interruption of both drugs (>1%) were diarrhoea,*****hypertension,*****and vomiting for the PEM+LEN group.[11, 13] For the TPC ***** group, the most common AEs resulting in treatment interruption of treatment were . [13]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 56 of 137

Confidential

B.2.10.1.11. Dose reduction due to adverse events

The pembrolizumab dose was fixed at 200 mg Q3W and dose reduction was not allowed.[13] The starting dose for lenvatinib was 20 mg QD, however dose modifications were allowed according to the approved label and standard practice. The incidence of patients with AEs that led to dose reduction of lenvatinib was higher compared to dose reduction in the TPC group (66.5% vs 12.9%).[11]

The AEs that most frequently led to lenvatinib dose reduction (incidence >5%) were hypertension, diarrhoea, PPES, proteinuria, decreased appetite, fatigue, and weight decreased, all of which are known to be associated with lenvatinib.[11] The AEs that most frequently (incidence >1%) led to dose reduction of doxorubicin or paclitaxel were *****, all of which are known to be associated with chemotherapy.[13]

B.2.10.1.12. Safety overview

The safety profile of PEM+LEN is as expected and consistent with previously reported studies

At the first interim analysis, the safety results of the KEYNOTE-775 trial demonstrate that PEM+LEN offers a manageable and predictable AE profile specific to the individual products, and that this safety profile was consistent across patients enrolled to the trial.[11] These data were consistent with those observed during the phase I/IIb KEYNOTE-146 trial (Appendix F)[23] ; both studies demonstrate a similar safety profile.

The safety profiles observed were generally consistent with those expected for pembrolizumab and for lenvatinib as monotherapy agents.[11] Known AEs are generally managed with supportive medications and dose modifications, with approaches to manage any AEs for each monotherapy are well established through their usage and practice in prior indications. Based on these analyses, the safety profile of pembrolizumab in combination with lenvatinib can be considered well characterised.

B.2.10.2. KEYNOTE-146

A summary of safety data from KEYNOTE-146 are provided in Appendix F. These data were broadly consistent with those observed during the KEYNOTE-775 trial.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 57 of 137

Confidential

B.2.11. Ongoing studies

The KEYNOTE-775 trial is ongoing, with the final analysis database lock estimated in ***** (events dependent - see Table 7). There is also an active trial investigating the clinical benefits of PEM+LEN versus carboplatin and paclitaxel in adults with advanced or recurrent EC who have not received prior platinum-based chemotherapy outside of the adjuvant and/or neoadjuvant setting: LEAP-001.[24] Data from this trial are not expected *****

B.2.12. Innovation

Lenvatinib is a potent multiple RTK inhibitor that acts as an anti-angiogenic agent. Pembrolizumab is a potent and highly selective PD-1 inhibitor that acts as an immunotherapeutic agent (Table 2). The combination of an immunotherapeutic agent (pembrolizumab) with an anti-angiogenic agent (lenvatinib) is thought to remodel the tumour microenvironment, enhancing cancer cell recognition by the immune system and thus priming response to pembrolizumab.[25-27] These synergistic molecular effects allow a rapid, stronger and more durable tumour response, with the combination of PEM+LEN building on their individual agent effect.

The KEYNOTE-775 trial is the first randomised, controlled Phase III study evaluating a novel combination therapy in the previously treated advanced EC setting that has demonstrated positive results for both primary endpoints of OS and PFS across a broad range of participants.[13] Efficacy results demonstrate that treatment with PEM+LEN is superior to TPC, irrespective of a person’s stage, histology, or biomarker status. While the main health-related benefits of PEM+LEN will be captured in the quality-adjusted life year (QALY) calculation, the provision of an effective novel treatment option will provide additional benefits to patients, carers and healthcare professionals alike that are not captured in a QALY-only approach. This is especially important when considering the backdrop of historic unmet need for this population (see Section B.2.13.1).

The introduction of PEM+LEN to the current pathway of care would represent a stepchange in EC management, providing a novel treatment option with proven efficacy to a group of people with poor prognosis, for whom there is a significant unmet medical need and no current standard of care (see Section B.2.13.1).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 58 of 137

Confidential

B.2.13. Interpretation of clinical effectiveness and safety evidence

B.2.13.1. Summary of unmet medical need

Little progress has been made in EC management over the past decade, and there is no standard of care beyond platinum-based chemotherapy (carboplatin and paclitaxel) for advanced or recurrent EC, as acknowledged in the BGCS Guidelines.[1, 2, 12]

There is currently no consensus on a standard treatment for people with advanced or recurrent EC who have received prior systemic therapy, with limited efficacious treatments available for this patient population, and no treatment option demonstrating superiority over another in a robust clinical trial setting.[12] People with advanced or recurrent EC who have disease progression on or following prior treatment with platinum-based chemotherapy are unlikely to live beyond a year with current treatment options (see Section B.2.13.4). This estimate is based on empirical data from KEYNOTE-775 and retrospective UK chart review (data on file), and supported by clinical expectations based on observations in real-world practice.[2, 11]

There is a clear unmet need for additional, novel treatment options with proven efficacy and manageable safety profiles for adults with advanced or recurrent EC, beyond platinum-based chemotherapy. PEM+LEN addresses this unmet need.

B.2.13.2. Principal findings from the evidence base

The KEYNOTE-775 trial provides the pivotal evidence supporting the use of PEM+LEN in people with advanced, or recurrent EC who have disease progression on or following prior treatment with a platinum-containing therapy and who are not candidates for curative surgery or radiation.

Efficacy results from the KEYNOTE-775 trial demonstrate that treatment with PEM+LEN is superior to TPC and provides a statistically significant and clinically meaningful improvement in OS, PFS and ORR.[11] Kaplan–Meier curves for PFS and OS can be seen to separate early and definitively, remaining consistently separated through the duration of study follow-up. The PFS rate at 2 years was more than five times higher in the PEM+LEN arm compared with the TPC arm (20.9% versus 3.8%), and the OS rate at 2 years was twice as high (42.0% vs 21.4%).[13] This is despite extensive use of subsequent systemic anticancer therapy in the TPC arm that is not available to patients in the UK. The ORR rate

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 59 of 137

Confidential

was also twice as high in the PEM+LEN arm compared with the TPC arm (31.9% vs 14.7%) with an estimated difference of 17.2%.[11]

The KEYNOTE-146 trial provides supportive evidence of the longer-term benefit of PEM+LEN in people with pre-treated metastatic EC, demonstrating an unprecedented 5- year survival rate for patients treated with PEM+LEN of approximately 30%.[23] This is discussed further in Section B.3.3 and Section B.3.8.

Safety results across the KEYNOTE-775 and KEYNOTE-146 trials showed PEM+LEN to offer a manageable safety profile, with most adverse events (AEs) resolved with supportive medication, dose interruption or reduction of lenvatinib and less commonly, with dose interruption of pembrolizumab.[11, 17] Safety profiles observed were generally consistent with those expected for pembrolizumab and lenvatinib as monotherapy agents, for which approaches to manage AEs are well established.

Importantly, HRQL data collected in KEYNOTE-775 show no clinically meaningful drop in patient quality of life during treatment with PEM+LEN, and there were no differences in HRQL for participants treated with PEM+LEN versus chemotherapy.[16]

B.2.13.3. Strengths and limitations of the evidence base

B.2.13.3.1. Study design

The KEYNOTE-775 trial is a phase III RCT of high-quality design and conduct with several steps taken to minimise the risk of bias. This included the restriction of dMMR participants and a staged statistical analyses approach to ensure ITT data were not positively biased in favour of the PEM+LEN arm.

Due to the differences in dosage and administration between pembrolizumab, lenvatinib, doxorubicin and paclitaxel, KEYNOTE-775 is an open-label study.[11] This study design (where investigators and patients were not blinded to treatment assignment) may introduce bias to which subjective measures, such as PROs, could be particularly susceptible to. To counteract this, treatment randomisation occurred centrally using an interactive response technology (IRT) system. Patients were stratified according to MMR status, and within the pMMR stratum, patients were further stratified according to ECOG performance status, geographic region, and prior history of pelvic radiation. Furthermore, efficacy endpoints as per RECIST 1.1 were determined by a blinded independent central investigator through BICR assessment.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 60 of 137

Confidential

B.2.13.3.2. Study applicability to clinical practice

The KEYNOTE-775 trial provides data of direct relevance to the decision problem with regard to population, comparator and outcomes.

B.2.13.3.2.1. Population

Patients included in the KEYNOTE-775 trial had advanced, recurrent or metastatic EC and had disease progression on or following prior systemic, platinum-based chemotherapy treatment.[11] This is consistent with the most common position of use for PEM+LEN in clinical practice.

A broad range of participants were enrolled in terms of histology, MMR status and treatment history with PEM+LEN providing clinically meaningful and durable responses across all patients.[11] As such, biomarker testing is not a barrier to treatment, potentially resulting in a faster initiation. As the prognosis for patients with unresectable advanced or recurrent EC is particularly poor, earlier treatment initiation may lead to a significant impact on patients’ survival and overall quality of life.

