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

Single Technology Appraisal

Olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

Committee Papers

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL

Olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

Contents:

The following documents are made available to consultees and commentators:

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

1. Company submission from AstraZeneca

2. Clarification questions and company responses

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

• a. Prostate Cancer UK b. Tackle Prostate Cancer

4. Expert personal perspectives from:

• a. Peter Isard – patient expert, nominated by Prostate Cancer UK

• 5. Evidence Review Group report prepared by Warwick Evidence

6. Evidence Review Group – factual accuracy check

7. Technical Report

8. Technical engagement responses from experts:

• a. Professor Johann de Bono – clinical expert, nominated by AstraZeneca b. Dr Sree Rodda – clinical expert, nominated by the British Uro-Oncology Group

9. Technical engagement responses from consultees and commentators: a. Prostate Cancer UK

• b. Tackle Prostate Cancer

10. Technical engagement response from AstraZeneca 11. Clarification questions on the company technical engagement response and company responses

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

13. Factual accuracy check of ERG critique

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

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

Document B

Company evidence submission

June 2020

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Contents

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Tables

Table 1. Technology being appraised. ..................................................................... 11 Table 2. The decision problem. ................................................................................ 17 Table 3. Clinical effectiveness evidence overview of PROfound and CARD studies. ................................................................................................................................. 27 Table 4. Key inclusion and exclusion criteria for the PROfound study. .................... 31 Table 5. Patient characteristics for PROfound Cohort A+B, Cohort A, prior taxane subgroup .................................................................................................................. 38 Table 6. Definition of populations. ............................................................................ 44 Table 7. Description of outcomes and methodology for statistical analysis. ............. 46 Table 8. Overview of quality assessments for the PROfound study ......................... 49 Table 9. Key efficacy outcomes in Cohort A, Cohort A+B, Cohort A+B prior taxane subgroup (used in ITC analysis). .............................................................................. 53 Table 10. Median OS and HR for investigators’ choice of NHA arm, adjusted for treatment switching; overall HRRm (Cohort A+B) and Cohort A .............................. 63 Table 11. Time to first SSRE in patients in Cohort A+B. .......................................... 69 Table 12. OS data in the prior taxane subgroup of PROfound at DCO2 (Cohort A+B) ................................................................................................................................. 79 Table 13. Summary of the covariates available for testing for effect modification .... 85 Table 14. Assessment of effect modifiers for switching adjusted anchored analysis. ................................................................................................................................. 87 Table 15. Summary of results for rPFS and OS: Bucher anchored ITC olaparib versus cabazitaxel .................................................................................................... 91 Table 16. Summary of treatment exposure, dose interruptions, and dose modifications: Cohort A + B SAS and prior taxane subgroup, DCO1 (4[th] June 2019) ................................................................................................................................. 94 Table 17. Adverse events in any category, DCO1 (4[th] June 2019) in Cohort A+B SAS/prior taxane subgroup. ..................................................................................... 96 Table 18. End-of-life criteria ................................................................................... 105 Table 19. Description of modelled patient population ............................................. 107 Table 20. Features of the economic analysis ......................................................... 114 Table 21. Summary of main clinical efficacy data sources for each intervention in the economic analysis. ................................................................................................. 117 Table 22. Number of events in the HRRm (Cohort A+B) prior taxane subgroup of PROfound. ............................................................................................................. 119 Table 23. AIC and BIC values for parametric models for rPFS (HRRm [Cohort A+B] – prior taxane). .......................................................................................................... 121 Table 24. Number of events in the HRRm (Cohort A+B) prior taxane subgroup of PROfound. ............................................................................................................. 123 Table 25. AIC and BIC values for parametric models for OS (HRRm [Cohort A+B] – prior taxane, olaparib). ........................................................................................... 124 Table 26. OS estimates for olaparib (HRRm [Cohort A+B] - prior taxane, olaparib) ............................................................................................................................... 126 Table 27. OS for cabazitaxel (predicted via ITC) .................................................... 129 Table 28. Final rPFS and OS models selected for economic evaluation (HRRm [Cohort A+B] – prior taxane)................................................................................... 130 Table 29. Summary of TTD approach selected for base case and scenario analysis ............................................................................................................................... 133

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 30. AIC and BIC values for parametric models for TTD (HRRm – prior taxane, olaparib) ................................................................................................................. 136 Table 31. Grade 3 and above AEs affecting at least 5% of patients included in base case analysis. ......................................................................................................... 138 Table 32. Distribution of SREs identified from mCRPC clinical trials...................... 140 Table 33. Probability of SREs occurring with each intervention in the base case analysis. ................................................................................................................. 140 Table 34. Mean (LSM) HSUV estimates by progression status (HRRm population) ............................................................................................................................... 143 Table 35. Utility values based on targeted literature search in mCRPC setting (not specific to post-NHA population). ........................................................................... 145 Table 36. AE utility decrements and duration of AEs applied in the economic model ............................................................................................................................... 149 Table 37. SRE utility decrements and duration of SREs applied in the economic model ..................................................................................................................... 150 Table 38. Mean health state utility values used in the economic model (base case).[a] ............................................................................................................................... 153 Table 39. Summary of intervention and comparator costs included in the economic analysis (base case). .............................................................................................. 155 Table 40. Mean relative dose intensity ................................................................... 158 Table 41. Unit cost per pack/vial for olaparib and comparators in the economic model ............................................................................................................................... 162 Table 42. Unit cost per pack/vial for premedication and concomitant medications in the economic model ............................................................................................... 163 Table 43. Proportion of patients in the cabazitaxel arm receiving each concomitant medication in the base case analysis ..................................................................... 164 Table 44. Administration costs applied for olaparib and cabazitaxel in the economic model. .................................................................................................................... 165 Table 45. Data informing subsequent treatment costs applied in the economic analysis. ................................................................................................................. 170 Table 46. Data informing Unit costs associated with subsequent treatment applied in the economic analysis ............................................................................................ 171 Table 47. Unit costs associated with subsequent treatment applied in the economic analysis .................................................................................................................. 171 Table 48. Proportion of patients receiving best supportive care after progressing on olaparib or cabazitaxel in the economic model (based on TA391 and clinical expert opinion). ................................................................................................................. 173 Table 49. Summary of health-state unit costs for resource use ............................. 174 Table 50. Summary of health-state resource use frequency per month ................. 174 Table 51. Summary of AE and SRE unit costs ....................................................... 177 Table 52. Summary of variables applied in the economic model (base case analysis) ............................................................................................................................... 179 Table 53. Assumptions applied in the economic analysis. ..................................... 183 Table 54. Base-case results (costs and health outcomes discounted at 3.5%). ..... 187 Table 55. PSA results. ............................................................................................ 188 Table 56. Scenario analyses .................................................................................. 190

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figures

Figure 1. PROfound study design. ........................................................................... 31 Figure 2. Summary of the hierarchical testing structure for PROfound. ................... 45 Figure 3. An overview of the PROfound trial and its placement in the current submission. .............................................................................................................. 52 Figure 4. Primary outcome: Kaplan–Meier plot of BICR-assessed rPFS in Cohort A. ................................................................................................................................. 54 Figure 5. Key secondary outcome: Kaplan–Meier plot of BICR-assessed rPFS in patients Cohort A+B ................................................................................................. 56 Figure 6. Key secondary outcome: Kaplan–Meier plot of interim OS in patients in Cohort A. .................................................................................................................. 58 Figure 7. Secondary outcome: Kaplan–Meier plot of interim OS in patients in Cohort A+B .......................................................................................................................... 59 Figure 8. Kaplan–Meier plot of counterfactual overall survival in Cohort A+B (RPSFTM Weibull method, no re-censoring). ........................................................... 62 Figure 9. Secondary outcome: Kaplan–Meier plot of interim TTPP by investigator assessment in patients in overall HRRm population (Cohort A+B). ......................... 68 Figure 10. Secondary outcome: confirmed BICR-assessed radiological ORR in Cohort A+B. .............................................................................................................. 70 Figure 11. Mean change from baseline EQ-5D-5L scores up to week 64 (Cohort A+B). ........................................................................................................................ 72 Figure 12. Kaplan–Meier plot of BICR-assessed rPFS in Cohort A+B patients whom had prior taxane treatment ....................................................................................... 75 Figure 13. Kaplan–Meier plot of interim OS in patients whom had prior taxane treatment in Cohort A+B ........................................................................................... 76 Figure 14. Kaplan–Meier plot of counterfactual overall survival in patients who had prior taxane treatment in Cohort A+B (RPSFTM Weibull method, no re-censoring). 78 Figure 15. Evidence network for the ITC (OS and rPFS) ......................................... 83 Figure 16: PROfound rPFS: Schoenfeld (left) and log-cumulative hazards (right) plots .......................................................................................................................... 88 Figure 17: CARD rPFS: Schoenfeld (left) and log-cumulative hazards (right) plots . 89 Figure 18. PROfound OS: Schoenfeld (left) and log-cumulative hazards (right) plots ................................................................................................................................. 90 Figure 19. CARD OS: Schoenfeld (left) and log-cumulative hazards (right) plots .... 90 Figure 20. Structure of the cost-effectiveness model ............................................. 110 Figure 21. Olaparib BICR rPFS, Kaplan–Meier plot (HRRm [Cohort A+B] – Prior taxane). .................................................................................................................. 120 Figure 22. Modelled rPFS for olaparib based on PROfound (HRRm [Cohort A+B] – prior taxane). .......................................................................................................... 120 Figure 23. Modelled rPFS for cabazitaxel based on ITC HR vs olaparib as reference curve. ..................................................................................................................... 122 Figure 24. OS Kaplan–Meier plot (HRRm [Cohort A+B] – prior taxane subgroup, olaparib) ................................................................................................................. 123 Figure 25. Modelled OS based on PROfound (HRRm [Cohort A+B] – prior taxane subgroup, olaparib). ............................................................................................... 124 Figure 26. Modelled OS for cabazitaxel based on ITC HR vs olaparib as reference curve. ..................................................................................................................... 127

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 27. Selected distribution for extrapolating OS for olaparib (PROfound, Kaplan–Meier estimate and loglogistic model; prior taxane subgroup) and cabazitaxel (via ITC).[a, b] .......................................................................................... 131 Figure 28. Selected distribution for extrapolating rPFS for olaparib (PROfound, Kaplan–Meier estimate and Weibull model; prior taxane subgroup) and cabazitaxel (via ITC).[a,b] ............................................................................................................. 132 Figure 29. TTD, Kaplan–Meier plot (HRRm – prior taxane) ................................... 135 Figure 30. Modelled treatment duration based on PROfound (HRRm – prior taxane, olaparib). ................................................................................................................ 136 Figure 31. PSA results scatterplot, costs and QALYs for olaparib and cabazitaxel (1,000 simulations). ................................................................................................ 189 Figure 32. Cost-effectiveness acceptability curve (olaparib versus cabazitaxel). ... 189 Figure 33. Tornado plot: 10 most influential inputs (DSA results). ......................... 190

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

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

The regulatory submission for the olaparib indication relevant to this appraisal was provided to the European Medicines Agency (EMA) on XXXXXXXXXXXXX. The anticipated marketing authorisation is aligned to the overall population of the pivotal Phase III PROfound study:

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. XXXXXXXXXX.

This section provides an overview of the health condition, the current clinical pathway of care, a description of olaparib (the technology being appraised), and the decision problem addressed in the company submission.

B.1.1 Overview of prostate cancer

Prostate cancer is the most common cancer in men in the UK[1] with a total of 42,668 people in England and Wales were diagnosed with prostate cancer between 2017 and 2018.[2] Although the majority of patients (83%) present with early stage disease at the time of diagnosis,[2] a substantial proportion of patients eventually develop resistance to therapy and progress to castration-resistant prostate cancer, or metastatic prostate cancer,[3] or (for ~40% of patients) metastatic castration-resistant prostate cancer (mCRPC).[4,5]

mCRPC - the focus of this appraisal - is associated with substantially increased symptom burden, deterioration in health-related quality of life (HRQoL), and increased mortality (with >3 higher risk of death) versus non-metastatic disease.[6-9] Almost all patients dying from prostate cancer have mCRPC,[10] and fewer than half of patients with mCRPC in the UK survive for 5-years.[11] Therefore, there is significant and urgent unmet clinical need for life-extending therapies for the treatment of mCRPC.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Approximately 20%−30% of patients with mCRPC have mutations in genes involved in the homologous recombination repair (HRR) pathway, such as BRCA1, BRCA2, ATM .[12,13] While not all HRR mutations (HRRm) are fully characterised, those patients with BRCA2 and CDK12 -mutations (two of the most well-characterised of the HRR mutations) have more aggressive disease, and worse prognosis, than those without.[14-26] Additionally, mutations in BRCA-2 have been reported to lead a reduced response to taxane chemotherapy, highlighting the urgent need for new types of treatment in these patients.[27]

The presence HRRm render tumours sensitive to targeted therapy with poly(adenosine diphosphate)-ribose polymerase (PARP) inhibitors, such as olaparib, which specifically target and kill HRR-deficient tumour cells via a mechanism involving synthetic lethality (described in Section B.1.3).[28,29]

Olaparib is the only PARP inhibitor supported by Phase III trial data, demonstrating a statistically-significant extension in radiographic

progression-free survival (rPFS), as well as a significant benefit overall

survival (OS)[1] , versus investigators’ choice of NHA in patients with deleterious or suspected deleterious HRR mutations who have progressed after first-line treatment for mCRPC with an NHA,[30,31] and provides a much-needed new therapeutic option to improve patient outcomes in this setting .

B.1.2 Clinical pathway of care

The growth and survival of prostate tumours is dependent on androgens, particularly testosterone and dihydrotestosterone.[32] Due to the dependence of prostate tumours on androgen receptor signalling, all prostate cancers are treated with androgen deprivation therapy (ADT). However, eventually these tumours may develop resistance to ADT, characterised by increased levels of testosterone that drive cancer progression. Overall, around 40% of all patients with prostate cancer go on to develop metastatic castration-resistant disease (i.e. mCRPC),[4,5,33] which is an incurable form of cancer.[34]

1 In patients with BRCA1/2 and ATM mutations. Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

The use of docetaxel in the pre-mCRPC setting is increasing. In the 2019 National Prostrate Cancer (NPCA) Audit,[2] which used data from 2017−2018, docetaxel was used in the treatment of approximately 1 in 4 patients with metastatic hormonesensitive prostate cancer (mHSPC). Since these data were collected, NICE Guideline 131 (NG131, May 2019) for “Prostate Cancer: Diagnosis and

Management” recommended the addition of docetaxel to standard-of-care ADT for the treatment of patients with high-risk HSPC (used synonymously with locally-advanced disease in NG131) or mHSPC , leading to a change in treatment paradigm and more patients receiving docetaxel in this setting. Six UK clinical experts were consulted by AstraZeneca to inform the company submission and highlighted that ~75% of patients currently receive docetaxel in the pre-mCRPC setting (AstraZeneca data on file). Data from the GETUG-AFU15 trial showed that, although the addition of docetaxel to ADT led to improved patient outcomes in this setting, prolonging median time to biochemical disease progression by ~11 months (versus ADT alone), the majority (50%) of patients still progressed within two years of starting therapy.[35]

Treatment with new hormonal agents (NHAs), specifically enzalutamide or abiraterone acetate, are the standard-of-care and recommended by NICE for patients with mCRPC .[36-38] With docetaxel moving earlier in the treatment pathway from May 2019 (per NG131),[36] most patients in England currently receive NHA as initial treatment for mCRPC after treatment with docetaxel and ADT for advanced HSPC. Although NHAs have substantially improved outcomes in patients with mCRPC, many (~60%) do not respond to these therapies and ~50% experience disease progression within 6–12 months of initiating therapy.[39,40]

Treatment options for mCRPC after NHA progression are limited. The six UK clinical experts consulted by AstraZeneca confirmed that the standard-of-care for patients after disease progression on docetaxel and an NHA is cabazitaxel (AstraZeneca data on file), which is recommended by NICE as a treatment option “in combination ‐ with prednisone or prednisolone for patients with metastatic hormone relapsed prostate cancer whose disease has progressed during or after docetaxel chemotherapy”.[41] Although NICE guidelines still recommend docetaxel post-NHA, in practice docetaxel is now predominantly used earlier in the treatment pathway (as

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 9 of 208

explained above); the 6 UK clinical experts consulted highlighted that re-treatment with docetaxel is not preferred in patients where cabazitaxel is a treatment option (AstraZeneca data on file).

Prognosis for patients with mCRPC who have progressed following treatment with NHA is extremely poor, with a median OS of just 13.6 months reported in the CARD study, the pivotal clinical trial for cabazitaxel in the post-NHA setting.[42]

Treatment options are further limited once patients have exhausted cabazitaxel as a treatment option (i.e. patients have progressed on cabazitaxel or cannot receive it [e.g. due to frailty or contraindications]). Prostate cancer has a high propensity to metastasise to bone tissue. NICE guidelines recommend radium ‐ 223 dichloride (referred to radium-223 henceforth) as an option for treating patients with “hormonerelapsed prostate cancer, with symptomatic bone metastases and no known visceral metastases, only if they have already received docetaxel or docetaxel is contraindicated or is not suitable for them”.[43] Although it is possible to use radium223 dichloride instead of cabazitaxel, in those patients who have symptomatic bone metastases (and no known visceral metastases) and have received prior docetaxel for hormone-sensitive disease, clinical expert opinion from 6 UK-based clinical experts indicates that in practice it is often reserved for later-lines of treatment, unless treatment with a taxane is not suitable (AstraZeneca data on file).

Olaparib, the intervention relevant to this appraisal, offers a new targeted treatment option for those patients who have deleterious or suspected deleterious HRRm and whose disease has progressed after treatment with an NHA. Based on the current clinical practice in England, where ~75% of patients receive docetaxel with ADT for metastatic HSPC (and radium-223 dichloride is reserved for later lines of treatment in eligible patients) (AstraZeneca data on file),[44] it is anticipated that olaparib will primarily displace treatment with cabazitaxel following disease progression after NHA.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

B.1.3 Description of the technology being appraised

The summary of product characteristics (SmPC) and European public assessment report (EPAR) can be found in Appendix C.

Table 1. Technology being appraised.

==> picture [459 x 561] intentionally omitted <==

----- Start of picture text -----
UK approved Olaparib (LYNPARZA [®] )
name and brand
name
Mechanism of Olaparib inhibits poly(adenosine diphosphate)-ribose polymerase
action (PARP) proteins. [45 ] PARP enzymes are essential for repairing
commonly-occurring DNA single-strand breaks (SSBs) in human cells.
Olaparib works by trapping PARP enzymes at the site of SSBs, thereby
preventing their repair. Persistent SSBs in the DNA are eventually
converted into more harmful double-strand breaks (DSBs) during the
process of DNA replication. Normal cells can repair DNA DSBs through
the homologous recombination repair (HRR) pathway. However, cells
with HRR defects/deficiencies are unable to accurately repair these
breaks, leading to the accumulation of DNA damage and eventually cell
death (or apoptosis). [46-48]
Mode of action of PARP inhibitors, including olaparib
HR [+] , homologous repair positive; HR [–] , homologous repair negative; HRRm, mutation in
the homologous recombination repair pathway; PARP, poly(adenosine diphosphate)-
ribose polymerase.
Source: Guha et al , 2011 [46]
----- End of picture text -----

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

ADP, adenosine diphosphate ribose; ATM , ataxia-telangiectasia mutated gene; BRCA , BReast CAncer gene; CHMP, Committee for Medicinal Products for Human Use; DSB, double stranded breaks; FIGO, Fédération Internationale de Gynécologie et d’Obstétrique [Federation of Gynaecology and Obstetrics]; HER2, human epidermal growth factor receptor 2; HRR, homologous recombination repair; PARP, poly (ADP-ribose) polymerase; SSB, single stranded breaks.

B.1.4 Decision problem

As highlighted above, the regulatory submission for the olaparib indication relevant

to this appraisal was provided to the European Medicines Agency (EMA) on X.. XXXXXXXXXX. The anticipated marketing authorisation is aligned to the overall population of the pivotal Phase III PROfound study:

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

The overall population of patients included in the PROfound study had qualifying mutations in one or more of 15 HRR genes ( BRCA1, BRCA2, ATM, BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, PPP2R2A, RAD51B, RAD51C, RAD51D, and RAD54L ).[30] Enrolment into the study was not restricted by prior taxane use (given the unmet clinical need for patients who had progressed after treatment with a NHA, regardless of whether they had received prior treatment with a taxane), or baseline metastases (bone, visceral, or other).[30] However, the clinical- and costeffectiveness evidence presented in this submission focuses on the predefined subgroup of patients who had received prior treatment with a taxane and NHA, to align with the anticipated positioning of olaparib in the current clinical pathway of care in England (where the majority of patients receive a taxane [docetaxel] for nonmetastatic or metastatic HSPC, before receiving NHA for mCRPC, as described in Section B.1.2).

The PROfound study assessed the efficacy and tolerability of olaparib versus investigators’ choice of NHA, since re-treatment with NHA (i.e. enzalutamide after progression of abiraterone, and vice versa) are approved treatment options in this setting (by both the EMA[52,53] and the US FDA[54,55] ) and is a standard-of-care in many countries where the PROfound study was conducted.[56] This strategy also ensured that patients for whom treatment with chemotherapy was unsuitable were not excluded from the study. Since re-treatment with NHA is not reimbursed in

England,[57] an anchored Bucher indirect treatment comparison (ITC) was conducted versus cabazitaxel, the most-commonly used treatment option and current standardof-care for patients whose disease has progressed after treatment with a taxane (i.e. docetaxel) and an NHA – the focus of the company submission.

The decision-problem specified by NICE for this appraisal also includes docetaxel and radium-223 dichloride as comparators of interest. These are discussed below.

• Docetaxel: As explained in Section B.1.2, since the publication of NG131 in May 2019,[36] the vast majority (~75%) of patients now receive docetaxel earlier in the treatment pathway (added to ADT, for HSPC; AstraZeneca data on file). Based on the opinion of 6 UK clinical experts, re-treatment with docetaxel for mCRPC is not routinely used in clinical practice, with patients receiving cabazitaxel following

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

disease progression after NHA (AstraZeneca data on file). Therefore, treatment with docetaxel in the post-NHA setting, which is relevant to this appraisal, is only considered for the minority (~25%) of patients who have not previously received docetaxel for prostate cancer, and who are willing and able/fit enough to endure this treatment . To fulfil the NICE scope, we investigated the feasibility of comparing olaparib to docetaxel in the subset of patients with previous NHA use via an indirect comparison; however, no suitable published randomised controlled trials (RCTs) for docetaxel in mCRPC after diseaseprogression on NHA were identified in the systematic literature review (SLR) conducted to inform this submission, thus limiting our ability to conduct any such analyses. Whilst we are unable to provide comparative evidence, it is worth highlighting that treatment with olaparib resulted in clinically-meaningful extensions to both rPFS and OS in the overall study population of PROfound, which included patients regardless of whether they had received prior treatment with a taxane (rPFS, HR, 0.39 [95% CI, 0.29–0.53] and 0.77 [95% CI, 0.50–1.22] in the “prior taxane” and “no prior taxane” subgroups , respectively; olaparib versus investigators’ choice of NHA). These data are further described in Section B.2.6 and Appendix E.

 Radium-223: As described in SectionB.1.2, radium-223 is recommended as a treatment option for patients with symptomatic bone metastases and no known visceral metastases, who have already received docetaxel treatment., However opinion from 6 UK-based clinical experts indicates that, in practice, radium-223 is reserved for later-lines of treatment (i.e. after NHA and cabazitaxel), unless treatment with a taxane is not suitable (AstraZeneca data on file).[58] Radium223 is thus only an appropriate comparator for olaparib in the latter circumstance. This positioning is also supported by data from a recent UK national radium-223 audit, which also reported on its use in later lines of treatment.[59] Comparisons between olaparib and radium-223, in patients for whom treatment with docetaxel is unsuitable, is limited by two factors:

• 1. There are no published RCTs for radium-223 dichloride for the treatment of patients whose disease has progressed after an NHA (section B.2.1.3).
• 2. We would have to make the assumption that patients in PROfound who did not receive a taxane prior to NHA were “unsuitable” for treatment with

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

docetaxel, which may be inappropriate (since 19.5% of patients in the “no prior taxane” subgroup of PROfound did receive a taxane as subsequent treatment following disease progression on study treatment), and bias any analyses conducted.

Although a comparative analysis of olaparib versus radium-223 dichloride was not possible for this small group of patients due to these limitations, it is worth highlighting that data from PROfound show efficacy for olaparib in patients with baseline bone metastasis, with a HR of 0.57 (95% CI: 0.35–0.94) for rPFS versus investigators’ choice of NHA.

Given the limitations described above in addressing the relative clinical- and costeffectiveness of olaparib versus docetaxel (in patients with no docetaxel use prior to NHA) and radium-223 dichloride (in patients who have bone metastases, no known visceral metastases, and in whom docetaxel is contraindicated), this submission focuses on the use of olaparib versus cabazitaxel in patients who have received a prior taxane and a NHA. The final scope issued by NICE and the decision-problem addressed in the company submission are detailed in Table 2.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 2. The decision problem.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca 2020 All rights reserved Page 17 of 208

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 18 of 208

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca 2020 All rights reserved Page 19 of 208

group of patients who have not received a taxane prior to NHA under equality provisions

ATM, ataxia-telangiectasia mutated; BRCA, BReast CAncer gene; CSR, clinical study report; EMA, European Medicines Agency; HRR, homologous recombination repair; NHA, new hormonal agent; NICE, National Institute of Health and Care Excellence; NHS, National Health Service; N/A, not applicable OS, overall survival; (r)PFS, radiographic progression-free survival; PFS2, second progression-free survival; RCT, randomised controlled trial.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca 2020 All rights reserved Page 20 of 208

B.1.5 Equality considerations

Although a formal comparative clinical and cost-effectiveness analysis versus docetaxel (for patients with no docetaxel use prior to NHA) and radium-223 (for patients with bone metastases and no known visceral metastases, in whom docetaxel was contraindicated) was not feasible due to the limited published RCT evidence identified (Section B.2.1), it is worth highlighting that olaparib has shown efficacy in the overall HRRm population of the PROfound study, which included patients regardless of prior taxane use and baseline metastases (bone, visceral, or other; Section B.2.6.2.3 and Section B.2.7).[30]

In addition to meaningfully extending survival and providing the hope of long-term response (for at least a small group of patients), olaparib also represents a more patient-centric treatment option than docetaxel and radium-223 dichloride, on account of its oral administration (negating the need for patients to travel to the hospital to receive treatment) and its tolerability profile , with most adverse events (AEs) being non-serious and manageable, without requiring discontinuation of treatment.[61]

Data from PROfound also showed no meaningful deterioration in patients’ HRQoL over the course of olaparib treatment compared with investigator’s choice of NHA; instead, patients benefitted from a reduced burden of pain as well as reduced incidence of symptomatic skeletal-related events, which are significant causes of morbidity in patients with mCRPC.[62,63] On account of these reasons, and given that many patients may not be able to access or wish to undergo treatment with docetaxel and/or radium-223 due to the factors described below,[37] we request that NICE consider granting access to olaparib for those patients who have not received a taxane prior to NHA or for whom treatment with a taxane is

unsuitable under equality considerations.