The median patient age in the KEYNOTE-775 trial was 64 years[11] , which is generally aligned with the median age of UC patients in England, where the majority of patients diagnosed in 2019 were aged ≥ 55 years.[5] Clinical expert consultation suggested trial participants are slightly younger than UK patients and that the proportion of patients with dMMR is slightly higher in the UK, but they noted that baseline characteristics are generally representative of UK clinical practice.[2]

It should be acknowledged that there is a current evidence gap for PEM+LEN in adults with advanced or recurrent EC who have not received prior platinum-based chemotherapy outside of the adjuvant and/or neoadjuvant setting. This evidence gap will be filled by the ongoing LEAP-001 trial (see Section B.2.11).

B.2.13.3.2.2. Comparator

In the absence of a standard of care beyond platinum-based chemotherapy for advanced or recurrent EC, the TPC arm of KEYNOTE-775 was chosen to represent clinical practice and is considered generalisable to the decision problem being addressed, with TPC treatments representing the majority share of real-world treatment use in the UK (data on file).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 61 of 137

Confidential

This trial was not powered to analyse results by treatment option, as both doxorubicin and paclitaxel represent ‘standard’ clinical practice. Investigators must have selected and recorded the TPC option prior to randomisation, in the event that the participant was assigned to the TPC arm, limiting any selection bias.

B.2.13.3.2.3. Outcomes

The outcomes measured in KEYNOTE-775 can be considered relevant to patients, carers and healthcare professionals alike. Dual primary efficacy endpoints of PFS and OS reflect well established trial endpoints and monitoring endpoints adopted in clinical practice, but also outcomes that have the most impact on patients and carers. HRQL endpoints further assess the impact of disease on patients and also allow formal utility analyses to support economic modelling.

Current survival data from KEYNOTE-775 are limited to IA1 but show OS improvements with PEM+LEN that are statistically significant and clinically meaningful.[11] The OS HR of 0.62 (95% CI: 0.51, 0.75; p<0.0001) crossed the pre-specified boundary for statistical significance at IA1 of ≤0.0064. The KM curves for OS clearly separated, starting at around 3 months and remained consistently separated throughout the duration of the evaluation period. This was despite high use of subsequent anticancer therapy in the TPC group, including subsequent treatment with PD1/PDL1 regimens that are not currently available in the UK (see Section B.3.5.4) that likely dilutes the real-world survival benefit PEM+LEN offers.

Interim survival data from KEYNOTE-775 are supported with longer-term data from the earlier phase KEYNOTE-146 trial, that clearly demonstrate the longer-term benefit of PEM+LEN treatment for advanced or recurrent EC with an unprecedented 5-year survival rate of approximately 30%.[23] To put this into context, the 2-year survival rate with current care is around 20% (Table 23).

B.2.13.4. End-of-life criteria

PEM+LEN should be considered an end-of-life treatment, meeting the NICE criteria for such designation, as summarised in Table 23.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 62 of 137

Confidential

Table 23: End-of-life criteria

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 63 of 137

Confidential

==> picture [29 x 14] intentionally omitted <==

Cost effectiveness

B.3.1. Published cost-effectiveness studies

No relevant studies were identified in a systematic search for cost-effectiveness analyses of PEM+LEN for patients with advanced, metastatic, or recurrent EC who have received prior systemic therapy (reported in Appendix G). Based on an additional hand-search of economic models in the published domain, one abstract was identified that demonstrated the cost-effectiveness of PEM+LEN versus chemotherapy in the Swedish setting. The same model structure as per this submission with the Sweden specific inputs. Therefore, it is not informative for decision making in the UK.[28]

At the time of writing this submission, there were no published NICE appraisals for treatments for advanced, metastatic or recurrent EC. NICE have very recently recommended dostarlimab for a group of patients with previously treated advanced or recurrent EC with MSI-H or dMMR (ID3802, 08 February 2022). This appraisal was based on a single arm study (GARNET [an ongoing phase 1, open-label, single-arm, multicentre study of the efficacy and safety of dostarlimab]) and supplementary UK RWE conducted by the manufacturer. Given the proximity of the publication to this submission, there was not sufficient time to review the available evidence from ID3802 in full. A top-level assessment of data available from ID3802 showed that all efficacy and HRQL data have been redacted from the public domain, and it is therefore not informative for this appraisal of PEM+LEN.

Published NICE appraisals in similar gynaecological cancers were considered, where relevant, to inform the approach taken for the economic evaluation for PEM+LEN. An expanded targeted literature review of NICE technology appraisals considering uterine, cervical and ovarian cancers was conducted, and economic evaluation methods of published NICE appraisals are presented in Table 24.

B.3.2. Economic analysis

As no published cost-effectiveness studies relevant for the UK were identified by the SLR and no published NICE appraisals in EC, an economic model was developed to assess the cost-effectiveness of PEM+LEN. Previous NICE appraisals in gynaecological cancers were considered to inform the approach taken for the economic evaluation (Table 24).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 64 of 137

Confidential

B.3.2.1. Patient population

PEM+LEN is indicated for the treatment of advanced or recurrent endometrial carcinoma in adults who have disease progression on or following prior treatment with a platinumcontaining therapy in any setting and who are not candidates for curative surgery or radiation, as described in Section B.1.2. The economic analysis addresses this patient population directly in line with the decision problem, which is consistent with the KEYNOTE-775 trial population and the final scope issued by NICE.[29]

B.3.2.2. Model structure

The economic model developed to assess the cost-effectiveness of PEM+LEN in this indication follows a partitioned survival modelling approach. Previous NICE technology appraisals (TAs) in gynaecological cancers have included variations of the standard threestate partitioned survival structure (Table 24), where state membership is defined by disease status (progression-free [PF] versus progressed disease [PD]) and mortality status (Section B.3.1). Importantly, the partitioned survival modelling approach facilitates direct use of clinical trial evidence available from KEYNOTE-775, including the primary outcomes of progression-free survival (PFS) and overall survival (OS) endpoints. The cohort model structure accurately captures survival and HRQL implications for patients and cost and resource use implications for the National Health Service (NHS) (as reported in Sections B.3.4 and B.3.5, respectively), in line with the NICE reference case.[30]

Figure 10 illustrates the health states and possible transitions in each model treatment arm. The health states capture disease progression status (PF or PD) and treatment status (on or off treatment). Treatment-dependent costs and health outcomes associated with each arm are captured within each mutually exclusive heath state.

Patients with advanced or recurrent EC who have received prior systemic therapy enter the model in the PF state and are assumed to be on treatment. In each model cycle those in the PF stage can either remain in the PF state or move into the PD or death state. Those in the PD state can remain in the PD state or move into the death state. Death is included as an absorbing health state.

Additionally, the following adjustments are applied to maintain logical consistency in the patient flow of the model:

• The mortality risk at each model cycle is bound by age-matched general population predictions, sourced from the latest available Office for National Statistics Life Tables[31]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 65 of 137

Confidential

• A limit is built into the model to ensure that PFS cannot exceed OS. The limit is applied to the per cycle hazard of progression/death and hazard of death; if the hazard of death exceeds that of progression/death, the maximum hazard is assumed. This only occurs towards the end of the time horizon, after there are approximately 10% of PEM+LEN patients and 0% of TPC patients remaining alive. In addition, the hazard of death on both arms is capped by the hazard of death experienced by the general population.

• By the same logic, time on treatment (TOT) also cannot exceed OS; if TOT is estimated to be greater than OS at any time on any model arm, TOT is assumed to be equal to OS

Figure 10: Economic model structure

==> picture [448 x 102] intentionally omitted <==

==> picture [448 x 103] intentionally omitted <==

A 1-week cycle length is considered sufficiently short to reflect dosing regimens and accurately capture key clinical outcomes for patients. Given the short cycle length, a halfcycle correction is not applied to any cost or health outcomes. Table 24 summarises key features of the economic analysis, alongside corresponding features of completed appraisals in other gynaecological appraisals.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 66 of 137

Confidential

Table 24: Features of the economic analysis

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 67 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 68 of 137

Confidential

B.3.2.3. Intervention technology and comparators

B.3.2.3.1. Intervention

PEM+LEN combination treatment is implemented in the analysis according to the EMA and MRHA marketing authorisation described in Section B.1.2, and the KEYNOTE-775 trial (please see Section B.2.3.1 and Section B.2.4 for further information about PEM+LEN and the clinical trial protocol).

In KEYNOTE-775, the maximum treatment duration of pembrolizumab was 2 years from first dose (Section B.2.3.1). The cost-effectiveness analysis incorporates this treatment rule as described in Sections B.3.3 to B.3.5. The dose and administration schedule for PEM+LEN follows details obtained from the SmPC for PEM+LEN and the KEYNOTE-775 trial protocol, as summarised below[3, 14] :

• Pembrolizumab is administered at a dose of 200 mg once every 3 weeks until progression or for up to a maximum of 35 cycles (2 years)

• Lenvatinib is administered at a dose of 20 mg daily until progression

Scenario analysis investigates pembrolizumab administered at a dose of 400 mg once every 6 weeks until progression or for up to a maximum of 35 cycles (2 years). Scenario results are presented in Section B.3.9.2.

B.3.2.3.2. Comparators

As described in Section B.3.5.1, the most relevant comparators included in the costeffectiveness model are paclitaxel and doxorubicin. This is incorporated in the TPC arm in the economic model, where treatment is implemented using time-to-event data from the TPC arm of the KEYNOTE-775 trial (containing 25.5% paclitaxel and 74.5% doxorubicin). Both paclitaxel and doxorubicin treatments are used in UK clinical practice as described in Section B.1.3.2.