• ~20% of men are considered clinically unsuitable for chemotherapy at diagnosis,

• Many others are simply unable to receive it for reasons beyond clinical factors, such as:

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• The presence of a carer or loved one for support (both for attending hospital appointments for infusions and managing potential side effects),

• Proximity to treatment centres/access to public transport, regardless of whether they live alone or have a carer,

• The willingness and emotional endurance to tolerate the toxicity of chemotherapy, which is often underestimated, and

• Religious beliefs, for instance, due to the alcohol content present in docetaxel.

Collectively, these factors could prevent mCRPC patients from seeking treatment with docetaxel or radium-223 and thus compromise their prognosis in the absence of any alternative life-extending therapy.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

B.2 Clinical effectiveness

B.2.1 Identification and selection of relevant studies

A systematic literature review (SLR) was conducted in January 2020 in order to identify published clinical evidence on the use of health technologies in patients with mCRPC whose disease had progressed following treatment with an NHA, irrespective of HRR mutation status. The scope of the SLR was broader than that of the decision-problem, and did not restrict inclusion by:

• Patients with mCRPC who had HRR gene mutations , to capture all studies in a post-NHA setting that could inform indirect comparisons with existing drugs that are not targeted to HRR mutations.

• The treatments and comparators specified in the final NICE scope, to ensure that no relevant studies were accidentally excluded.

Full details of the SLR methodology and results are provided in Appendix D.

The SLR identified 157 studies. Of these, a total of 14 studies, reported across 23 publications reported outcomes with olaparib,[64,65] cabazitaxel,[42,66-73] docetaxel[74-81] and radium-223[82-84] (i.e. the intervention and comparators specified in the NICE scope) in the population of interest.

Of the studies that reported on cabazitaxel, docetaxel, and radium-223 dichloride, the comparators of interest for this appraisal, just one (the CARD study of cabazitaxel versus NHA; described below), was appropriate for inclusion in the evidence base for this appraisal. Reasons for excluding the remaining studies are briefly described below and in further detail in Section B.2.9 (further details on all studies are provided in Appendix D).

==> picture [38 x 10] intentionally omitted <==

Olaparib

The SLR identified three studies that assessed the efficacy and safety of olaparib in the population of interest for this appraisal (PROfound and TOPARP-A/B; three abstracts, two full text publications).[28,29,64,65,85] The TOPARP studies (comprising TOPARP-A and TOPARP-B) were Phase II single arm trials that evaluated the

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 23 of 208

300 mg and 400 mg dose of olaparib in patients who had received a prior taxane; enrolment was not restricted by whether patients had received a prior NHA. Given the availability of data from the much larger Phase III , PROfound randomisedcontrolled trial in exact population and treatment setting relevant to this appraisal , data from TOPARP were not included in the evidence synthesis.

==> picture [39 x 11] intentionally omitted <==

Cabazitaxel

Eight publications that reported outcomes on cabazitaxel were identified by the SLR. Of these, only one study - CARD (NCT02485691) – was deemed relevant for an indirect treatment comparison with the PROfound trial (as described in Section B.2.9).[2][42,67] CARD is an ongoing Phase IV RCT that assessed the efficacy and safety of cabazitaxel compared with an NHA (enzalutamide or abiraterone plus prednisolone) in patients with mCRPC, who had received previous treatment with docetaxel and an NHA.[66] As all patients enrolled in the CARD trial were required to have received previous docetaxel, the patient population is closely aligned with the prior-taxane subpopulation of the PROfound study, although not restricted to those patients who have mutations in HRR genes.[30] Detailed baseline characteristics in the PROfound and CARD studies is provided in Section B.2.3.7 and Appendix D (Section D.1.4), respectively.

The remaining six publications that reported outcomes in patients who received cabazitaxel were either small early phase or single-arm studies (often conducted in a single country or centre)[68-72] or cabazitaxel combination studies (with budesinone, prednisone, prednisolone, or abiraterone), or did not report on the outcomes of interest (split by those who had received a prior NHA, in case of a mixed population) and were therefore deemed unsuitable for inclusion in the evidence base for this appraisal, as described in Section B.2.9.[73]

2 It is worth noting that the TROPIC study, the registrational clinical trial for cabazitaxel in mCRPC (which pre-dates CARD by nine years), did not evaluate its efficacy and safety in the post-NHA setting relevant to this appraisal. This study was thus not identified in the SLR or included in the evidence synthesis. As such, the CARD study provides a more relevant evidence base for comparative analysis versus olaparib due to 1) being more-recent, 2) being conducted in the post-NHA setting relevant to this appraisal, and 3) having an NHA comparator arm (like PROfound), thus making an anchored indirect treatment comparison versus olaparib possible. Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 24 of 208

==> picture [39 x 10] intentionally omitted <==

Docetaxel

Although the SLR identified eight publications that included docetaxel, none of these studies were relevant to the decision-problem, since they: 1) they either did not include docetaxel as a monotherapy arm (four publications)[76-78] , or 2) did not include patients based on progression on prior NHA therapy (a pre-requisite for randomisation in the PROfound study; three publications)[74,79] or 3) did not report appropriate data on the key survival outcomes of OS and rPFS results (three publications).[75,80,81] An overview of studies is presented in Appendix D. Radium-223

The SLR did not identify any studies that reported on outcomes after treatment with radium-223 dichloride monotherapy in mCRPC patients whose disease had progressed after treatment with an NHA. Of note, the SLR did identify the radium223 international Early Access Program (iEAP; NCT01618370); however, this included patients regardless of whether they had received a prior NHA. Furthermore, only an abstract was available from the study at the time the SLR was conducted, providing insufficient evidence for further analysis and inclusion in the evidence base (as described in section B.2.9). An overview of the study is presented in Appendix D.

==> picture [39 x 10] intentionally omitted <==

Summary

In line with the final NICE scope for this appraisal, evidence from studies of olaparib, cabazitaxel, docetaxel, and radium-223 were considered. The SLR identified two studies of interest: the PROfound trial, which compared olaparib against NHA rechallenge in a post-NHA setting in patients with one or more HRR mutations, and the CARD study, which compared cabazitaxel against NHA re-challenge in a post-NHA setting irrespective of HRR mutation status.[64,65,67] No suitable studies were identified which would facilitate robust indirect comparisons against either radium-223 or docetaxel. This is further discussed in Section B.2.9.

B.2.2 List of relevant clinical effectiveness evidence

Brief details of PROfound and CARD studies, the two clinical trials included in the evidence synthesis described in Section B.2.9 are presented in Table 3. Further details are on the PROfound study, the pivotal RCT for olaparib and the main source of data used in the economic analysis, are provided in Sections B.2.3 to B.2.7.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Details of the CARD study are provided in Appendix D (section D.1.4); comparative evidence of cabazitaxel versus olaparib is provided in Section B.2.9.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 3. Clinical effectiveness evidence overview of PROfound and CARD studies.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 27 of 208

AE, adverse event; CTC, circulating tumour cell; DoR, duration of response; G-CSF, granulocyte-colony stimulating factor; HRQoL, health-related quality of life; HRR, homologous recombination repair; mCRPC, metastatic castration-resistant prostate cancer; NHA, new hormonal agent; ORR, objective response rate; OS, overall survival; PFS2, second progression-free survival; PSA, prostate-specific antigen; q3w, every 3 weeks; rPFS, radiographic progression-free survival; SRE, skeletal-related events; TTPP, time to pain progression.

Source: PROfound Clinical Study Report, version 1, 23 October 2020,[61] de Bono et al , 2020[30] and de Wit et al , 2019.[67]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca 2020 All rights reserved Page 28 of 208

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

An overview of the PROfound study (the pivotal Phase III study for olaparib in the population of interest), including trial design, eligibility criteria, trial drugs and concomitant medications, primary and key secondary outcomes, and baseline characteristics is provided in the sections below (B.2.3.1−B.2.3.7). Further details are available in the Clinical Study Protocol and the Clinical Study Report (Sections 8−10).

Details of the CARD study (from which the efficacy data for cabazitaxel were derived and were used in the anchored ITC versus olaparib as described in Section B.2.9) are provided in Appendix D, section D.1.4.

==> picture [38 x 10] intentionally omitted <==

Trial design

PROfound (NCT02987543) is a Phase III, randomised, open-label, multicentre study designed to assess the efficacy of olaparib compared with investigators’ choice of enzalutamide or abiraterone (investigators’ choice of NHA) in patients with mCRPC and had a confirmed HRR mutation, whose disease had progressed following an NHA (abiraterone or enzalutamide).[30,87] PROfound is the pivotal study investigating the efficacy and tolerability of olaparib in the population of patients relevant to this appraisal, and formed the basis of the EMA regulatory submission for olaparib in the mCRPC indication.

The PROfound study included multiple oral agents in the comparator arms. At the time of study design, it was believed that the differences in administration and safety profiles would enable investigators to differentiate between the different study treatments. Thus, PROfound was designed as an open-label study. The primary endpoint of rPFS was evaluated by BICR assessment, thus removing the risk of any investigator bias from the results.

Patients were randomised using an interactive voice response system/interactive web response system in a 2:1 ratio to receive either olaparib tablets, or investigators’ choice of NHA. As highlighted in Section B.1.2, investigator’s choice of NHA was

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

chosen as the comparator since treatment with NHA (i.e. enzalutamide after progression of abiraterone, or vice versa) is an approved treatment option (by both the EMA and the US FDA) and is a standard-of-care in many countries where the PROfound study was conducted.[88] This strategy also ensured that patients for whom treatment with chemotherapy was unsuitable were not excluded from the study.

The PROfound study did not restrict enrolment by prior taxane use (given the unmet clinical need for patients who had progressed after treatment with an NHA, regardless of whether they had received prior treatment with a taxane), or baseline metastases (bone, visceral, or other). Randomisation was stratified by previous taxane use (yes, no) and measurable disease at baseline (yes, no), to ensure that patients were well-balanced across treatment arms.

Based on the results of the Phase II TOPARP-A trial (which showed that patients with HRR gene mutations demonstrated notably higher response rates and marked improvements in rPFS and OS following olaparib treatment, when compared with patients without HRR gene mutations),[28] a decision was made to restrict enrolment to those mCRPC patients who had a qualifying tissue HRR gene mutations. The overall HRR-mutated (HRRm) population of PROfound included two Cohorts:

• Cohort A: patients with at least one mutation in either BRCA1, BRCA2 or ATM genes.

• Cohort B: patients with mutations in BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, PPP2R2A, RAD51B, RAD51C, RAD51D or RAD54L genes.[87]

The primary endpoint of PROfound was rPFS (by BICR) for olaparib versus investigators’ choice of NHA in Cohort A. The rationale for using only BRCA1 , BRCA2 and ATM tissue mutations (i.e. Cohort A) for the primary endpoint was based upon the prevalence of these mutations and/or how well mutations in these genes have been characterised to date.[61] It was expected that qualifying mutations would be detected in the tumour tissue of approximately 1 in 10 patients with mCRPC.

Treatment switching from investigators’ choice of NHA to olaparib was permitted in the study following BICR-assessed radiographic disease progression. A summary of

the study design is provided in Figure 1. Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 30 of 208

Figure 1. PROfound study design.

==> picture [471 x 215] intentionally omitted <==

a Cohort B HRR pathway genes include BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, PPP2R2A, RAD51B, RAD51C, RAD51D and RAD54L .

b Treatment switching was permitted post BICR-assessed radiographic disease progression and adjusted for using naïve and sophisticated adjustment methods (see section B.2.6.3.1).

ATM , gene for ataxia–telangiesctasia mutated; BARD1 , gene for BRCA1 associated RING domain 1; BICR, blinded independent central review; bid, twice daily; BRCA1 , gene for breast cancer type 1 susceptibility protein; BRCA2 , gene for breast cancer type 2 susceptibility protein; BRIP1 , gene for BRCA1 interacting protein C-terminal helicase 1; CDK12 , gene for cyclin-dependent kinase 12; CHEK1 , gene for checkpoint kinase 1; CHEK2 , gene for checkpoint kinase 2; FANCL , gene for Fanconi anaemia, complementation group L; FMI, Foundation Medicine, Inc.; HRR, homologous recombination repair; mCRPC, metastatic hormone-resistant prostate cancer; NHA, new hormonal agent; PALB2 , gene for partner and localiser of BRCA2; PPP2R2A , gene for protein phosphatase 2 regulatory subunit B alpha; RAD51B , RAD51C , RAD51D , genes for RAD51 paralogues B, C and D; RAD54L , gene for RAD54-like protein.