The dose and administration schedule for TPC implemented in the model per KEYNOTE775 are summarised below[14] :

• Paclitaxel is administered at a dose of 80 mg/m[2] once every 3 weeks until progression

• Doxorubicin is administered at a dose of 60 mg/m[2] 3 weeks in every 4 until progression or a lifetime cumulative dose of 500 mg/m[2]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 69 of 137

Confidential

Paclitaxel was administered once every 3 weeks in line with the KEYNOTE-775 trial regimen and expected regimen in clinical practice. The use of weekly paclitaxel is only supported by anecdotal evidence; based on the clinical efficacy in ovarian cancer and its good tolerability.[1] Therefore, no scenarios investigating alternative paclitaxel dosing patterns were investigated, although the impact of these scenarios is likely to be negligible.

The TPC arm in the model base case is the most relevant comparator for this appraisal, as it covers treatments used by the majority of patients in the UK. For completeness, we have explored the impact of additional scenarios:

• The NICE scope lists doxorubicin as a comparator. The base case in the economic model is considered conservative because it applies the lower cost of the generic formulary for doxorubicin based on the eMIT database. Clinical experts consulted for this appraisal confirmed that the more expensive branded liposomal/pegylated doxorubicin (Caelyx®) is primarily used in UK clinical practice. Caelyx is considered to have equivalent effectiveness but is preferred to non-pegylated doxorubicin due to being associated with a better toxicity profile.[2] The base case cost-effectiveness analysis uses generic non-pegylated doxorubicin, and, when using the more expensive costs of Caelyx® in scenario analysis to better reflect the type of doxorubicin used in the UK, the incremental cost-effectiveness ratios (ICERs) slightly improve in favour of PEM+LEN (Section B.3.5.1)

• An exploratory scenario is included which incorporates carboplatin in combination with paclitaxel (as re-challenge) and carboplatin monotherapy – in addition to paclitaxel and doxorubicin – in a mixed chemotherapy comparator arm.[1, 2] This scenario is included in recognition of a small number of UK patients who may be eligible to receive further platinum-based chemotherapy following initial response to treatment. As there are no robust data for these treatments in previously treated advanced EC, it was necessary and appropriate to assume equivalent efficacy between the mixed chemotherapy options and TPC arm of KEYNOTE-775. Treatment costs are weighted according to their expected use in clinical practice as informed by the ECHO study (as detailed in Section B.2.9.3) (*****% paclitaxel, *****% doxorubicin, *****% carboplatin, and *****% carboplatin plus paclitaxel).[22] The proportions used exclude investigational treatments and those that are not currently reimbursed in the UK. Results of these scenarios are presented in Section B.3.8.2

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 70 of 137

Confidential

• Further scenario analyses investigate paclitaxel (as part of the TPC and mixed chemotherapy analysis) administered at a dose of 80 mg/m[2] every week until progression and the inclusion of a paclitaxel stopping rule after 6 cycles, based on feedback from UK clinical experts.[2]

Finally, hormone therapy and best supportive care in this setting are palliative in nature, primarily used to relieve symptoms for patients without translating into any survival benefits. Patients who can only receive symptom-relief are unlikely to be fit enough for any active anti-cancer treatment, and therefore the treatments described are not considered as appropriate comparators for this appraisal.

The approval of PEM+LEN in this indication would represent a step change improvement in the clinical care pathway for patients with advanced EC. PEM+LEN offers an alternative mechanism of action to currently available chemotherapies and has demonstrated longterm effectiveness in KEYNOTE-775 and KEYNOTE-146.

B.3.3. Clinical parameters and variables

B.3.3.1. Overview of clinical data sources and outcomes in the economic model

The following clinical outcomes are included in the economic model:

•  OS

•  PFS

•  TOT

• HRQoL

•  AEs

The clinical outcomes used to inform the economic analysis are based on patient-level data from the Phase III KEYNOTE-775 trial investigating PEM+LEN versus TPC in the base case. This trial provides the most robust and best available evidence for conducting the comparative cost-effectiveness analysis.

B.3.3.2. Approach to time-to-event analysis – KEYNOTE-775

Key efficacy outcomes (OS, PFS and TOT) for PEM+LEN and TPC were modelled using patient-level data from KEYNOTE-775 (data cut-off date of 26 October 2020). The median duration of follow-up was approximately 12 months in the PEM+LEN arm and

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 71 of 137

Confidential

approximately 11 months in the TPC arm.[13] In line with the NICE reference case, to assess the cost-effectiveness of PEM+LEN over a lifetime horizon it was necessary to extrapolate the patient-level data beyond the trial period.[30, 38, 39]

The methods used to extrapolate OS, PFS and TOT followed guidance outlined in NICE Decision Support Unit (DSU) Technical Support Documents (TSDs) 14 and 21.[38, 39] As patient-level data are available for both PEM+LEN and TPC in KEYNOTE-775, and there are a large number of observations for each arm, it is appropriate to extrapolate survival outcomes using individually fitted curves for each trial arm. The parametric models used in the base-case analysis were assessed systematically for each endpoint, based on the following approach:

• Following NICE DSU TSD 14, six parametric models (exponential, Weibull, log-normal, log-logistic, Gompertz and generalized gamma) were fitted to the observed data from KEYNOTE-775 and assessed for suitability considering[38] :

  • Visual fit to the observed Kaplan–Meier (KM) data within the trial period for KEYNOTE-775

  • Assessment of the underlying hazard functions

  • Statistical goodness of fit indicated by Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values

• Additional flexible two-piece KM plus parametric models were explored where necessary, in accordance with NICE DSU TSD 21, and assessed for suitability based on the same process[39]

• Clinical validation of extrapolated models was sought from UK clinical experts, and extrapolations were compared against relevant evidence from the KEYNOTE-146 trial (see Section B.2.6.2 for further details)

The most appropriate and clinically plausible models for OS, PFS and TOT were used in the base case analysis, with alternative clinically plausible models tested in scenario analyses. Models in the base case and scenarios are summarised in Table 25 for OS and PFS and Table 26 for TOT. Full details are provided in Section B.3.3.3 for OS, Section B.3.3.4 for PFS, and Section B.3.3.5 for TOT.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 72 of 137

Confidential

Table 25: Summary of OS and PFS models selected for economic analysis (PEM+LEN and TPC)

Key : KM, Kaplan–Meier; PFS, progression-free survival; OS, overall survival; PEM+LEN, pembrolizumab with lenvatinib; TPC, treatment of physician’s choice. Notes : Alternative plausible assumptions tested in scenario analyses; given there were no other plausible OS curves, minimal scenarios for OS were included (see Appendix R for further detail)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 73 of 137

Confidential

Table 26: Summary of TOT models selected for economic analysis (PEM+LEN and TPC)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 74 of 137

Confidential

B.3.3.3. Overall survival

Figure 11 shows the OS KM data for all patients in KEYNOTE-775 and the corresponding underlying number at risk over time in the PEM+LEN and TPC arms of the trial. There is a clear separation in the KM plots from 3 months onwards, with increasing improvements in survival benefit for PEM+LEN that is sustained over the entire remaining observed period. Based on visual inspection, OS in the PEM+LEN arm appears to plateau at approximately *****% survival from 21 months onwards while there remained a substantial number of patients at risk. This is considerably higher than in the TPC arm, where approximately *****% of patients remained alive from the same time point. Figure 11 also includes the 5-year follow-up data observed in KEYNOTE-146, which demonstrates the prolonged long-term benefit of PEM+LEN in this indication. This is anticipated to continue to benefit patients throughout their lifetime.

Table 27. Number of events and level of maturity of OS in KN-775

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 75 of 137

Confidential

Figure 11: PEM+LEN and TPC – OS, KM plot

==> picture [385 x 264] intentionally omitted <==

Key: n, number; OS, overall survival; PEM+LEN, pembrolizumab with lenvatinib; TPC, treatment of physician’s choice.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 76 of 137

Confidential

As NICE TSD 14 states, generally, when patient-level data are available, it is unnecessary to rely upon the proportional hazards assumption and apply a proportional hazards modelling approach. Nonetheless, the proportional hazards assumption was tested for completeness.[38] Details of the Schoenfeld residuals and log-cumulative hazard plots are provided in Appendix P which confirm that it is appropriate to extrapolate OS outcomes based on individually fitted curves for each trial arm.

An overlay of the independent one-piece parametric models and observed KM data from KEYNOTE-775 and KEYNOTE-146 are shown in Figure 12 for PEM+LEN (see Appendix P for further details). Based on assessment of the single-fitted curves against the observed data from KEYNOTE-775 and longer-term study KEYNOTE146 (approximately 5-year follow-up), it is clear that the independent one-piece parametric survival curves are inappropriate for decision-making and should be disregarded on the basis of clinical implausibility:

• Observed long-term data from KEYNOTE-146 demonstrate that for patients treated with PEM+LEN, the 4 to 5-year survival rate is likely to be around 30% while the highest survival rates predicated by any independent one-piece parametric model was just 17%; this substantially underestimates the observed benefit of PEM+LEN (Figure 12)[17, 40]

• The hazard of death associated with each parametric survival model further confirms the importance of exploring more accurate methods of modelling lifetime survival across both arms, as the independent one-piece models are unable to capture the complex hazard profile, particularly beyond 26 weeks (Figure 13 and Figure 14, see further details provided in Appendix P)

  • The observed hazard of death increases until these timepoints before decreasing for the remaining observed period for PEM+LEN and plateauing for TPC; it is clear that the implied hazard profiles from some (most) of the parametric curves fit poorly to this shape for both PEM+LEN and TPC

  •  The difference between the hazards based on the observed data and independent one-piece parametric fits is greater at later points in time in both arms, which emphasises the importance of exploring more accurate methods of modelling lifetime survival across both arms

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 77 of 137

Confidential

Figure 12: OS independent one-piece parametric survival curves for PEM+LEN

*** Key** : KM, Kaplan–Meier; OS, overall survival; PEM+LEN, pembrolizumab with lenvatinib.

Figure 13: PEM+LEN OS hazard function

*** Key:** OS, overall survival; PEM+LEN, pembrolizumab with lenvatinib. Note: The shaded region refers to 95% confidence intervals for the smooth spline estimates.

Figure 14: TPC OS hazard function

*** Key:** OS, overall survival; TPC, treatment of physician’s choice. Note: The shaded region refers to 95% confidence intervals for the smooth spline estimates.

B.3.3.3.1. Two-piece extrapolation

Alternative survival models were explored to improve the validity of OS extrapolations for PEM+LEN and TPC. Consistent with guidance published in NICE DSU TSD 14 and 21, independent two-piece survival models (i.e. KM plus parametric extrapolation) were analyzed.[38, 41]

The independent two-piece models provided a substantially better fit to the observed data and are therefore preferable for the base-case analysis. An overlay of the independent two-piece parametric models and observed KM data from KEYNOTE775 are shown in Figure 15 for PEM+LEN and Figure 16 for TPC. KEYNOTE-146 OS data have also been included in Figure 15 to inform the validity of long-term survival estimates. The AIC and BIC statistics corresponding to the independent twopiece parametric models fitted to KEYNOTE-775 are provided in Table 28.

Selection of the KM cut-off

Following visual inspection of the hazard of death for the PEM+LEN and TPC arm, a clear inflection point appears by 30 weeks indicating an appropriate timepoint for switching from KM to extrapolated data for the independent two-piece models. This observation is validated by supplementary Chow tests which suggest that the likely inflection point in OS is around 26 weeks, as shown in Appendix P It is also generally preferable to select earlier cut-off time points to retain as much statistical power as Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 78 of 137

Confidential

possible for the two-piece survival analyses, as the number of patients at risk naturally decreases over time.

Therefore, 26 weeks was selected as an appropriate cut-off, where 337/411 and 300/416 patients were at risk on the PEM+LEN and TPC arms, respectively. Scenarios with different cut-offs were explored to understand the impact of this decision on the results, as shown in Section B.3.8.3.

Figure 15: OS independent two-piece parametric survival curves for PEM+LEN

*** Key** : KM, Kaplan–Meier; OS, overall survival; PEM+LEN, pembrolizumab with lenvatinib.

Figure 16: OS independent two-piece parametric survival curves for TPC

*********

Key : ECHO, endometrial cancer health outcomes study; KM, Kaplan–Meier; OS, overall survival; TPC, treatment of physician’s choice; UK, United Kingdom.

Table 28: Fit statistics of overall independent two-piece survival extrapolation

Selection of the distribution for extrapolation beyond 26 weeks

Following determination of the 26-week cut-off for OS as an appropriate timepoint for the independent two-piece models, the distribution used for the extrapolated portion was partly informed by a comparison of the hazard of death associated with each

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 79 of 137

Confidential

parametric survival model and the observed hazards (Figure 17 and Figure 18, respectively). Full details are provided in Appendix P.

The log-logistic model exhibits the closest properties to the observed hazard of death for PEM+LEN (decreasing hazards) and provides a plausible fit to the observed longer-term data from KEYNOTE-146. The exponential, Weibull and generalized gamma models do not adequately capture the complex hazard trend of PEM+LEN over the observed trial period and should be discounted for PEM+LEN extrapolation, as explained below and summarised in Appendix R:

• The hazard for TPC remains constant up to a time point of ~75 weeks (Figure 18). Although the observed hazard decreases after 75 weeks, this is largely attributed to low remaining numbers at risk in the data leading to variability in the results (Table 29). Simpler models, such as the exponential and Weibull, exhibit similar properties to the observed hazard of death for TPC over the period between 26 to 75 weeks, where there are a sufficiently large number of patients remaining at risk to inform the analysis. This assessment indicates that the log-logistic function is appropriate for PEM+LEN and the exponential function is appropriate for TPC

• When comparing the observed hazard of independent one-piece models (Figure 13 and Figure 14) with independent two-piece models (Figure 17 and Figure 18), the hazard of death associated with independent two-piece models over time appear to fit better to the remaining observed data than their independent onepiece counterparts. Independent one-piece models overfit to the initial steep increasing section of the smoothed spline hazard estimate, leading to an overestimation of long-term hazards. When cutting the data at 26 weeks, we simplify the survival section used to inform long term extrapolations, better fitting to medium- and long-term hazards

The survival models in the TPC arm were also validated using real-world survival data based on the UK cohort of the ECHO study (Section B.2.9.3), as shown in Figure 16. This demonstrated high consistency with outcomes from the TPC arm of the KEYNOTE-775 trial. As discussed in Section B.2.9.3, median OS from the start of second line of systemic therapy was ***** months in ECHO compared with a median OS of 11.4 months in KEYNOTE-775. Of note, ECHO included some

patients who received investigational treatments not routinely available in UK clinical Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 80 of 137

Confidential

practice as a subsequent treatment (such as PD-1/PD-L1 and VEGF/VEGFR inhibitors). Based on this, it is likely that the ECHO OS data may overestimate the ‘true’ comparative real-world outcomes for patients treated with second-line systemic therapy (TPC), and it is also expected that there are some differences compared to the TPC arm in KEYNOTE-775 which did allow subsequent line treatment PD-1/PDL1 and VEGF/VEGFR inhibitors.

Figure 17: PEM+LEN OS hazard function, with breaking point at Week 26

*** Key:** OS, overall survival; PEM+LEN, pembrolizumab with lenvatinib. Note: The shaded region refers to 95% confidence intervals for the smooth spline estimates.

Figure 18: TPC OS hazard function, with breaking point at Week 26

*** Key:** OS, overall survival; TPC, treatment of physician’s choice. Note: The shaded region refers to 95% confidence intervals for the smooth spline estimates.

Table 29: Number of patients at risk, KN-775

Following guidance in NICE DSU TSD 14, there is strong justification for the use of different survival distributions for each arm in the model.[38] The different mechanism of action for PEM+LEN and TPC are expected to translate into different survival trajectories and hazard profiles for patients:

• PEM+LEN is the first and only immuno-oncology (IO) drug to demonstrate superior efficacy based on Phase III clinical trial evidence in advanced, metastatic, and recurrent EC. Currently, there are no standard treatment options with highly limited treatment benefit associated with chemotherapies. There is clinical rationale supporting different hazard profiles for IO therapies compared with standard chemotherapy options (Section 0)

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 81 of 137

Confidential

• Immunotherapy differs from conventional anticancer treatments such as chemotherapy because it treats the immune system, and this is expected to translate into differential outcomes

• The anti-angiogenic effect of lenvatinib (multi-TKI) in combination with the immune-stimulatory effect of pembrolizumab (anti-PD-1) results in a tumour microenvironment that may help overcome primary and acquired resistance to immunotherapy, improving tumour responses compared to either treatment alone. In preclinical murine models, PD-1 plus TKI inhibitors have demonstrated enhanced anti-tumour activity compared to either agent alone[26, ] 42

• In line with outcomes observed with pembrolizumab across multiple advanced malignancies, and with other IOs with similar mechanism of action, PEM+LEN is expected to offer a sustained treatment effect that is distinct from conventional chemotherapy options in EC.[43, 44] KEYNOTE-146 provides evidence of long-term effect after 5 years of follow-up, although the duration of sustained treatment effect beyond that time period is less clear and the impact on the results could be tested

• Standard chemotherapy regimens are not expected to improve long-term outcomes, offering limited survival benefit for patients in the short-term only as is highlighted by the currently poor survival outcomes in this indication

• Clinical experts consulted for this appraisal support the above expectations of long-term benefit associated with PEM+LEN in this patient population

• Based on Phase III randomised clinical trial evidence, PEM+LEN has already demonstrated the potential to improve long-term survival in KEYNOTE-775. The extent of benefit observed with longer follow-up in KEYNOTE-146 has not been previously seen in this indication

• The observed underlying hazard profiles of PEM+LEN and TPC from KEYNOTE775 follow distinct and different trajectories (Figure 12, Figure 13, Figure 17 and Figure 18). One major difference is that the hazards in the PEM+LEN arm have a strong decreasing trend after 26 weeks, which does not occur in the TPC arm

Finally, the clinical plausibility of long-term extrapolations was validated by oncologists experienced in endometrial cancer treatment in the UK.[2] As part of the

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 82 of 137

Confidential

clinical validation process, the goodness of fit statistics, survival curves and estimates of long-term survival were presented for each extrapolation. All participants were more comfortable predicting plausible extrapolations for the TPC arm, given their experience in treating patients with chemotherapy regimens in this treatment area.[2] In particular:

• TPC arm: Due to some participants in the TPC arm of KEYNOTE-775 receiving subsequent PD-1/PD-L1 and VEGF/VEGFR inhibitor therapies that are not routinely available in the UK, the OS estimates may be overestimated in the trial. This could underestimate the incremental benefit of PEM+LEN versus TPC although it is not possible to test and adjust for the potential impact without introducing substantial uncertainty in the analysis. The bottom group of curves in Figure 16 provide a more plausible estimate of survival reflecting current UK clinical practice, with the exponential or Weibull extrapolations fitting well to the observed KEYNOTE-775 data. Furthermore, the hazard of death is not expected to reduce or change dramatically over time

• PEM+LEN arm: The log-logistic extrapolation fits well to the observed KEYNOTE775 and longer-term KEYNOTE-146 data and provides a plausible estimate of long-term survival. Furthermore, a reduction in hazard for patients treated with PD-1/VEGFR inhibitor combination treatments such as PEM+LEN would be expected

B.3.3.3.2. Summary of modelled extrapolations

The most appropriate and clinically plausible models for OS are used in the basecase analysis, shown in Figure 19.