Source: PROfound CSP version 4, 7 March 2019.[87]

==> picture [39 x 10] intentionally omitted <==

Eligibility criteria

Key eligibility criteria for the PROfound study are summarised in Table 4 with further details available in the PROfound Clinical Study Protocol, version 1, 23 October 2019 (section 3.1 and section 3.2). Following enrolment, all patients underwent central assessment to determine HRR pathway gene mutation status using a genetic assay, as described in Section B.2.3.6.[87]

Table 4. Key inclusion and exclusion criteria for the PROfound study.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

ADT, androgen deprivation therapy; AML, acute myeloid leukaemia; CT, computed tomography; ECG, electrocardiogram; ECOG PS, European Cooperative Oncology Group performance status; HRRm, homologous recombination repair pathway mutation; LHRH, leuteinising hormone-releasing hormone; (m)HRPC, (metastatic) hormone-resistant prostate cancer; MDS, myelodysplastic syndrome; MGUS, monoclonal gammopathy of unknown significance; MRI, magnetic resonance imaging; NHA, new hormonal agent; PARP, poly(adenosine diphosphate)-ribose polymerase; PSA, prostate specific antigen; RECIST, Response Evaluation Criteria in Solid Tumours. Source: PROfound CSP version 4, 7 March 2019.[87]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

==> picture [39 x 10] intentionally omitted <==

Settings and locations where the data were collected

The PROfound study was conducted in 111 centres across: France, Japan, Canada, Turkey, Australia, South Korea, Netherlands, United states, Italy, Taiwan, Brazil, Argentina, Israel, Germany Spain, Sweden, United Kingdom (five sites, four patients) , Austria, Denmark and Norway.[30,87]

==> picture [39 x 11] intentionally omitted <==

Trial drugs and concomitant medications

Patients in the PROfound study were randomised to receive either:

• Olaparib tablets (orally; 300 mg twice daily [bid]), or

• Investigators choice of NHA with either enzalutamide (160 mg orally once daily [od]) or abiraterone acetate (1000 mg orally od with prednisone 5 mg orally bid [prednisolone was permitted for use instead of prednisone, if necessary])

All study treatments were given continuously until BICR-assessed radiographic disease progression occurred, or until the patient discontinued treatment owing to AEs or consent was withdrawn.[87]

A list of permitted concomitant treatments is included in section 7.7 of the Clinical Study Protocol, version 4, 7 March 2019.[87] Overall, the concomitant treatments administered were generally representative of those commonly prescribed to manage side effects of olaparib or enzalutamide/abiraterone and were not considered to have impacted the study results.[61]

==> picture [39 x 10] intentionally omitted <==

Outcomes used in the economic model or specified in the scope, including primary outcome

A full list of primary, secondary, and exploratory objectives of the PROfound study is available in the Clinical Study Report, version 1, 23 October 2019 (Table 2, page 51).[61] A brief description of the primary endpoint (rPFS by BICR in Cohort A) and key secondary endpoints (specified in the NICE scope and/or used in the economic model) is provided below.

B.2.3.5.1 Primary endpoint (Cohort A)

The primary endpoint in the PROfound study was rPFS, defined as time from randomisation until the date of objective disease progression (as assessed by BICR

using RECIST version 1.1 [for soft tissue disease] or Prostate Cancer Working Group 3 [PCWG3, for bone disease]) or death (by any cause in the absence of progression) regardless of whether the patient withdrew from randomised therapy or received another anti-cancer therapy prior to disease progression.[87]

B.2.3.5.2 Secondary outcomes

The secondary outcomes of PROfound that were used in the economic model and/or specified in the NICE scope (Table 2) are described below:[87]

• BICR-assessed rPFS (in Cohort A+B) : as described above.

• Overall Survival (OS) : time from randomisation to death from any cause regardless of whether the patient withdrew from randomised therapy or received another anti-cancer therapy.

• Second progression-free survival (PFS2): time from randomisation to the earliest investigator-assessed progression event subsequent to that used for the primary variable or death.

• Time to pain progression (TTPP): Time from randomisation to the time point at which worsening in pain was observed as assessed by BPI-SF item 3. This was assessed according to whether patients were symptomatic at baseline.

• Time to first symptomatic skeletal-related event (SSRE): time from randomisation to first SSRE was defined by the use of radiation therapy to prevent or relieve symptoms, occurrence of new radiologically confirmed symptomatic pathological bone fractures (vertebral or non-vertebral) or spinal compression, or surgical intervention for bone metastasis.

• Objective response rate (ORR) by BICR : for patients who had measurable disease at baseline determined by BICR, objective response rate assessed by BICR (RECIST 1.1 and PCWG3), is defined as the number (%) of patients with at least one visit response of complete (CR) or partial response (PR), in their soft tissue disease assessed by RECIST 1.1, in the absence of progression on bone scan assessed by PCWG3.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• Health-related quality of life : the patient-reported FACT-P will be used to assess health-related quality of life. The questionnaire was administered, at baseline, weeks 8, 16 and 24, and every 8 weeks thereafter to all patients who have not withdrawn consent. The following outcome measures were calculated from the FACT-P questionnaire; the resulting value is the total score for the associated questions or scaled scores:

  • Physical well-being subscale (PWB)

  • Social/family well-being subscale (SWB)

  • Emotional well-being subscale (EWB)

  • Functional well-being subscale (FWB)

  • Prostate cancer subscale (PCS)

  • Total Functional Assessment of Cancer Therapy- General (FACT-G) score, sum of PWB, SWB, EWB and FWB.

  • Trial Outcome Index (TOI), sum of PWB, FWB and PCS.

  • Functional Assessment of Prostate Cancer Symptoms Index 6 (FAPSI6)

  • Total FACT-P score (sum of scores of all the sub-scales)

• EQ-5D-5L : The EQ-5D-5L index comprises 5 dimensions of health (mobility, self-care, usual activities, pain/discomfort and anxiety/depression). For each dimension, respondents select which statement best describes their health on that day from a possible 5 options of increasing levels of severity (no problems, slight problems, moderate problems, severe problems and unable to/extreme problems). A unique EQ-5D health state is referred to by a 5-digit code allowing for a total of 3,125 health states (for example, state 11111 indicates no problems on any of the 5 dimensions). These data were converted into a weighted health state index by applying scores from EQ-5D value sets elicited from general population samples (the base case was the UK valuation set).

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

==> picture [39 x 10] intentionally omitted <==

Biomarker analyses

An investigational clinical trial assay, based on the FoundationOne CDx nextgeneration sequencing test and developed in partnership with Foundation Medicine, was used to prospectively identify patients with qualifying deleterious or suspected deleterious alterations in at least 1 of the 15 prespecified genes, selected for their direct or indirect role in HRR, namely: BRCA1, BRCA2, ATM, BRIP1, BARD1,

CDK12, CHEK1, CHEK2, FANCL, PALB2, PPP2R2A, RAD51B, RAD51C, RAD51D , and RAD54L . Tumour testing was conducted centrally with the use of archival or recent biopsy tissue from primary or metastatic disease. The presence of a deleterious or suspected deleterious alteration according to the central tumour test was required for inclusion in the study. Further details of HRRm testing are provided in section 5.7 of the Clinical Study Protocol, version 4, 7 March 2019.[87]

==> picture [39 x 11] intentionally omitted <==

Baseline characteristics

Patient baseline characteristics are summarised Table 5 for:

• Cohort A+B (the overall population of PROfound, and the population of interest for this appraisal),

• Cohort A (the population that the primary endpoint of rPFS was analysed in) and,

• Cohort A+B prior taxane subgroup (on which the comparative clinical and cost-effectiveness evidence versus cabazitaxel in based on, which aligns with the anticipated positioning of olaparib in the current clinical pathway of care in England where the majority of patients receive a taxane [docetaxel] for HSPC, before receiving NHA for mCRPC).

Key baseline characteristics were largely well-balanced between treatment arms, and between the Cohort A+B, Cohort A, and the Cohort A+B prior taxane subgroup (Table 5).

Important prognostic variables, such as measurable disease at baseline and prior taxane use, were included as stratification factors to ensure balance across treatment arms (the size of the study did not allow for further stratification factors).

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 36 of 208

The primary endpoint in the study, rPFS (BICR), and the key secondary endpoint of OS were analysed with these stratification factors included in the model as covariates.

Minor discrepancies were observed in specific characteristics at baseline, such as ECOG status, and PSA. To assess if the results were impacted by these differences in key patient characteristics across the PROfound study treatment arms, a sensitivity analysis was performed which adjusted for XXXXXXXXXXXXXX XXXXXXXXXXXXXX.XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXX..XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX .XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 5. Patient characteristics for PROfound Cohort A+B, Cohort A, prior taxane subgroup

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 39 of 208

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 40 of 208

a Subgroup adjusting for previous taxane (yes, no), collected via IVRS

b As per investigator assessment. Patients with multiple sites of disease within the same category of extent of disease are counted only once in that category. c Derived from eCRF data.

d Proportions expressed as % of the total number of patients in the analysis set with single mutations: Cohort A+B (234 for olaparib and 118 for investigator’s choice of NHA), Cohort A (148 for olaparib and 76 for investigator’s choice of NHA), Cohort A+B prior taxane (163 for olaparib and 78 for investigator’s choice of NHA). ATM ataxia telangiectasia mutated; BARD1 BRCA1 associated ring domain protein; bid twice daily; BRCA breast cancer susceptibility gene; BRIP1 BRCA1 interacting protein C-terminal helicase 1; CDK12 cyclin-dependent kinase 12; CHEK1 checkpoint kinase 1; CHEK2 checkpoint kinase 2; FANCL FA complementation group; FAS full analysis set; HRR homologous recombination repair; NHA new hormonal agent; PALB2 partner and localizer of BRCA2; PPP2R2A protein phosphatase 2 regulatory subunit B alpha; RAD51B RAD51 paralog B; RAD51C RAD51 paralog C; RAD51D RAD51 paralog D; RAD54L RAD54 like.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 41 of 208

e Reported as a patient who received prior cisplatin and fluorouracil and paclitaxel.

f A detailed overview of co-mutations is given in Appendix M.

bid, twice daily; eCRF, electronic case report form; IVRS, interactive voice response system; NR, not reported; SD, standard deviation Source: de Bono et al 2020,[30] Clinical Study Report Edition 1 – 23 October 2019[61] and de Wit 2019[67]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca 2020 All rights reserved Page 42 of 208

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

All analyses were performed in accordance with a comprehensive statistical analysis plan (SAP), which details the analyses to be conducted, summaries produced, and the analysis sets upon which they would be based (Sections 1-3 of the PROfound SAP).[89]

The main hypothesis evaluated in the PROfound study was that single agent olaparib at 300 mg bid has superior efficacy and an acceptable tolerability profile as compared with enzalutamide or abiraterone in mCRPC patients with deleterious or suspected deleterious HRR gene mutations and whose disease has progressed after treatment with an NHA such as enzalutamide or abiraterone.[87]

==> picture [38 x 10] intentionally omitted <==

Analysis sets

The primary endpoint of the study was rPFS in Cohort A. The study planned to randomise approximately 240 patients (2:1 ratio of olaparib:investigators’ choice of NHA), with the rPFS analysis occurring once approximately 143 rPFS events (confirmed by BICR) had occurred.[87]

It was expected that the targeted sample size of 240 patients in Cohort A with approximately 143 rPFS events (i.e. 60% maturity) would provide 95% power to demonstrate a statistically significant difference in rPFS at a 2-sided alpha level of 5%, assuming that true treatment effect was a HR of 0.53, translating an ~4.5-month improvement in median rPFS with olaparib, over an assumed 5-month median duration of rPFS on enzalutamide or abiraterone. The PROfound trial met its primary endpoint in Cohort A at DCO1 (4[th] June 2019), demonstrating a statisticallysignificant and clinically-meaningful rPFS benefit versus investigators’ choice of NHA (HR, 0.34 [95% CI, 0.25–0.47], p < 0.0001).[87]

Cohort B of PROfound was an exploratory cohort and was designed to include ~100 patients with qualifying HRR gene mutations other than BRCA1, BRCA2, and ATM .[87]

Analysis of efficacy and patient-reported outcomes was conducted on the full analysis set (FAS) of Cohort A, Cohort B, Cohort A+B, whilst safety analyses were based on the safety analysis set (FAS) for each Cohort, as defined in Table 6.

Table 6. Definition of populations.

FAS, full analysis set; PRO, patient-reported outcome. Source: PROfound CSP version 4, 7 March 2019.[87 ]

==> picture [39 x 10] intentionally omitted <==

Outcome measures and statistical analysis

All calculations were performed with statistical analysis software (SAS[®] ) Version 9.4 (SAS Institute, Inc, Cary, North Carolina, US), unless otherwise stated. Further information on sample size calculation and analysis of key outcome variables (including supporting sensitivity and subgroup analyses, and censoring) are described in detail in Section 9.8 of the PROfound Clinical Study Report, version 1, 23 October 2019.[61]

The PROfound study used a multiplicity strategy for statistical testing of primary endpoints and key secondary endpoints (Figure 2). Per study protocol, the primary analysis was performed when ~143 rPFS (60%) events in Cohort A had occurred, based on BICR assessment. Upon achieving statistical significance on the primary endpoint rPFS in Cohort A, testing of each of the secondary endpoints, i.e. ORR (Cohort A), rPFS (Cohort A + B), time to pain progression (Cohort A), and overall survival (Cohort A) was performed sequentially with the 2-sided 5% level of alpha recycled from the primary rPFS (Cohort A) endpoint (Figure 2). The data cut-off for the primary rPFS analysis (DCO1) occurred on 4[th] June 2019.[61]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Per the study protocol, the final analysis of OS was performed when ~146 (61%) OS events had occurred in Cohort A (DCO2, 20[th] March 2020). Safety summaries were also updated at the time of this analysis.[87] Top-line data from the final OS analysis are presented in Section B.2.6.3.1; further analyses are currently underway.[31] These will be provided to NICE as soon as possible.

Figure 2. Summary of the hierarchical testing structure for PROfound.

==> picture [299 x 138] intentionally omitted <==

==> picture [299 x 139] intentionally omitted <==

BICR, blinded independent central review; ORR, objective response rate; OS, overall survival; rPFS, radiographic progression-free survival

Source: Clinical Study Report, version 1, 23 October 2019.[87]

An overview of the analysis methods for key efficacy outcomes used in the economic model and/or included in the final NICE scope is provided in Table 7.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 7. Description of outcomes and methodology for statistical analysis.

a Although it is expected that there will be enough rPFS events in each strata (where strata are defined as categories formed from – prior taxane * measurable disease * treatment) to allow a meaningful analysis, if any stratum for either treatment arm contains less than 5 events, Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 47 of 208

then a pooling strategy will be employed. The order of preference for pooling will be (prior taxane * treatment), (measurable disease * treatment), unstratified. In addition, for analyses on Cohort A+B, Cohort will be added as a stratification factor provided that the addition does not lead to <5 events in any strata. Prior taxane and measurable disease will use data collected via IVRS. The pooling strategy will be employed for Cohort A, Cohort B and Cohort A+B separately. All sensitivity analyses and secondary endpoints (except for ORR which only includes prior taxane) will use the same strata as the primary model, for that endpoint, unless there are <5 events per stratum and then an unadjusted model will be used.