For PEM+LEN, KEYNOTE-775 KM data were used directly until a 26-week cut-off, after which the log-logistic function was fitted to OS time-to-event data reported in KEYNOTE-775. The KM + log-logistic model provides the most appropriate fit to PEM+LEN KEYNOTE-775 and KEYNOTE-146 survival, observed PEM+LEN hazard, and provides a clinically plausible long-term extrapolation.

For TPC, KEYNOTE-775 KM data were used directly until a cut-off at 26 weeks, after which the exponential function was fitted to OS time-to-event data reported in

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 83 of 137

Confidential

KEYNOTE-775. The KM + exponential model provides an appropriate fit to TPC KEYNOTE-775 survival and hazard and provides a clinically plausible long-term extrapolation.

Alternative plausible assumptions tested in scenario analyses. Given there were no other plausible OS curves, minimal scenarios for OS were included (see Table 25 and Section B.3.8.2 and Appendix R for a summary of the plausibility of the assessed curves).

Figure 19: Selected OS curve fits for PEM+LEN and TPC

*** Key** : ECHO, endometrial cancer health outcomes study; KM, Kaplan–Meier; OS, overall survival; PEM+LEN, pembrolizumab with lenvatinib; TPC, treatment of physician’s choice; UK, United Kingdom.

B.3.3.3.3. Mixed chemotherapy scenario

As mentioned in Section B.3.2.3.2, a scenario analysis is considered where carboplatin plus paclitaxel and carboplatin monotherapy are included alongside paclitaxel and doxorubicin in a mixed chemotherapy comparator arm. In the absence of efficacy data for these treatments in previously treated advanced EC, the mixed chemotherapy arm is assumed to have equal efficacy to the TPC arm as detailed in B.3.3.3.1 above. This approach was validated with UK clinicians.[2]

B.3.3.4. Progression-free survival

Figure 20 shows the PFS KM data for all patients in KEYNOTE-775 and the corresponding underlying number at risk over time in the PEM+LEN and TPC arm of the trial. There is clear separation in the KM plots from two months onwards, with sustained improvements in PFS benefit for PEM+LEN that is sustained over the entire remaining observed period. PFS in the PEM+LEN arm appears to plateau at approximately 21% from 18 months onwards. This plateau is considerably higher than in the TPC arm, where approximately 4% of patients remained progression free from the same time point with much fewer patients at risk.

Table 30. Number of events and level of maturity of PFS in KN-775

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 84 of 137

Confidential

Figure 20: PEM+LEN and TPC – PFS, KM plot

==> picture [447 x 312] intentionally omitted <==

Key: n, number; PEM+LEN, pembrolizumab with lenvatinib; PFS, progression-free survival; TPC, treatment of physician’s choice.

As with OS, the proportional hazards test for PFS also suggests that the proportional hazards assumption can be rejected given that the log-cumulative hazard of PEM+LEN and physician’s choice chemotherapy PFS cross. Details of Schoenfeld residuals and log-cumulative hazard plots for the PFS proportional hazards test are presented in Appendix P.

Following a similar analytical approach to the OS analysis of independent one-piece parametric curves in B.3.3.3, independent two-piece extrapolations were also preferred for PFS. This section provides details of the independent two-piece models

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 85 of 137

Confidential

relevant to the model base case; details of the independent one-piece models are available in Appendix P.

B.3.3.4.1. Two-piece extrapolation

A 10-week cut-off point was used, at which point there had been a large number of progression/death events, resulting in a steep initial drop off in the PFS KM which corresponded to the first scheduled radiological assessment of progression defined in the KEYNOTE-775 protocol (week 8 ± 7 days). Scenarios with different cut-offs were explored to understand the impact of this decision on the results, as shown in Section B.3.8.2.

An overlay of the independent two-piece parametric models and observed KM data from KEYNOTE-775 are shown in Figure 21 for PEM+LEN and Figure 22 for TPC. The AIC and BIC statistics corresponding to the independent two-piece parametric models fitted to KEYNOTE-775 are provided in Table 31. Hazard plots for PFS are presented in Appendix P.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 86 of 137

Confidential

Figure 21: PFS independent two-piece parametric survival curves for

PEM+LEN

==> picture [456 x 112] intentionally omitted <==

==> picture [456 x 112] intentionally omitted <==

==> picture [456 x 111] intentionally omitted <==

Key : KM, Kaplan–Meier; PEM+LEN, pembrolizumab with lenvatinib; PFS, progression-free survival.

Figure 22: PFS independent two-piece parametric survival curves for TPC

Key: KM, Kaplan–Meier; PFS, progression-free survival; TPC, treatment of physician’s choice.

Table 31: Fit statistics of PFS independent two-piece survival extrapolation

Confidential

When considering the observed PFS data for the TPC arm, there is minimal variability between PFS extrapolations given the maturity of the KEYNOTE-775 data. All independent two-piece models provide plausible fits.

When considering the observed data for the PEM+LEN arm, all independent twopiece models underestimate PFS after 1 year, and are therefore considered conservative. The Gompertz, generalized gamma, log-normal and log-logistic models may be considered a conservative, but plausible estimates of long-term PFS, given the appearance of a plateau at around 20% in KEYNOTE-775 PFS KM data after 1 year.

Clinical plausibility of the PFS extrapolations were validated. The goodness of fit statistics, survival curves and estimates of long-term PFS were presented to clinical experts to validate the plausibility of each extrapolation[2] :

• TPC arm: The experts agreed that all independent two-piece extrapolations seem to fit well to the observed KEYNOTE-775 data and provide plausible estimates of long-term PFS

• PEM+LEN arm: There was some variability in responses from clinical experts, suggesting that the exponential, log-normal, Gompertz or generalized gamma could provide plausible estimates

B.3.3.4.2. Summary of modelled extrapolations

The most appropriate and clinically plausible models for PFS are used in the base case analysis, shown in Figure 23.

For PEM+LEN, KEYNOTE-775 KM data were used directly until a 10-week cut-off, after which the log-logistic function was fitted to PFS time-to-event data reported in

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 88 of 137

Confidential

KEYNOTE-775. This provides the most plausible fit to the observed data and provides a clinically plausible conservative long-term extrapolation.

For TPC, KEYNOTE-775 KM data were used directly until a 10-week cut-off, after which the log-logistic function was fitted to PFS time-to-event data reported in KEYNOTE-775. The KM + log-logistic model provides a plausible fit to the observed data and provides a clinically plausible long-term extrapolation.

Alternative models tested in scenario analyses are summarised in Table 25 and Section B.3.8.2.

Figure 23: Selected PFS curve fits for PEM+LEN and TPC

*****

Key : KM, Kaplan–Meier; PEM+LEN, pembrolizumab with lenvatinib; PFS, progression-free survival; TPC, treatment of physician’s choice.

B.3.3.4.3. Mixed chemotherapy scenario

As mentioned in Section B.3.2.3.2 and Section B.3.3.3.1, in the absence of efficacy data for these treatments in previously treated advanced EC, the mixed chemotherapy arm is assumed to have equal efficacy to the TPC arm from KEYNOTE-775.

B.3.3.5. Time on treatment

TOT is modelled based on KEYNOTE-775 to determine the cohort of patients remaining on treatment at each model cycle to accurately accrue treatment-related costs. TOT for PEM+LEN is implemented individually in the PEM+LEN arm. Given the relatively short treatment duration and low treatment costs associated with paclitaxel and doxorubicin, implementing TOT data for these components individually would have minimal impact on the results, so they are not split in the TPC arm in favour of simplicity. Median TOT for pembrolizumab, lenvatinib and TPC are presented in Table 32.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 89 of 137

Confidential

Table 32: KN-775 median TOT

B.3.3.5.1. One-piece extrapolations

An overlay of the independent one-piece parametric models and observed KM data from KEYNOTE-775 are shown in Figure 24 for pembrolizumab, Figure 25 for lenvatinib and Figure 26 for TPC. AIC and BIC statistics corresponding to the parametric models fitted to KEYNOTE-775 are provided in Table 33.

Figure 24: TOT parametric curves for pembrolizumab (PEM+LEN arm)

*** Key** : KM, Kaplan–Meier; PEM+LEN, pembrolizumab with lenvatinib; TOT, time on treatment.

Figure 25: TOT parametric curves for lenvatinib (PEM+LEN arm)

*** Key** : KM, Kaplan–Meier; PEM+LEN, pembrolizumab with lenvatinib; TOT, time on treatment.

Figure 26: TOT parametric curves for TPC

*** Key:** KM, Kaplan–Meier; TOT, time on treatment; TPC, treatment of physician’s choice.

Table 33: Fit statistics of TOT extrapolation

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 90 of 137

Confidential

The generalized gamma model provides a plausible fit to observed data from KEYNOTE-775 for pembrolizumab and lenvatinib in the PEM+LEN arm. When considering the observed data for the pembrolizumab component in the first 2 years, the exponential, Weibull, and Gompertz also seemed plausible. These distributions also fit well to the observed data for the lenvatinib component, where all curves except the Gompertz provided plausible long-term extrapolations.

The generalized gamma model provides a plausible fit to the TOT data in the TPC arm of KEYNOTE-775, as well as the exponential, Weibull, and Gompertz distributions.

The goodness of fit statistics, survival curves and estimates of long-term TOT were presented to clinical experts to validate the clinical plausibility of each extrapolation[2] :

• PEM+LEN arm: Clinical experts suggested that if patients exhibit good response to PEM+LEN treatment they would stop treatment at 2 years, in line with the license for pembrolizumab and use in KEYNOTE-775 trial. Experts also suggested that the majority of patients would not be receiving treatment with lenvatinib after 10 months, but they would expect a small number to receive treatment beyond discontinuation of pembrolizumab

• TPC arm: Clinical experts noted that chemotherapy is typically administered for a maximum of six cycles (18–24 weeks) in UK clinical practice. As median TOT from KEYNOTE-775 KM data is 14.86 weeks, TOT estimates are in line with expectations in clinical practice. However, scenarios are investigated using this 6- cycle stopping rule for paclitaxel (and carboplatin plus paclitaxel and carboplatin monotherapy), further described in Section B.3.3.5.3 and results presented in Section B.3.8.3

B.3.3.5.2. Summary of modelled extrapolations

The most appropriate and clinically plausible models for TOT are used in the basecase analysis, shown in Figure 27.

For PEM+LEN, pembrolizumab and lenvatinib components were extrapolated separately to accurately capture the costs associated with each intervention. The

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 91 of 137

Confidential

generalized gamma function was fitted to pembrolizumab TOT time-to-event data reported in KEYNOTE-775, up to a maximum of 24 months. The generalized gamma function was also fitted to lenvatinib TOT time-to-event data reported in KEYNOTE775. The generalized gamma model provides a good fit for both components.

For TPC, the generalized gamma function was fitted to TOT time-to-event data reported in KEYNOTE-775. The generalized gamma model provides a plausible fit to TPC KEYNOTE-775 TOT and provides a clinically plausible long-term extrapolation.

Alternative models and extrapolation methods tested in scenario analyses are summarised in Table 25 and Section B.3.8.2.

Figure 27: Selected TOT curve fits

*** Key** : KM, Kaplan–Meier; PEM+LEN, pembrolizumab with lenvatinib; TOT, time on treatment; TPC, treatment of physician’s choice.

B.3.3.5.3. Treatment stopping rules

Treatment stopping rules are also included in the model base case in line with the administration of treatments in KEYNOTE-775.[13] Pembrolizumab treatment is limited to a maximum duration of 24 months (35 cycles); therefore, all patients in the PEM+LEN arm of the model were assumed to discontinue pembrolizumab treatment from 24 months onwards. Doxorubicin treatment is limited to a lifetime cumulative dose of 500 mg/m²; based on a dose of 60 mg/m[2] per administration, the maximum cumulative dose is reached at 5.75 months, after which the proportion of patients receiving doxorubicin in the TPC (or mixed chemotherapy in scenario analysis) arm is set to 0%.[14, 45]

A scenario was implemented to investigate the impact of implementing a maximum duration of six cycles of paclitaxel administration (affecting only costs in the model), in line with expected clinical practice. Given the short time on treatment and low cost of paclitaxel, this scenario has a negligible impact on cost-effectiveness results (Section B.3.8.2).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 92 of 137

Confidential

A further scenario was implemented to reflect the close relationship between progression status and treatment status and the expected use of PEM+LEN in clinical practice until treatment progression (applied in addition to the maximum duration of treatment outlined above). This scenario incorporates a treat-toprogression rule by ensuring that patients are not treated beyond progression at any model cycle (i.e. TOT cannot exceed PFS to maintain logic in the patient flow). The change slightly improved the ICER per quality-adjusted life year (QALY) for PEM+LEN versus TPC (Section B.3.8.2).

B.3.4. Measurement and valuation of health effects

It is well-documented that daily quality of life is severely impacted for women with advanced, metastatic, or recurrent EC. Common physical symptoms associated with aggressive disease and disease progression include pain and difficulty with urinating, unusual vaginal discharge, or bleeding (particularly after menopause), and pain during sexual intercourse (Section B.1.3.1). There is a substantial unmet need for tolerable and effective treatment options for patients in this indication, leading to a negative impact on emotional and mental well-being.

PEM+LEN is the first and only IO to demonstrate superior efficacy based on Phase III clinical trial evidence in advanced, metastatic, and recurrent EC. Currently, there are no standard treatment options, with highly limited evidence supporting treatment benefit with chemotherapies. PEM+LEN represents a step-change improvement in the treatment pathway for patients with advanced, metastatic or recurrent EC who have received prior systemic therapy (Section 0).

In line with the NICE reference case, and to incorporate the important impact on HRQoL described above, EQ-5D data collected from the KEYNOTE-775 trial were analysed and used in the economic model.[30] The model includes two approaches, as summarised below, and detailed in the following sections:

• Base case: A time-to-death utility approach captures patients’ HRQoL over time, which is clinically appropriate as time states closer to death are likely to be associated with worse HRQoL and a lower utility value. Utilising a time-to-death approach can remove the dependence on subjective clinical assessment of

progression status

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 93 of 137

Confidential

• Scenario analysis: A health-state utility approach based on progression assigns a value to each health state using data collected before and after progression events in the trial. It is logical that patients who are progression-free will generally experience a higher level of HRQoL than those who have progressive disease

Additionally, decrements to HRQoL due to AEs and natural decline of age-related HRQoL were considered in line with the NICE reference case[30]

B.3.4.1. Health-related quality-of-life data from clinical trials

The KEYNOTE-775 protocol specified patient completion of the EQ-5D-5L questionnaire[13] :

• For patients receiving PEM+LEN: On Day 1 of each 21-day study cycle, for the equivalent of 4 cycle lengths, and at the end of treatment (EOT) visit

• For patients receiving paclitaxel: On Day 1 of each 28-day study cycle, for the equivalent of 4 cycle lengths, and at the EOT visit

• For patients receiving doxorubicin: On Day 1 of each 21-day study cycle, for the equivalent of 4 cycle lengths, and at the EOT visit

Of note, completion of the HRQoL questionnaires following the EOT visit (i.e. posttreatment discontinuation) was not mandatory, although continued recording of questionnaire responses was encouraged via a Web Diary.

B.3.4.2. Mapping

Consistent with the current NICE position statement on the use of EQ-5D measurements in technology appraisals, EQ-5D-5L data from KEYNOTE-775 was analysed and mapped to EQ-5D-3L using the van Hout crosswalk, as described in Section B.3.4.1.[46]

Mixed effects regression models account for within- and between-patient variation and are therefore more representative of the range of HRQoL experienced by patient populations over time compared with a simple means-based approach. The regression analysis includes a random effect to account for the correlation between multiple observations from the same patient. Linear mixed effects regression models were formally fitted to the available data from KEYNOTE-775 to estimate utility

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 94 of 137

Confidential

values. Based on clinical drivers of HRQoL, the presence of AEs (Grade 3 and above), progression status (PF or PD) and patients’ time to death were included as explanatory variables (see Appendix P for further details).

B.3.4.2.1. Base case analysis: Utilities by time-to-death

Table 35 presents the results of the HRQoL analysis based on the presence of Grade 3 and above AEs and patients’ time to death (split into five categories), which reflects the known decline in HRQoL during the terminal phase of the disease. A time-to-death approach captures the decrease in utility as patients move closer to death, driven by the underlying impact of the disease over time, removing the dependence on clinical assessment of progression status. The utility values in the economic model are driven by the underlying impact of the disease over time based on patients’ time to death applied to both the PEM+LEN and TPC arms, independent of treatment choice. This approach has been previously accepted in other oncology appraisals by NICE, including but not limited to TA531 in metastatic squamous nonsmall cell lung cancer and TA357 in advanced melanoma.[43, 44]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 95 of 137

Confidential

Table 34: Model 2: Coefficients for HRQoL analyses

Table 35: Mean utility values based on time to death included in the economic model

B.3.4.2.2. Scenario analysis: Utilities by progression status

Table 36 presents the coefficients of the HRQoL analysis, which shows that utility values decrease with the presence of Grade 3 and above AEs and disease

progression, in line with expectations. The mean values are assumed to be driven by disease status irrespective of treatment choice and are applied to both PEM+LEN and TPC arms in the model (Table 37).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 96 of 137

Confidential

Table 36: Model 1: Coefficients for HRQoL analyses

Table 37: Model 1: Mean health state utility values applied in the economic

model (scenario analysis)

B.3.4.3. Health-related quality-of-life studies

A systematic search for published studies reporting relevant HRQoL data for patients with recurrent early-stage, advanced/metastatic or surgically unresectable endometrial cancer was conducted, and full details are presented in Appendix H.