BICR, blinded independent central review; BPI-SF, Brief Pain Inventory – Short Form; CI, confidence interval; FACT-G, Functional Assessment of Cancer Therapy – General; FACT-P, Functional Assessment of Cancer Therapy – Prostate; FAPSI-6, 6-item Functional Assessment of Cancer Therapy Advanced Prostate Symptom Index; FWB, functional wellbeing; g BRCA m, germline BRCA mutation; HR, hazard ratio; KM, Kaplan–Meier; OS, overall survival; PCWG2, Prostate Cancer Working Group 2; PCS, prostate cancer subscale; PFS2, second progression-free survival; PRO, patient-reported outcome; PWB, physical wellbeing; RECIST, Response Evaluation Criteria in Solid Tumours; rPFS, radiographic progression-free survival; SAP, statistical analysis plan; SSRE, symptomatic skeletal-related event; TOI, Trial Outcome Index; TTPP, time to pain progression. Source: PROfound CSP version 4, 7 March 2019.[87]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca 2020 All rights reserved Page 48 of 208

B.2.5 Quality assessment of the relevant clinical effectiveness evidence

The PROfound study was performed in accordance with the ethical principles that have their origin in the Declaration of Helsinki[90] and that are consistent with International Council for Harmonisation (ICH)/Good Clinical Practice (GCP),[91] applicable regulatory requirements and the AstraZeneca policy on Bioethics.

A complete quality assessment in accordance with the NICE-recommended checklist for assessment of bias in RCTs is presented in Table 8 and Appendix D. The risk of bias in the PROfound study is confirmed as being low.

Table 8. Overview of quality assessments for the PROfound study

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

BoR, best objective response; CSP, Clinical Study Protocol; CSR, Clinical Study Report; DoR, duration of response; ECOG, Eastern Cooperative Oncology Group; EFR,evaluable for response;

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

FAS, full analysis set; ORR, objective response rate; OS, overall survival; PSA, prostate-specific androgen; RECIST, Response Evaluation Criteria In Solid Tumours; rPFS, radiographic progressionfree survival; SAP, statistical analysis plan; SAS, safety analysis set Source: PROfound Clinical Study report, version 1, 23 October 2019,[61] PROfound Clinical Study Protocol, version 4, 7 March 2019[87] and de Bono et al . 2020.[30]

B.2.6 Clinical effectiveness results of the relevant trials

As discussed in section B.2.1, clinical effectiveness evidence on olaparib (the intervention of interest) and cabazitaxel (the main comparator of interest in current clinical practice in England) were derived from the PROfound[30,61] and CARD studies,[67] respectively. This section focuses on data on olaparib from the pivotal Phase III PROfound study. Key data on cabazitaxel from the CARD study are summarised in Appendix D, section D.1.4.; comparative analysis of olaparib versus cabazitaxel via an anchored indirect treatment comparison is described in section B.2.9.

==> picture [38 x 11] intentionally omitted <==

PROfound: summary of clinical data

An overview of the data provided in Sections B.2.6 and B.2.7 is presented in Figure 3.

The key efficacy outcomes in Cohort A+B (the overall HRRm population in the PROfound study), Cohort A (wherein the primary endpoint of BICR rPFS was evaluated), and Cohort A+B prior taxane subgroup of PROfound (the main focus for comparative clinical- and cost-effectiveness analysis versus cabazitaxel, and the population reflective of the positioning of olaparib in clinical practice, where the majority of patients receive docetaxel [with ADT] for HSPC) are summarised in Table 9.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 3. An overview of the PROfound trial and its placement in the current submission.

==> picture [463 x 131] intentionally omitted <==

==> picture [463 x 130] intentionally omitted <==

==> picture [463 x 131] intentionally omitted <==

aIncludes BICR-assessed rPFS (primary analysis) and investigator-assessed rPFS (sensitivity analysis)

bIncludes unadjusted OS, and OS with treatment switching adjustment

cBaseline patient characteristics are presented in section B.2.3.7, not section B.2.7

ATM, gene for ataxia–telangiesctasia mutated; BRCA1 , gene for breast cancer type 1 susceptibility protein; BRCA2 , gene for breast cancer type 2 susceptibility protein; DoR, duration of response; HRRm, homologous recombination repair mutation; NICE, National Institute for Care and Health Excellence; ORR, objective response rate; OS, overall survival; PFS2, second progression-free survival; rPFS, radiographic progression-free survival; SSRE, symptomatic skeletal-related events; TTPP, time to pain progression.

Table 9. Key efficacy outcomes in Cohort A, Cohort A+B, Cohort A+B prior taxane subgroup (used in ITC analysis).

aAlpha spend was 0.01 at the interim analysis (DCO1); therefore, statistical significance was not reached.

bResults presented using RPSFTM method (Weibull, with no re-censoring). Other methods explored as per NICE DSU16 (Section A.7.2).92

bid, twice daily; CI, confidence interval; HR, hazard ratio; NC, not calculable; ORR, objective response rate; OS, overall survival; PFS2, second progression-free survival; rPFS, radiographic progression-free survival; RPSFTM: Rank Preserving Structural Failure Time Models; TTPP, time to pain progression. Source: de Bono et al 2020,[30] Clinical Study Report, version 1, 23 October 2019[61] and PROfound treatment switching analysis report.[93]

==> picture [39 x 10] intentionally omitted <==

rPFS (BICR), DCO1 (4[th] June 2019)

B.2.6.2.1 Cohort A (primary endpoint)

The primary analysis of rPFS took place when 174 progression events had occurred (71% maturity) in Cohort A. The PROfound study met its primary endpoint at the time of this analysis, with olaparib treatment achieving a statistically-significant and clinically-meaningful improvement in median BICR rPFS compared with investigators’ choice of NHA in Cohort A (7.39 months vs 3.55 months; HR, 0.34 [95% CI, 0.25–0.47]; p < 0.0001, Figure 4). There was clear separation of the Kaplan-Meier (KM) curves in favour of olaparib; this separation started at

approximately 2 months (coinciding with the first planned tumour assessment) and was maintained throughout most of the study follow-up period. These data are supported by landmark rPFS assessments at 6- and 12-months, wherein ~60% and 28% of patients in the olaparib arms were alive and progression-free, respectively (versus, ~23% and 9% of patients in the investigators’ choice of NHA arm).[30,61]

Figure 4. Primary outcome: Kaplan–Meier plot of BICR-assessed rPFS in Cohort A.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

a Disease progression, as assessed by BICR and defined by RECIST version 1.1 and/or PCWG3 or death (by any cause in the absence of progression) regardless of whether the patient withdrew from randomised therapy or received another anti-cancer therapy before progression.

BICR, blinded independent central review; bid, twice daily; CI, confidence interval; HR, hazard ratio; PCWG3, Prostate Cancer Working Group 3; RECIST, Response Evaluation Criteria In Solid Tumours; rPFS, radiographic progression-free survival.

Source: de Bono et al 2020[30] and CSR edition 1, 23 October 2019.[61]

B.2.6.2.2 Cohort A+B (key secondary endpoint)

In the overall HRRm population (Cohort A+B), olaparib was associated with a statistically-significant and clinically-meaningful improvement in median BICR rPFS, versus investigators’ choice of NHA (median rPFS, 5.82 months vs 3.52 months, respectively; HR, 0.49 [95% CI: 0.38–0.63; p < 0.0001]), as shown in Figure 5. As with Cohort A, there was clear separation of the Kaplan–Meier curves in favour of olaparib; this separation started at approximately 2 months (coinciding with the first planned tumour assessment), and was maintained for the majority of the study follow-up period. ~50% and ~25% of patients remained alive and progressionfree in the olaparib arm at the time of 6-month and 12-month landmark rPFS assessments , respectively (versus just 22% and 14% of patients in the ’ investigators choice of NHA arm), which is remarkable in this heavily pre-treated mCRPC population of patients.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 5. Key secondary outcome: Kaplan–Meier plot of BICR-assessed rPFS in patients Cohort A+B

==> picture [452 x 125] intentionally omitted <==

==> picture [452 x 124] intentionally omitted <==

a Disease progression, as assessed by BICR defined by RECIST version 1.1 and/or PCWG3 or death (by any cause in the absence of progression) regardless of whether the patient withdrew from randomised therapy or received another anti-cancer therapy before progression.

BICR, blinded independent central review; bid, twice daily; CI, confidence interval; HR, hazard ratio; PCWG3, Prostate Cancer Working Group 3; RECIST, Response Evaluation Criteria In Solid Tumours; rPFS, radiographic progression-free survival.

Source: de Bono et al 2020[30] and CSR edition 1, 23 October 2019.[61]

B.2.6.2.3 Pre-specified sensitivity and subgroup analyses

The rPFS benefit achieved with olaparib treatment was maintained in a range of prespecified sensitivity and subgroup analysis (see sections 11.1.1. and 11.1.2.2. of the PROfound Clinical Study Report, version 1, 23 October 2020[61] and Appendix E), demonstrating a robust and consistent treatment effect across potential or expected

prognostic factors. Subgroup analyses of eight pre-specified baseline characteristics (including the stratification factors of prior taxane [yes/no] and baseline metastases

[yes/no]) showed clinically-meaningful reductions in the risk of radiological disease Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

progression or death in patients who received olaparib (ranging from 39% to 75% in Cohort A and from 23% to 88% in Cohort A+B). These data, as well as analyses by single HRRm status, are further described in Appendix E. The prior taxane subgroup of Cohort A+B – the focus of the comparative clinical- and cost-effectiveness evidence versus cabazitaxel, is further discussed in Section B.2.7.

B.2.6.2.4 Post-progression anticancer therapies (Cohort A + B), DCO1 (4[th] June 2019)

Subsequent anticancer therapies were received by a lower percentage of patients in the olaparib arm compared with the investigators’ choice of NHA arm: 29.0% versus 68.7% in Cohort A, and 35.2% versus 63.4% in Cohort A+B (see Appendix M). Of these patients, treatment with a subsequent PARP inhibitor (olaparib) occurred in 1.9% (3/162) and 1.2% (3/256) of patients in the olaparib arm of Cohort A and Cohort A+B, respectively, versus, 61.4% (51/83) and 57.3% (75/131) of patients in the investigators’ choice of NHA arm. The high percentage of patients in the investigators’ choice of NHA arm who had treatment with a subsequent PARP inhibitor is consistent with the study design, which allowed patients to switch to olaparib treatment following BICR-assessed progression on investigators’ choice of NHA. Other commonly reported subsequent treatments included hormonal therapy/taxane chemotherapy, which is consistent with clinical practice. A full list of subsequent therapies for Cohort A and Cohort A+B is provided in Appendix M.

==> picture [39 x 10] intentionally omitted <==

OS, DCO1 (4[th] June 2019)

At the 4[th] June 2019 data cut-off (DCO1), OS data were immature (Cohort A: 38% data maturity; Cohort A+B: 41% data maturity).

In Cohort A, 56.8% of patients in the olaparib arm and 48.2% of the investigators’ choice of NHA arm were alive and in the survival follow-up. Olaparib was associated with a clinically-meaningful median OS benefit of 3.4 months compared with investigators’ choice (median OS, 18.50 months versus 15.11 months; HR, 0.64

[95% CI, 0.43–0.97]; p = 0.0173). This benefit was observed despite most patients

(61.4%) in the investigators’ choice of NHA arm switching to olaparib treatment upon BICR-confirmed disease progression and confounding the OS analysis. The interim OS analysis was not statistically significant because the alpha spend at DCO1 was

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

0.01; however, a statistically significant survival benefit was achieved during the final OS analysis (DCO2), making olaparib the first and only PARP-inhibitor to show a significant OS benefit in mCRPC patients with a BRCA1 , BRCA2 and ATM gene mutations; these data are discussed in Section B.2.6.4.

Figure 6. Key secondary outcome: Kaplan–Meier plot of interim OS in patients in Cohort A.

==> picture [452 x 240] intentionally omitted <==

aAlpha spend was 0.01 at the interim analysis; therefore, statistical significance was not reached. bCalculated using the Kaplan–Meier method.

CI, confidence interval; HR, hazard ratio; OS, overall survival.

Source: Clinical Study Report Edition 1 – 23 October 2019.[61 ]

For the Cohort A+B, 53.9% of patients in the olaparib arm and 44.3% of patients in

the investigators’ choice of NHA arm were alive and in the survival follow-up at the

time of DCO1. As in Cohort A, treatment with olaparib was associated with a

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

clinically-meaningful median OS benefit of 3.4 months in Cohort A+B compared with investigators’ choice of NHA, despite most eligible patients

(81.8%) in the investigators’ choice of NHA arm switching to olaparib following BICR-assessed progression (median OS, 17.51 months versus 14.26 months, respectively; HR, 0.67 [95% CI, 0.49–0.93]; Figure 7). The Kaplan–Meier curves for olaparib and investigators’ choice of NHA separated early (in favour of olaparib) and remained separated for the majority of the follow-up period (Figure 7).