No HRQoL studies relevant to the UK setting were identified for use in the economic model, and, as previously described, at the time of writing this submission there were no published NICE appraisals for treatments for advanced, metastatic or recurrent EC. Previous NICE appraisals in similar gynaecological cancers (uterine, cervical and ovarian) were hand-searched (Table 38), with the results demonstrating a high degree of consistency with the analysis in this appraisal, suggesting these values are robust and valid for decision-making.

The EQ-5D analysis from the KEYNOTE-775 trial remains the most relevant and robust source of data for this appraisal and follows methods outlined in the NICE reference case (see Section B.3.4.2).[13, 30]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 97 of 137

Confidential

Table 38: HRQoL data from previous NICE appraisals for the treatment of gynaecological cancers

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 98 of 137

Confidential

B.3.4.4. Adverse reactions

The impact of AEs on HRQoL is incorporated into the economic model. The occurrence of AEs and number of episodes of AEs per patient from KEYNOTE-775 were used to estimate the incidence probabilities. Treatment-specific Grade 3+ AEs with incidence of greater than 5% of patients on either arm of KEYNOTE-775 were included.[13] The number and proportion of patients who experienced AEs are shown in Table 39.[13] The mean number of AE episodes per patient is combined with the number of participants with one or more Grade 3+ AEs to capture recurrence and reflect the true number of each AEs experienced in KEYNOTE-775. The recurrence of AEs was assumed to be the same regardless of treatment received.

The duration of AEs and AE utility decrements were used to calculate a QALY decrement due to AEs applied at each model cycle. AE data are based on analysis of data from KEYNOTE-775, as shown in Table 39, with AE utility decrements described in Section B.3.4.1.

Scenario analysis demonstrated that the inclusion or exclusion of the AE utility decrements had minimal impact on the results (Section B.3.8.2).

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 99 of 137

Confidential

Table 39: Grade 3+ adverse events occurring in 5% or more patients in either arm in KN-775, and number of episodes per patient included in the economic model

Table 40: Adverse event durations

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 100 of 137

Confidential

AE utility decrements were informed using data from KEYNOTE-775 (Section 6.4.1.1). AE utility decrements were calculated by taking the “No Grade3+ AE” coefficient in the regression models as a decrement (Table 34 and Table 36). So, the AE utility decrement was ***** for the time to death approach in the base case and ***** for the health state approach in scenario analysis.

Applying the utility decrement of -0.040 for PEM+LEN and TPC to the cycle probability of each event produced AE cycle decrements of ***** and ***** for patients receiving PEM+LEN and TPC, respectively.

B.3.4.5. Health-related quality of life data used in the cost-effectiveness analysis

In line with the NICE reference case, trial-based utility values from KEYNOTE-775 are included in the economic model.[30] EQ-5D-5L data were collected in KEYNOTE775 and mapped to EQ-5D-3L values using public preference tariffs per the UK time trade-off (TTO) valuation set. The impact of AEs on HRQoL were also included. Additionally, at each model cycle, utility values are adjusted account for the natural decline in quality of life associated with age based on a standard published regression algorithm commonly used in NICE technical appraisals.[47]

Table 41 summarises the utility values used in the base case cost-effectiveness analysis.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 101 of 137

Confidential

Table 41: Summary of utility values used for cost-effectiveness analysis

B.3.5. Cost and healthcare resource use identification, measurement and valuation

A systematic search for published studies that reported cost and healthcare resource use data for patients with recurrent early-stage, advanced/metastatic or surgically unresectable endometrial cancer was conducted, and full details are presented in Appendix I. In total, 19 unique studies from 22 publications were identified. Of these, only three were applicable to the UK including two which were specific to England, however data were highly limited.[49-51] The studies were not deemed useful for the

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 102 of 137

Confidential

analysis as the type of resource use information were not published in sufficient detail for inclusion in the economic model, and the study populations were highly restricted compared with the patient population for which PEM+LEN is indicated:

• One study estimated the mean total cost of palliative care of advanced uterine cancer patients in the UK[49]

• A prospective cohort study conducted in England for patients with endometrial cancer estimated the average cost of treatment of Stage IV disease 5-years after diagnosis[50]

• A cost–consequence analysis based on a randomised controlled trial conducted in England estimated the mean healthcare cost associated with recurrent Stage 1 endometrial cancer patients[51]

The following cost categories are incorporated in the economic model, as described in this section:

• Drug acquisition costs

• Drug administration costs

• Health state resource use costs (e.g. ongoing monitoring and follow-up)

• AE costs

• Cost of testing

• Subsequent treatment costs

• End-of-life care costs

B.3.5.1. Intervention and comparators’ costs and resource use

B.3.5.1.1. Drug acquisition costs

Treatment costs are calculated based on the recommended dosing regimen for each drug for the modelled treatment duration detailed in Section B.3.3.5. The

recommended dose per administration, administration schedule and list prices for relevant treatments are presented in Table 42 (see Appendix Q for further PAS details). The total calculated drug acquisition costs per administration are also provided.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 103 of 137

Confidential

As previously described, PEM+LEN is implemented in the economic model according to the EMA and MHRA marketing authorisation and the KEYNOTE-775 trial protocol.[14] TPC is also implemented as per KEYNOTE-775 (see Section B.3.2.3). Treatments included in the mixed chemotherapy scenario analysis were implemented as per KEYNOTE-775 for paclitaxel and doxorubicin and the literature for carboplatin and carboplatin plus paclitaxel. Drug acquisition costs were based on the full recommended dose, which is a conservative approach that is likely to lead to higher incremental costs for PEM+LEN in favour of TPC as it does not account for dose interruptions and reductions over the treatment period that would be expected in clinical practice.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 104 of 137

Confidential

Table 42: Drug formulation, dose, administration, proportion of doses received and total drug acquisition cost per week, intervention and active comparators

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 105 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 106 of 137

Confidential

Key: AUC, area under the curve; eMIT, electronic market information tool; MIMS, monthly index of medical specialities; KN-775, KEYNOTE-775; Q3W, once every 3 weeks; Q6W, once every 6 weeks; RDI, relative dose intensity. Notes: Costs presented using list prices for all treatments.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 107 of 137

Confidential

B.3.5.1.2. Observed dosing data for lenvatinib (KEYNOTE-775)

In order to accurately estimate the total costs associated with the PEM+LEN arm, it is necessary to implement actual observed dosing data for the lenvatinib component from KEYNOTE-775. This is directly aligned to the currently approved label for lenvatinib (on which standard dosing for lenvatinib is based) and consistent with the administration of lenvatinib per the KEYNOTE-775 clinical trial protocol on dose reductions and modifications for optimal AE management. Clinical experts confirmed that this approach is valid (aligned with use of lenvatinib in practice).[2]

In the economic model, patients in the PEM+LEN arm started at a dose of 20 mg for lenvatinib. Patient-level dosing data from KEYNOTE-775 were used to calculate the total cost of lenvatinib per week as observed in KEYNOTE-775 (Appendix P), aligned with the realistic dose expected in the real-world setting.

B.3.5.1.3. Drug administration costs

In the economic model, drug administration costs are accrued for the duration of treatment in the PEM+LEN and TPC arms (Section B.3.3.5). The unit costs of the intravenous (IV) administration of pembrolizumab, paclitaxel and doxorubicin (and carboplatin and carboplatin plus paclitaxel in scenario analysis) were sourced from NHS reference costs (Table 43).[56] Lenvatinib is administered orally and is assumed to incur no cost.

For simplicity, pre-medication costs are not included in the analysis, leading to conservative estimates of cost-effectiveness for PEM+LEN compared with TPC. As mandated by the SmPC for paclitaxel, patients must be given corticosteroids, antihistamines and H2-receptor antagonists prior to paclitaxel administration, in order to prevent severe hypersensitivity reactions, and excluding these medications results in underestimating the costs associated with TPC in the model.[57]

Table 43: Administration costs

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 108 of 137

Confidential

B.3.5.2. Health-state unit costs and resource use

Costs associated with ongoing disease management, monitoring and patient followup are included in the economic model. Healthcare resources were included that were specific to each health state (i.e. PF or PD). Costs were applied to each resource and accrued according to the time spent in each health state. In line with the NICE reference case, the relevant unit costs were sourced from either the Personal Social Services Research Unit (PSSRU) or the NHS reference cost documentation and reflect 2019–2020 prices.[56, 58]

As there were no UK-specific resource use or cost studies identified in the systematic search or in published NICE technology appraisals in EC at the time of writing this submission (Section B.3.5 and Appendix I), resource use items and frequency of use were obtained by conducting hand searches of other health technology appraisals in related disease areas including uterine, cervical and ovarian cancers. HCRU used in this model has been accepted by NICE previously and UK clinical experts advised us of their generalisability for this indication.[2] As these searches showed a high degree of consistency in resource use values, the same values have been assumed in this appraisal for PEM+LEN versus TPC.[2]

Table 44 details the resource use and unit cost estimates for disease management used in the base case analysis, alongside supporting sources.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 109 of 137

Confidential

Table 44: Resource use and costs associated with model health states

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 110 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 111 of 137

Confidential

B.3.5.3. Adverse reaction unit costs and resource use

Table 45 shows the per-episode unit costs associated with resolving AEs included in the economic model, primarily sourced from NHS National Cost Collection Data Version 2 (2019/20).[56]

Total average AE management costs are calculated based on the incidence, recurrence and duration of Grade 3+ AEs that were observed in more than 5% of patients in KEYNOTE-775 as detailed in Sections B.2.9.3 and B.3.4.4. Costs are accrued based on the per-cycle probability of each AE in the PEM+LEN and TPC arm of the economic model, resulting in a calculated cost of £10.74 and £42.19 per cycle, respectively. For the mixed chemotherapy scenario analysis, AE rates are assumed equal to TPC as a simplifying assumption, so weekly costs for mixed chemotherapy was calculated to be £42.19 per cycle.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 112 of 137

Confidential

Table 45: AE costs

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 113 of 137

Confidential

B.3.5.4. Miscellaneous unit costs and resource use

B.3.5.4.1. Subsequent therapy

After progressing on PEM+LEN or TPC, some patients will receive subsequent anticancer treatment in NHS centres (see Section B.1.3.2). These costs are considered in the economic model as a calculated one-off cost, applied at the point of treatment discontinuation if alive at each model cycle. The total average cost per model cycle is based on the proportion of patients receiving subsequent therapies, the distribution of each subsequent treatment, drug acquisition and administration costs ( Table 47 ), and the observed duration of subsequent treatments associated with the PEM+LEN and TPC arms in KEYNOTE-775 (Table 48).