Figure 7. Secondary outcome: Kaplan–Meier plot of interim OS in patients in Cohort A+B

==> picture [452 x 125] intentionally omitted <==

==> picture [452 x 126] intentionally omitted <==

a Calculated using the Kaplan–Meier method. bid, twice daily; CI, confidence interval; HR, hazard ratio; OS, overall survival. Source: de Bono et al 2020[30] and Clinical Study Report Edition 1 – 23 October 2019.[61]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

B.2.6.3.1 Treatment switching adjustment

As stated above, switching from investigators’ choice of NHA to olaparib treatment was permitted upon BICR-confirmed radiographic progression up to DCO1, if deemed appropriate for the patient (see section B.2.3.1). Subsequently, a substantial number of patients in the investigators’ choice of NHA arm switched treatment to olaparib (51/83 patients in Cohort A, and 75/131 patients in Cohort A+B), thus confounding the OS analysis. Given that olaparib is not currently approved or reimbursed in this treatment setting in the UK (i.e. after disease progression on two lines of NHA), treatment switch adjustment analyses were performed per NICE DSU TSD16 guidance,[92] to estimate the true OS benefit of olaparib compared with investigators’ choice of NHA.

Multiple naïve and sophisticated adjustment methods were explored for the treatment switching analysis, in line with guidance from NICE DSU TSD16. The sophisticated methods included:

• Rank Preserving Structural Failure Time Model (RPSFTM)

• Inverse Probability of Censoring Weights (IPCW)

• Two-stage estimation (TSE).

The treatment switching analyses were performed using R (R Foundation). Detailed methodological considerations for the choice of adjustment method are provided in the technical report.[93] Of the aforementioned methods, the TSE approach was excluded, as an appropriate secondary baseline could not be identified, with the method considered to provide biased results. IPCW and RPSFTM methods were

compared, with the RPSFTM approach deemed the most appropriate on the basis that it is not dependent on data, particularly time-varying data, to predict switching. The RPSFTM approach also utilises all data for switchers and non-switchers, compared with the IPCW approach, which involve analysis on reduced sample sizes. This issue of reduced sample size is particularly important in the case of the PROfound data, due to the relatively small sample size of the investigators’ choice of NHA arm when divided into switchers/non-switchers.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

In addition to this, the IPCW approach is also dependent on the ‘no unmeasured confounders’ assumption, i.e. that all baseline covariates and time-dependent confounders that predict switching and outcomes are included. This assumption may not hold when there is relatively little prognostic data collected post-randomisation, limiting the scope of time-varying covariables that can be included in an analysis, as is the case with the PROfound data. In contrast to the IPCW method, the RPSFTM approach does not rely on the ‘no unmeasured confounders’ assumption; however, it does rely on clinical and biological plausibility of the randomisation and common treatment effect assumptions, described in the technical report. The randomisation assumption was shown to hold in the analysis through plots comparing the counterfactual OS KM curves of the reference and comparator arms (see technical report for details). To investigate the common treatment effect assumption, a sensitivity analysis was included where a proportion of the olaparib treatment effect was applied to those switching to olaparib from investigators’ choice of NHA. This showed that if the treatment effect were to decrease post-progression, it would still result in an overall benefit for the patients who switch, suggesting that the analysis is robust to changes in treatment effect over time.

The choice of model used to calculate the acceleration factor for the RPSFTM is based on the plausibility of the assumptions each model makes and the analyst’s preferences. For example, the log rank model gives equal weight to survival times, but a Weibull or Cox model may be preferred as they can include the trial stratification factors. The acceleration factor was consistent across each model; the fully-parametric Weibull model was preferred .

Re-censoring was performed to assess the impact of informative censoring on the results. In this analysis of the PROfound data, the results were consistent with and without re-censoring (Table 10). Therefore, to utilise the longer-term counterfactual data, re-censoring was not considered in the preferred analysis.

Using the RPSFTM method, the adjusted median OS for the investigators’ choice of NHA arm ranged from XXXXXXXXXX (according to test/censoring) in Cohort A+B; corresponding HRs for olaparib versus investigators’ choice ranged from XXXXXXX in Cohort A+B. Results for these analyses are summarised in Figure 8, with the

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca 2020 All rights reserved Page 61 of 208

preferred approach (RPSFTM Weibull model with no re-censoring) highlighted. Using this preferred approach, the OS gain demonstrated in with olaparib versus switch adjusted investigator’s choice of NHA was XXXXXXXX. in Cohort A (XXXXXX XXXXXXXXXXXX.) and XXXXXXXXX in Cohort A+B (XXXXXXXXXXXXXXXXXXX,

see Figure 8). These results were highly consistent with the IPCW method, which produced an adjusted median OS of XXXXXXXXX in Cohort A (XXXXXXXXXXXX. XXXXXX..) and XXXXXXXXX. in Cohort A+B (XXXXXXXXXXXXXXXXXXX.). This consistency of the IPCW method, which applies different assumptions to the

RPSFTM method, supports the estimated ‘true’ OS difference for olaparib compared with investigators’ choice of NHA.

Figure 8. Kaplan–Meier plot of counterfactual overall survival in Cohort A+B - (RPSFTM Weibull method, no re censoring).

==> picture [349 x 352] intentionally omitted <==

Treatment-switching adjusted Kaplan–Meier curves for Cohort A are available upon request. KM, Kaplan–Meier; RPSFTM, Rank Preserving Structural Failure Time Model.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 10. Median OS and HR for investigators’ choice of NHA arm, adjusted for treatment switching; overall HRRm (Cohort A+B) and Cohort A

==> picture [452 x 553] intentionally omitted <==

----- Start of picture text -----
Test Re- Cohort A+B Cohort A
censorin
g Median OS OS HR (95% Median OS OS HR (95%
(months) for CI) olaparib [a] (months) for CI) olaparib [a]
investigator vs. investigator vs.
s choice of investigators s choice of investigators
NHA ’ choice of NHA ’ choice of
adjusted for NHA adjusted for NHA
switching switching
RPSFTM
Log rank Without XXX.. XXXXXXX. XXXX XXXXXXX.
XX. XXX.
With XXX.. XXXXXXX. XXXX XXXXXXX.
XX.. XXX.
Cox Without XXX.. XXXXXXX. XXXX XXXXXXX.
proportiona XX.. XXX.
l hazards
With XXX.. XXXXXXX. XXXX XXXXXXX.
XX.. XXX.
Weibull Without XXX.. XXXXXXX. XXXX XXXXXXX.
XX.. XXX.
With XXX.. XXXXXXX. XXXX XXXXXXX.
XX.. XXX.
IPCW
Adjusted N/A XXX.. XXXXXXX. XXXX XXXXXXX.
for XX.. XXX.
switching
using
IPCW
----- End of picture text -----

a Median OS with olaparib was 17.51 months as presented in Section B.2.6.2.

Note: these data are used in the ITC comparison, section B.2.9.

CI, confidence interval; HR hazard ratio; IPCW, Inverse Probability of Censoring Weights; ITC, indirect treatment comparison; NHA, new hormonal agent; OS, overall survival; RPSFTM, rank preserving structural failure time model.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

==> picture [39 x 10] intentionally omitted <==

OS, DCO2 (20[th] March 2020)

A final analysis of OS was planned for Cohort A at approximately 61% maturity of the data (DCO2). At DCO2, the survival benefit of olaparib over investigators’ choice was confirmed in the key secondary endpoint of OS in Cohort A (Figure 9). OS in Cohort A+B was a secondary endpoint, and at DCO2 again demonstrated a trend to survival benefit over investigators’ choice of NHA (Figure 10). Notably, this trend in OS benefit was observed despite >80% of patients in the investigators’ choice of NHA arm who were eligible to switch to olaparib in Cohort A+B receiving olaparib.

Figure 9. Secondary outcome: Kaplan–Meier plot of final OS in patients in Cohort A.

==> picture [446 x 126] intentionally omitted <==

==> picture [446 x 126] intentionally omitted <==

==> picture [452 x 127] intentionally omitted <==

----- Start of picture text -----
Key secondary outcome: final Olaparib 300 mg bid Investigators’ choice of
OS – Cohort A (n = 162) NHA
(n = 83)
Events, n (%) XXXX XXXX
Median OS, months (95% CI) XXXX XXXX
HR (95% CI) XXXX
OS at 6 months, % [a] XXXX XXXX
OS at 12 months, % [a] XXXX XXXX
----- End of picture text -----

• *0.047 alpha spent at the final OS analysis. Maturity rate: 60%

bid, twice daily; CI, confidence interval; HR, hazard ratio; NHA, new hormonal agent; OS, overall survival.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 10. Secondary outcome: Kaplan–Meier plot of interim OS in patients in Cohort A+B.

==> picture [446 x 132] intentionally omitted <==

bid, twice daily; CI, confidence interval; HR, hazard ratio; NHA, new hormonal agent; OS, overall survival.

Treatment switch analyses on data from DCO2 are currently in progress and will be supplied to NICE as soon as they become available.

==> picture [39 x 11] intentionally omitted <==

Other key secondary endpoints

rPFS and OS data in the PROfound study were supported by a range of clinicallyrelevant secondary endpoints, including time from randomisation to second progression or death (PFS2), time to pain progression (TTPP), time to first symptomatic skeletal-related event (SSRE), and objective response rate, all of which showed meaningful improvements for olaparib versus investigators’ choice of NHA. Collectively, these data highlight important patient benefits achieved with olaparib treatment, beyond extending survival, and are briefly summarised below. The focus

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

of discussion is the overall HRRm study population of PROfound, i.e. Cohort

A+ B ; further details on Cohort A are available in the Clinical Study Report, version 1, 23 October 2019, section 11.1. These data were not analysed in the prior taxane subgroup of Cohort A+B at the time of submission.[61]

B.2.6.5.1 PFS2, DCO1 (4[th] June 2019)

PFS2 is an intermediate endpoint between PFS and OS which reflects real-life treatment decisions and patient experience. Its use is recommended by the EMA to capture potential negative impacts on next-line therapy and to demonstrate that any potential tolerability concerns are outweighed by treatment benefit.[60] This is especially important in the heavily pre-treated mCRPC setting relevant to this appraisal, where even a small delay in disease progression and consequently overall survival is considered clinically meaningful.

Treatment with olaparib was associated with a XX-month improvement in

median PFS2 compared with investigators’ choice of NHA in Cohort A+B (XXXXX.. XXXXXXXXXXXX..; Figure 11). As with OS, this is a remarkable benefit given that a high proportion of patients had switched to olaparib upon disease

progression (by BICR) on investigators’ choice of NHA. Since olaparib is not available (approved or reimbursed) in England in this setting, this analysis underestimates the true PFS2 benefit of olaparib treatment and biases the results in favour of the control arm.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 11. Secondary outcome: Kaplan–Meier plot of interim PFS2 by investigator assessment in patients in the overall HRRm population (Cohort A+B).

==> picture [446 x 131] intentionally omitted <==

a Calculated using the Kaplan–Meier technique.

CI, confidence interval; HR, hazard ratio; PFS2, second progression-free survival. Source: CSR edition 1, 23 October 2019.[61]

The PFS2 advantage with olaparib in Cohort A was broadly consistent with that

observed in Cohort A+B (median PFS2 = XXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.];; see the Clinical Study Report,

version 1, 23 October 2019, section 11.1.3.7.).

B.2.6.5.2 Time-to-pain progression, DCO1 (4[th] June 2019)

TTPP is a patient-relevant endpoint in mCRPC, with many patients experiencing substantial pain within bone tissue due to the location of their

tumours (Section B.1.2). The Kaplan–Meier curves for TTPP in Cohort A+B (based on the Brief Pain Inventory - short form [BPI-SF] worst pain and opiate use items) separated from three months onwards and remained separated in favour of olaparib

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

for the duration of study follow-up. Median duration of TTPP was not reached in either arm at the time of DCO1 (HR, 0.64; 95% CI: 0.35–1.21; Figure 12). At 12 months, over three-quarters of the patients in the olaparib arm had no pain progression (versus just 50% of patients in the investigator’s choice of NHA arm).

Figure 12. Secondary outcome: Kaplan–Meier plot of interim TTPP by investigator assessment in patients in overall HRRm population (Cohort A+B).

==> picture [451 x 257] intentionally omitted <==

bid, twice daily; CI, confidence interval; HR, hazard ratio; NR, not reached; TTPP, time to pain progression.

Source: CSR edition 1, 23 October 2019.[61]

Treatment with olaparib was also associated with a statistically significant

delay in TTPP compared with investigators’ choice of NHA in Cohort A (median

– TTPP: not reached vs 9.92 months, respectively; HR, 0.44 [95% CI, 0.22 0.91];

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

p = 0.0192; see the Clinical Study Report, version 1, 23 October 2019, section 11.1.2).

B.2.6.5.3 Time to first symptomatic skeletal-related event, DCO1 (4[th] June 2019)

In mCRPC, the SSREs are a further indicator of worsening bone health due to tumour growth, and usually require further treatment. In the PROfound study, SSREs were defined as:

• Use of radiation therapy to prevent or relieve skeletal symptoms.

• Occurrence of new symptomatic pathological bone fractures (vertebral or nonvertebral) (radiologic documentation required).

• Occurrence of spinal cord compression (radiologic documentation required).

• Orthopaedic surgical intervention for bone metastasis.

The incidence of SSREs was lower in the olaparib arm than in the investigators’ choice of NHA arm in Cohort A+B (XXXXXXXXXXXXXXXXXXXXX.;; Table 11), highlighting an important benefit of olaparib treatment for patients in prolonging (potentially debilitating) and burdensome SSREs.

Table 11. Time to first SSRE in patients in Cohort A+B.

bid., twice daily; CI, confidence interval; HR, hazard ratio; SSRE, symptomatic skeletal-related event. Source: CSR edition 1, 23 October 2019.[61]

A similar benefit was observed in Cohort A: 15.4% of patients in the olaparib arm

and 22.9% of patients in the investigators’ choice of NHA arm had experienced a first SSRE and the HR favoured olaparib (0.37; 95% CI: 0.20, 0.70). Further details are available in the CSR (Section 11.1.3.8.).