In KEYNOTE-775, 28% and 48% of patients received subsequent treatment in the PEM+LEN and TPC arms, respectively (Section Error! Reference source not found. and B.2.13.2).[11] The proportion of patients receiving each type of subsequent treatment in the base case analysis is presented in Table 46, informed by the observed use of treatments in KEYNOTE-775 excluding treatments that are not reimbursed in the UK setting. As discussed in Section Error! Reference source not found. , of the patients who received subsequent treatment in the TPC arm, *****% received subsequent treatment with PD1/PDL1 regimens that are not currently available in the UK. As these treatments are anticipated to lead to improvements in long-term response and survival in the TPC arm, results in the base case analysis are conservative (patients in the TPC arm of the model receive the full benefits associated with subsequent PD1/PDL1 regimens, without applying any costs as these are not reimbursed in the UK setting). This simplifying assumption was applied in the base case analysis as a conservative, pragmatic approach to account for differences between within-trial and real-world treatment patterns without sacrificing power in the analysis.

For the mixed chemotherapy scenario analysis, subsequent treatment proportions are assumed equal to TPC as a simplifying assumption. Clinical experts confirmed that these values are appropriate for use in the base case analysis, noting that between 20% to 50% of patients could be offered subsequent therapy.[2]

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 114 of 137

Confidential

A scenario was included to test the impact of using an alternative distribution of each subsequent treatment as informed by the ECHO study (detailed in Section B.2.9.3), presented in Table 46.[22] In this scenario the proportion of patients that received subsequent treatment were consistent to those in KEYNOTE-775 as supported by clinical expert opinion during validation.[2] This scenario accounts for systemic therapy agents used in third line treatment, with the exclusion of investigational treatments that are not currently available in the UK. This has minimal impact on the results, demonstrating that the base case results are robust to alternative assumptions related to subsequent treatment costs and that these inputs are not a key driver of the results.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 115 of 137

Confidential

Table 46: Subsequent therapies received by patients in KN-775 in the base case analysis

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 116 of 137

Confidential

Table 47: Subsequent therapy – drug formulation, dose, administration, proportion of doses received and total drug acquisition cost per week, intervention and active comparators

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 117 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 118 of 137

Confidential

Table 48: Duration of subsequent therapies

B.3.5.4.2. Testing

In line with the final NICE scope and decision problem for this appraisal, the modelled population considers all patients with previously treated advanced, metastatic or recurrent EC. The treatment benefit of PEM+LEN compared with TPC was consistent across all the major subgroups tested in patients with advanced EC, including by histology (Section B.2.7).

Recent guidelines for Lynch syndrome (caused by a germline pathogenic variant in one of four DNA mismatch repair (MMR) genes: MLH1, MSH2, MSH6 and PMS2) have recommended that all patient’s tumours are tested for MSI/MMR status.[71] As PEM+LEN would be eligible for the indicated population without any additional testing requirements, there are no applicable testing costs to be included in the economic model, and, if included, the costs for testing would offset.

B.3.5.4.3. End-of-life costs

End-of-life care is applied as a one-off cost upon transition to death in the economic model. The cost per patient was sourced from Georghiou et al. (2014)[72] (a Nuffield Trust report that explored care costs towards the end of life) which reported a value of £6,015. This cost was inflated to the current cost year using PSSRU inflation indices to give a cost of £6,520.55.

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 119 of 137

Confidential

B.3.6. Summary of base case analysis inputs and assumptions

B.3.6.1. Summary of base-case analysis inputs

A tabular summary of the variables applied in the economic model, their base case values and the uncertainty distribution and magnitude assumed for each variable, is provided in Appendix P.

B.3.6.2. Assumptions

Key assumptions of the economic analysis are summarised in Table 49. The approach to modelling has been designed to make the best use of the available data to inform the decision problem. In the absence of data, assumptions are designed to minimise potential bias in the analysis. These two statements are illustrated by the likely direction of bias and justification for analysis assumptions, summarised in Table 49.

Table 49: Summary of assumptions of the economic analysis

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 120 of 137

Confidential

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 121 of 137

Confidential

B.3.7. Base-case results

Summary of key points:

• Based on list prices for all treatments, the estimated ICER for PEM+LEN versus paclitaxel and doxorubicin is £65,111 per QALY gained. Appendix Q presents the ICERs incorporating the patient access scheme [PAS] currently agreed for pembrolizumab, which show that PEM+LEN is highly likely be cost-effective when the confidential discounts are included

• These ICERs should be considered in the context of PEM+LEN being an innovative, end-of-life technology that presents a step-change improvement for patients with advanced or recurrent EC who have received prior platinumcontaining therapy in an area of high unmet need

• All relevant health outcomes and costs were included in line with the NICE reference case:

  • OS and PFS were the main efficacy inputs used in the economic model were directly based on KEYNOTE-775 data to model health outcomes over patients’ lifetime

  • Cost categories included treatment acquisition and administration costs, health state resource use costs (e.g. ongoing monitoring and follow-up), AE costs, subsequent treatment costs and costs associated with end-of-life care

B.3.7.1. Base-case incremental cost-effectiveness analysis results – LIST PRICE for P+L

All results presented in this section assume the list price for treatments. Please see Appendix Q for additional results that incorporate the patient access scheme [PAS] currently agreed for pembrolizumab and list price for lenvatinib.

The cost-effectiveness results for PEM+LEN versus TPC are presented in Table 50 and disaggregated results are available in Appendix J. The results show that PEM+LEN is estimated to offer a substantial incremental health benefit compared with TPC, offering an additional 3.53 LYs and 1.75 QALYs per patient lifetime (a total of ***** LYs and ***** QALYs for PEM+LEN compared with ***** LYs and ***** QALYs for TPC). This level of benefit supports the importance of PEM+LEN as a

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 122 of 137

Confidential

treatment for patients with advanced or recurrent EC who have disease progression on or following prior treatment with a platinum-containing therapy who would otherwise face a poor prognosis under highly limited treatment options. At list prices, treatment with PEM+LEN is associated with an additional cost of £*****, primarily driven by a longer duration of treatment for PEM+LEN coupled with the cost difference between treatments given that TPC is available in generic formulation. The resulting ICER in the base case analysis is £65,111 per QALY gained for PEM+LEN vs TPC.

Appendix Q presents the ICERs incorporating the PAS currently agreed for pembrolizumab, which show that the combination would be very likely be cost effective when these discounts are included. These ICERs should be considered in the context of PEM+LEN being an innovative, end-of-life technology that presents a step-wise improvement for patients with advanced or recurrent EC who have received prior platinum-containing therapy.

Table 50: Base-case results – List prices

B.3.8. Sensitivity analyses

B.3.8.1. Probabilistic sensitivity analysis

Probabilistic sensitivity analysis (PSA) was performed to account for joint uncertainties in the key model inputs, in which multiple input parameters were varied simultaneously over a number of iterations by sampling their values from uncertainty distributions. Whenever available, the standard error (SE) of the selected distribution was obtained directly from the same data source that informed the mean value. In the absence of data on the variability around health state cost values, variability was

Company evidence submission template for pembrolizumab with lenvatinib for previously treated advanced, metastatic or recurrent endometrial cancer [ID3811][©] MSD (UK) Ltd 2022. All rights reserved 123 of 137

Confidential

assumed as 10% of the mean value. The appropriate probability distributions were used, as described in Appendix P.

The results of the PSA based on 1,000 simulations are presented in Table 51, along with the results illustrated in a scatterplot in Figure 28. The mean outcomes from the PSA are highly consistent with the base case results presented in Section B.3.7 (£65,511 compared with £65,111 per QALY gained, respectively). Therefore, the outcomes from the cost-effectiveness model are considered robust to uncertainty from parameter distributions. Figure 29 shows the cost-effectiveness acceptability curve associated with PEM+LEN versus TPC.

Figure 28: PSA scatterplot, PEM+LEN versus TPC – List prices

*** Key:** ICER, incremental cost-effectiveness ratio; PEM+LEN, pembrolizumab + lenvatinib; PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year; TPC, treatment of physician’s choice.

Figure 29: Cost-effectiveness acceptability curve – List prices

*** Key:** ICER, incremental cost-effectiveness ratio; PEM+LEN, pembrolizumab + lenvatinib; PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year; TPC, treatment of physician’s choice.

Table 51: Mean probabilistic base case results, pairwise analysis – List prices