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

B.2.6.5.4 Overall response rate, DCO1 (4[th] June 2019)

Olaparib was associated with a clinically meaningful improvement in BICR confirmed ORR compared with investigators’ choice of NHA in Cohort A+B (odds ratio [OR], – 5.93 [95% CI, 2.01 25.40]; Figure 13). A total of 21.7% of patients in the olaparib arm achieved an objective response, compared with just three patients (4.5%) in the investigators’ choice of NHA arm (Figure 13). These results were consistent with those observed for olaparib compared with investigators’ choice of NHA in Cohort A (ORR, 33.3% vs 2.3%; OR, 20.86 [95% CI, 4.18–379.18]; p < 0.0001; see the Clinical Study Report, version 1, 23 October 2019, section 11.1.2.1)

Figure 13. Secondary outcome: confirmed BICR-assessed radiological ORR in Cohort A+B.

==> picture [302 x 185] intentionally omitted <==

a Radiological ORR based on BICR-assessed RECIST version 1.1 and bone scan data (using all scans regardless of whether they were scheduled or not) in patients with measurable disease. BICR, blinded independent central review; bid, twice daily; CI, confidence interval; CR, complete response; OR, odds ratio; ORR, overall response rate; PR, partial response; RECIST, Response Evaluation Criteria In Solid Tumours.

Source: CSR edition 1, 23 October 2019.[61]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

==> picture [39 x 10] intentionally omitted <==

HRQoL, DCO1 (4th June 2019)

B.2.6.6.1 EQ-5D (predefined exploratory endpoint)

Baseline and overall compliance rates for the EQ-5D-5L were similar for the two treatment arms in Cohort A+B (baseline compliance: olaparib, XXX% vs investigators’ choice of NHA, XXX%; overall compliance: XXX% vs investigators’ choice of NHA, XXX%).

There was no meaningful change observed in the mean individual domain scores from baseline through to Week 64 across both treatment arms (Figure 14). Across both treatment arms, for patients who had evaluable assessments throughout the study period, their health state stayed the same. There was no change observed in the VAS from baseline to Week 64 across both treatment arms. Overall, the EQ-5D5L data supported no worsening and no deterioration of individual domain scores or the VAS in the olaparib arm compared to the investigators’ choice of NHA arm in Cohort A+B (Figure 14) or in Cohort A (data not shown; see Clinical Study Report, version 1, 23 October 2019, section 11.1.5.1).[61]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 14. Mean change from baseline EQ-5D-5L scores up to week 64 (Cohort A+B).

==> picture [446 x 126] intentionally omitted <==

==> picture [446 x 125] intentionally omitted <==

==> picture [446 x 126] intentionally omitted <==

==> picture [446 x 126] intentionally omitted <==

bid, twice daily; EQ-5D, 5-dimension EuroQol questionnaire; SD, standard deviation; VAS, visual analogue scale.

Source: Clinical Study Report, version 1, 23 October 2019.[61 ]

B.2.6.6.2 FACT-P, DCO1 (4th June 2019)

Changes in Functional Assessment of Cancer Therapy – Prostate Cancer (FACT-P) total and subscale scores – a more-sensitive disease-specific PRO instrument - were

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

analysed using a mixed model repeated measures (MMRM) analysis of all of the post-baseline scores for each visit and were in favour of olaparib vs investigators’ choice of NHA.

The time to deterioration in FACT-P Total and all subscale scores (with the exception of functional well-being, FWB) for Cohort A+B numerically favoured patients in the olaparib arm compared with investigators’ choice of NHA arm, with HRs ranging from 0.68 to 0.94 (see Appendix M for details). For FWB, the HR suggested no detriment with olaparib treatment compared with investigators choice of NHA treatment.

Similar findings were observed in Cohort A, with time to deterioration in FACT-P Total and all subscale scores numerically favouring patients in the olaparib arm compared with the investigator’s choice of NHA arm (with HRs ranging from 0.74 to 0.95); as described in the Clinical Study Report, version 1, 23 October 2019, section 11.1.4.6.[61]

Collectively, these data (in conjunction with other PRO data provided in the Clinical Study Report, version 1, 23 October 2019[61] ) highlight that the efficacy of olaparib (for instance, in delaying radiographic disease progression, time to pain progression, and time to first symptomatic skeletal-related event) translates into meaningful improvements in patients’ health-related quality-of-life and support a favourable risk: benefit profile for olaparib for the treatment of mCRPC patients with qualifying HRR gene mutations.

B.2.7 Subgroup analysis

In current clinical practice in England, the majority of mCRPC patients (~75%) have already received a treatment with a taxane (docetaxel; in combination with ADT for HSPC), prior to receiving treatment with NHA. The “prior taxane” subgroup of the PROfound study is thus most representative of the population of patients who would be eligible to receive treatment with olaparib in clinical practice in England. This population also forms the basis of the comparative clinical-effectiveness and costeffectiveness analysis versus cabazitaxel, the most-commonly used treatment and standard-of-care in current practice, after disease progression on docetaxel and a NHA (presented in Section B.2.9 and Section B.3, respectively). This heavily pre-

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

treated population represents an area of significant unmet need, with median OS of just 13.6 months achieved in the CARD study (the largest RCT of cabazitaxel in the post-NHA setting).

In the PROfound study, 254 of 387 patients in the overall HRRm study population (Cohort A+B) had received prior treatment with a taxane: 170 patients in the olaparib arm and 84 patients in the investigators’ choice of NHA arm. Randomisation in the study was stratified by prior taxane use (yes/no), thus preserving balance amongst patients randomised to the treatment and control arm. The majority of patients (173 of 254; 68%) had received docetaxel only (as would be expected in clinical practice). Some patients had received both docetaxel and cabazitaxel; however, proportions of these patients was well-balanced across olaparib and investigators’ choice of NHA arms and is thus not expected to impact upon the study results/interpretation. Just 3 patients (of 254) had received cabazitaxel only. Detailed patient characteristics at baseline in the prior taxane group are shown in Section B.2.3.7, Table 5).

In this section, we describe key data (rPFS and OS) for olaparib versus investigators’ choice of NHA in the prior taxane group. These data were used in comparative clinical- and cost-effectiveness for olaparib versus cabazitaxel described in Section B.2.9 and B.3, respectively). It is worth noting that although the prior taxane group is the focus of this submission, olaparib also achieved meaningful clinical benefit in those patients who had not received a taxane prior to randomisation in the study (rPFS HR, 0.77; 95% CI: 0.50–1.22, Cohort A+B), highlighting an important

benefit with olaparib treatment in this group of patients, who (if

contraindicated or otherwise unsuitable for treatment with taxanes) have very limited treatment options .

==> picture [38 x 10] intentionally omitted <==

rPFS (BICR), DCO1 (4[th] June 2019)

At the time of the primary rPFS analysis (DCO1), 124 and 70 events, respectively, had occurred in the olaparib arm and the investigators’ arm of the Cohort A+B prior taxane group. Treatment with olaparib resulted in a remarkable 69% reduction in the risk of radiographic disease progression in this subgroup versus investigators’ choice of NHA (median rPFS, 5.82 months vs 2.56 months, respectively; HR; 0.39; 95% CI,

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

0.29–0.53; Figure 15). As with the overall Cohort A+B population, the KM-curves for olaparib and investigators’ choice of NHA in the prior taxane group separated early, in favour of olaparib, and remained separated for entire the duration of the follow-up period, supporting a sustained rPFS benefit with olaparib treatment and addressing a key unmet need in this advanced and heavily pre-treated patient population.

Figure 15. Kaplan–Meier plot of BICR-assessed rPFS in Cohort A+B patients whom had prior taxane treatment

==> picture [449 x 271] intentionally omitted <==

a Disease progression, as assessed by BICR defined by RECIST version 1.1 and/or PCWG3 or death (by any cause in the absence of progression) regardless of whether the patient withdrew from randomised therapy or received another anti-cancer therapy before progression.

BICR, blinded independent central review; bid, twice daily; CI, confidence interval; HR, hazard ratio; PCWG3, Prostate Cancer Working Group 3; RECIST, Response Evaluation Criteria In Solid Tumours; rPFS, radiographic progression-free survival.

Source: PROfound subgroup analyses, prior taxane Cohort A+B[94]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

==> picture [39 x 10] intentionally omitted <==

OS, DCO1 (4[th] June 2019)

At DCO1, 73 events (42.9%) had occurred in the olaparib arm and 49 events (58.3%) in the investigators’ choice of NHA arm. Treatment with olaparib led to a 4.4month median OS advantage compared with investigators’ choice of NHA in the prior taxane subgroup of Cohort A+B, despite patients switching from investigators’ choice of NHA to olaparib treatment upon BICR progression (median OS, 15.8 months vs - 11.4 months; HR; 0.61 [95% CI, 0.43 0.88]; Figure 16). The OS Kaplan–Meier curves for olaparib and investigators’ choice of NHA separated early and remained separated in favour of olaparib for the duration of follow-up period, a remarkable result, despite the level of switching (from investigators’ choice of NHA to olaparib, upon disease progression, as noted above) and the advanced, heavily pre-treated stage of disease.

Figure 16. Kaplan–Meier plot of interim OS in patients whom had prior taxane treatment in Cohort A+B

==> picture [452 x 254] intentionally omitted <==

a Calculated using the Kaplan–Meier method.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

bid, twice daily; CI, confidence interval; HR, hazard ratio; OS, overall survival. Source: PROfound subgroup analyses, prior taxane Cohort A+B[94]

Treatment switching from investigators’ choice of NHA to olaparib post-disease progression confounds the OS analysis, as described in Section B.2.6.3.1, and underestimates the true OS benefit for olaparib versus investigators’ choice of NHA. Since olaparib is not currently approved or reimbursed in England for mCRPC patients whose disease has progressed after NHA, treatment switching analyses were performed in line with NICE DSU TSD16 guidance,[92] to estimate the true OS benefit of olaparib more aligned to a setting where subsequent PARP-inhibitor treatment would not be available to patients whose disease had progressed after NHA.

The treatment switching analyses were conducted on Cohort A+B first (as described previously in Section B.2.6.3.1) and then, for the purposes of the anchored ITC (discussed in Section B.2.9), the counterfactual data were subset by prior taxane use, to reflect the population of patients who would be likely to receive treatment with olaparib in real-world practice and the population of patients included in the CARD study, the main clinical trial for cabazitaxel in the post-NHA setting.

Since prior taxane use was a stratification factor in the PROfound study, the randomisation assumption of the RPSFTM method was considered to hold in this subgroup (see Table 5 for an overview of baseline characteristics across the study populations). The common treatment effect assumption was not considered to be impacted by subsetting to a stratified subgroup. The XXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXX adjusted overall survival in the investigator’s choice of NHA arm to XX months, and produced a hazard ratio XXXXXXXXXXXXXXXXfor olaparib ’ versus treatment-switch adjusted investigators choice of NHA (Figure 17), thus - demonstrating a substantial and clinically meaningful survival benefit in favour of olaparib in patients with HRRm who have received a prior taxane and NHA . These data were used to inform the anchored indirect treatment comparison of olaparib versus cabazitaxel, and are described in further detail in Section B.2.9.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 17. Kaplan–Meier plot of counterfactual overall survival in patients who had prior taxane treatment in Cohort A+B (RPSFTM Weibull method, no recensoring).

==> picture [303 x 158] intentionally omitted <==

==> picture [303 x 158] intentionally omitted <==

Treatment-switching adjusted KM- curves for Cohort A are available upon request. RPSFTM, Rank Preserving Structural Failure Time Model

==> picture [39 x 11] intentionally omitted <==

OS, DCO2 (20[th] March 2020)

The results of the final OS analysis in the prior taxane subgroup of Cohort A+B was consistent with those reported above (for the interim analysis, DCO1) and confirmed a substantial and clinically-meaningful median OS benefit for olaparib versus investigators’ choice of NHA (median OS= XXXmonths and XXXmonths, respectively; XXXXXXXXXXXXXXXXXXXX,

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 12). OS data were XXX mature at the time of this analysis. Treatment switching adjustment analyses on these data are currently underway and will be provided to NICE as soon as they become available.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 12. OS data in the prior taxane subgroup of PROfound at DCO2 (Cohort A+B)

B.2.8 Meta-analysis

As described previously (in Section B.2.2.), the SLR identified three studies that assessed the efficacy and safety of olaparib in the population of interest for this appraisal (PROfound and TOPARP; three abstracts, two full text publications).[28,29,64,65,85]

The TOPARP studies were single-arm Phase II trials; TOPARP-A used the 400 mg dose of olaparib, and TOPARB-B included both 300 mg and 400 mg doses of olaparib. Neither study was explicitly set in a post-NHA population, although the majority of patients had received a prior NHA. Given the availability of data from the much larger international, Phase III, PROfound randomised-controlled trial in the population/treatment setting of interest for this appraisal, data from the TOPARP studies were not included in the evidence synthesis.

B.2.9 Indirect and mixed treatment comparisons

As discussed in Section B.2.1, an SLR was conducted in January 2020 in order to identify published clinical evidence on the use of health technologies in patients with mCRPC whose disease had progressed following treatment with an NHA, irrespective of HRR mutation status. The SLR 14 studies, reported across 23 publications, which reported on outcomes with olaparib,[64,65] cabazitaxel,[42,66-73] docetaxel[74-81] and radium-223[82-84] – the intervention and comparators specified in the final NICE scope - in the post-NHA setting. As described in Section B.2.1.3, the SLR did not identified any studies that reported outcomes on docetaxel or radium223 dichloride in the population relevant to the decision problem, i.e. patients with mCRPC whose disease had progressed after treatment with a NHA.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

The SLR identified eight publications that reported outcomes on cabazitaxel in the post-NHA setting. Of these, only one study - CARD (NCT02485691) – included a cabazitaxel arm as well as an NHA arm (as in PROfound), allowing for a comparative analysis between olaparib and cabazitaxel via an anchored indirect treatment comparison.

As described in Section B.2.3 and Appendix D (Section D.1.4.), CARD is an ongoing Phase IV RCT that assessed the efficacy and safety of cabazitaxel compared with an NHA (enzalutamide or abiraterone plus prednisolone) in patients with mCRPC, who had received previous treatment with docetaxel and an NHA.[66] As all patients enrolled in the CARD trial were required to have received previous docetaxel, the patient population is closely aligned with the prior-taxane subpopulation of the PROfound study, although there were some differences in distributions of prior treatments received and timing of disease progression on NHA in the two studies (see 0 and Appendix D for detailed baseline characteristics of PROfound and CARD). The CARD study was also not restricted to those patients who have mutations in HRR genes, which are associated with more aggressive disease and worse outcomes in mCRPC patients, as discussed in Section B.1.1.[30] The primary endpoint in CARD was rPFS (same as PROfound, although not assessed by BICR); OS was a secondary endpoint in both studies.[67] Collectively, this makes a comparison on outcomes relevant for an economic evaluation possible.

The remaining six publications identified in the SLR that reported outcomes in patients who received cabazitaxel were small single-arm studies (often conducted in a single country or centre; for example, Saad et al. 2014,[71] Saad et al. 2016,[70] Massard et al. 2017,[69] Louhanepessey et al. 2018,[68] and Shiota et al . 2020[12] ) or cabazitaxel combination studies (with and without budesinone; van Soest et al. 2015[73] ). In the absence of a common comparator with PROfound, only unanchored comparisons are feasible between these studies and the prior-taxane group of the PROfound trial. As indicated by the NICE DSU, unanchored comparisons should only be considered in the absence of anchored comparisons.[95] Since data from CARD provided the necessary evidence base for an anchored comparison; these studies were not considered relevant for evidence synthesis per NICE guidance.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Furthermore:

• Louhanepessey et al. 2018[68] and Massard et al. 2017[69] were abstract-only publications that only reported aggregate data with no associated Kaplan– Meier plots; no associated full text publications were identified.[68,69] Aggregate data are unsuitable for an ITC, particularly if no published Kaplan–Meier data are available.[68,69]

• Saad et al. 2016[70] and Saad et al. 2014[71] assessed outcomes from the Canadian cabazitaxel early access programme (NCT01254279), but did not report OS or rPFS, therefore, precluding an comparative analysis of these key endpoints.

In light of these factors, the CARD study was considered the most relevant source of evidence for cabazitaxel in the post-NHA setting and was used to inform the anchored ITC versus olaparib described below.

==> picture [38 x 11] intentionally omitted <==

Indirect treatment comparison (ITC) approach

The PROfound and CARD studies share the same common comparator arm of investigator’s choice of NHA, enabling the use of ITC methods to evaluate the relative efficacy of olaparib versus cabazitaxel.

Patient-level data were available for the pre-specified prior taxane subgroup of PROfound (Cohort A+B), which is similar to the population in the CARD study (i.e. post-taxane, post-NHA); while aggregate data were reported for the CARD study.

Given that there are some identified differences in the trial populations for PROfound and CARD (as noted above), the appropriateness for conducting an anchored PAIC was explored,[96] in line with the recommended process for selecting ITC approaches outlined in Appendix A of NICE DSU TSD 18.[95] This investigation was considered helpful in confirming whether any covariates available for matching in the PROfound and CARD studies could be plausibly considered an effect modifier (via statistical tests) and, if so, whether these might be imbalanced. This assessment determined

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

whether adjusting for differences could lead to more reliable estimates of relative efficacy between olaparib and cabazitaxel, as described below.

As noted previously, enrolment in the CARD study was not restricted by HRRm status (unlike PROfound) and as such, the HRRm status of patients in the study is not known. There is evidence to suggest that HRR mutations may be associated with worse outcomes on cabazitaxel treatment (relative to outcomes in those who do not carry these mutations). Although further research is needed to confirm this, it is thus possible that HRRm status is an effect modifier for cabazitaxel. If this is the case, then the inability to match PROfound and CARD populations by HRR mutations may bias the analysis in favour of cabazitaxel.

==> picture [39 x 10] intentionally omitted <==

Indirect comparison methodology

The ITC was conducted in accordance with the NICE DSU TSD 18 guidance,[96] as detailed below.

B.2.9.2.1 Evidence base

The ITC was conducted on the prior taxane subgroup of the PROfound overall HRRm population (Cohort A+B). As described previously, this subgroup better matches the prior taxane-exposed population in the CARD study, and ensures alignment to the UK population of patients who currently receive cabazitaxel (i.e. after prior treatment with taxane and NHA) and who would be eligible to receive olaparib treatment, if it were to be recommended.

To inform the economic evaluation, the outcomes of interest for this analysis were OS and rPFS. OS was defined as time from randomisation to death due to any cause in both studies. The definitions of rPFS for the PROfound and CARD trials were as follows:

• PROfound : Time from randomisation until objective radiological disease progression (by RECIST 1.1 or prostate cancer working group 3 criteria or death)[97]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• CARD : Time from randomisation until objective tumour progression (RECIST 1.1 criteria), progression of bone lesions (according to prostate cancer working group 2 criteria) or death.[98]

Individual patient-level data (IPD) for rPFS were taken directly from the PROfound Cohort A+B prior taxane subgroup, whilst IPD for OS were derived from the RFPSTM treatment-switching analysis described in Section B.2.7.2. Aggregate data from the CARD study were sourced directly from the study publication.[67]

Following NICE DSU TSD 18 guidance, an anchored ITC was performed for the comparison of PROfound with CARD, since both studies include NHA as the comparator arm.[95] The specific evidence network for this analysis is shown in Figure 18.

Figure 18. Evidence network for the ITC (OS and rPFS)

==> picture [285 x 165] intentionally omitted <==

B.2.9.2.2 Statistical methods

Following guidance from the NICE DSU TSD 18, the analysis was conducted in the following steps:

1. As described above, the prior taxane subgroup was derived from the Cohort A+B FAS of the PROfound study and used in the subsequent stages of the analysis:

   • a. rPFS data for this subpopulation were used directly.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• b. The counterfactual OS data were used after adjusting for treatment switching in the investigator’s choice of NHA arm to olaparib. The counterfactual data was generated using the RPSFTM approach described in Section B.2.7.2. This adjustment was considered appropriate due to the high proportion of patients randomised to the investigators’ choice of NHA arm of PROfound who switched to olaparib treatment after BICR-confirmed disease progression, as described previously.

2. Next, an assessment was conducted in the Cohort A+B prior taxane group to ascertain the extent to which the covariates available for matching were only prognostic, versus being potential effect modifiers for rPFS or OS, and to guide the choice of ITC methodology (i.e. unadjusted Bucher ITC or PAIC). This was conducted using multivariable cox regression with an interaction term between the randomised group and baseline variable to test for evidence of a statistically significant modification effect, with significance levels conservatively set at 80% so as not to exclude covariates that may be clinically important. The results of this analysis are described in Section � and determined the next steps for the ITC.

All analyses were conducted using R[®] version 3.6.1.[99] Kaplan–Meier data in the CARD study were digitised using the methods of Guyot et al . 2012.[100]

B.2.9.2.3 Identification of effect modifiers

The full list of covariates published in the de Wit et al , 2019[67] publication (CARD) were considered for matching. The variables available in PROfound were assessed against this list to determine the comparability of the data and therefore the feasibility of matching. Factors available for matching were assessed for effect modification, as summarised in Table 13.

In accordance with NICE DSU TSD 18 guidance on anchored PAICs, only effect modifiers should be considered for adjustment; therefore, it was necessary to exclude any factors that were deemed to be prognostic only. The list of covariates (summarised in Table 13) was assessed by an AstraZeneca medical oncologist with

experience in prostate cancer, who identified four factors that are prognostic factors Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 85 of 208

only (i.e. not effect modifiers), and thus should be excluded from consideration for the matching process, per NICE DSU guidance. This included neutrophil count per mm[3] , haemoglobin (g/L), alkaline phosphatase (IU/L), lactate dehydrogenase (IU/L), previous NHA (abiraterone). the remaining list of covariates (age, ECOG score [0-2], presence of lung or liver metastases*, mean baseline PSA level [ng/ml], M1 disease at diagnosis and Gleason score [8-10]) were then tested for evidence of effect modification at the 80% significance level, focusing on OS.

• Effect modifiers for OS were tested in the analysis since it is the main endpoint routinely used to demonstrate superiority of antineoplastic therapies and the most-important driver of cost-effectiveness (over a lifetime horizon) in the advanced mCRPC setting. It would be expected that the findings for OS would also be applicable for rPFS.

• The significance level was set to 80% (rather than the conventional 95% level), to maximise chances of identifying any variables that could be potentially effect modifying.

Table 13. Summary of the covariates available for testing for effect modification

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• Further details provided below.

Clarification around two variables (previous abiraterone treatment and visceral disease) noted in the Table above are as follows:

• Previous abiraterone treatment could not be tested appropriately as an effect modifier, due to differences in reporting across the PROfound and CARD trials. In the CARD study, data are only reported for the previous NHA treatment and separately for abiraterone and enzalutamide (it was not clear if one or more patients had received both NHAs). In contrast, in the PROfound study data were collected and reported for ≥1 prior treatment with abiraterone, enzalutamide, or abiraterone and enzalutamide.

• The reporting of visceral disease in both studies differed but was included for purposes of matching. In CARD, patients with visceral disease patients were categorised as follows: any patient with liver metastases was categorised as having liver metastases even if they had other metastatic sites; patients with lung metastases were denoted as having lung metastases, unless they also had liver metastases; and all other patients with visceral disease were categorized as having non-hepatic, non-pulmonary visceral metastases (such as adrenal, kidney, and others). In PROfound, visceral disease was not reported separately by liver, lung and other visceral metastases. Therefore, it was assumed all patients with visceral metastases in the CARD study had liver and/or lung metastases, and that it was appropriate to match with all visceral disease in the PROfound study. This assumption was considered reasonable by the AstraZeneca medical oncologist.

The results of the effect modifiers assessment for the treatment switching-adjusted OS endpoint are presented in Table 14 and show that:

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• There are no significant effect modifiers for OS at the 80% significance threshold. Therefore, a Bucher ITC,[101] unadjusted for variables , is the most appropriate and reliable for estimating the relative efficacy of olaparib against cabazitaxel.

• Age and PSA could be considered a prognostic factor for OS based on statistical significance, but not an effect modifier as the interaction term was not significant at the 80% level.

In the absence of evidence to support effect modification, a PAIC is not expected to lead to a reduction in bias, and may only serve to introduce uncertainty via “overmatching” in the estimates of relative efficacy (as noted in the NICE DSU guidance).

Table 14. Assessment of effect modifiers for switching adjusted anchored analysis.

==> picture [452 x 350] intentionally omitted <==

----- Start of picture text -----
Covariate Factor Interaction
(OS switching-adjusted) (OS switching-adjusted)
** = statistically significant, ** = statistically significant,
may be interpreted as may be interpreted as effect
prognostic factor modifier
Age XXXX XXX
XXXXXXXXX.
XXXXXXXXXX.
Visceral disease XXXX XXX.
XXXXXXXX XXXXXXXXX
M1 disease at diagnosis XXX.. XXX..
XXXXXXXXX XXXXXXXXX
Gleason 8-10 XXXX XXX..
XXXXXXXXX. XXXXXXXXX
ECOG 0-1 XXX. XXX
XXXXXXXXX. XXXXXXXXX
PSA* XXX.. XXX..
XXXXXXXXXX. XXXXXXXXX.
----- End of picture text -----

• *Binary covariate was used for modelling

• **Covariate was significant at 80% level

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

==> picture [39 x 10] intentionally omitted <==

Indirect comparison results

As outlined above, an unadjusted ITC is the most appropriate and reliable method for estimating the relative efficacy of olaparib against cabazitaxel, in the absence of any confirmed effect modifiers. The results of the unadjusted ITC show that olaparib is associated with rPFS and OS benefit versus cabazitaxel (XXXXXXXXXXXXXXXX. XXX]) and OS (XXXXXXXXXXXXXXXXXXX]), as described below.

B.2.9.3.1 rPFS

The proportional hazards assumption in the PROfound and CARD studies was assessed by visual inspection of the log-cumulative hazards plots and the Schoenfeld plots, and conducting Schoenfeld individual tests. Visual inspection of the log-cumulative hazards plots for rPFS indicates that proportional hazards assumption holds in both studies. This is further confirmed by the Schoenfeld individual tests, which resulted in p-values of 0.74 and 0.75 in the PROfound and CARD studies, respectively, indicating that there was no evidence against the null hypothesis of proportional hazards at the 95% significance level. Log-cumulative hazard plots and Schoenfeld plots for PROfound and CARD rPFS data are presented in Figure 19 and Figure 20, respectively.

Figure 19: PROfound rPFS: Schoenfeld (left) and log-cumulative hazards (right) plots

==> picture [447 x 69] intentionally omitted <==

==> picture [447 x 69] intentionally omitted <==

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Figure 20: CARD rPFS: Schoenfeld (left) and log-cumulative hazards (right) plots

==> picture [200 x 137] intentionally omitted <==

==> picture [215 x 136] intentionally omitted <==

Since the proportional hazards assumption was determined to hold across both the PROfound and CARD trials, the Bucher et al. method[101] was considered appropriate for conducting the ITC.

The unadjusted ITC was conducted by calculating the hazard ratios from the PROfound prior-taxane IPD and the recreated IPD from the digitised Kaplan Meier data in CARD. In the ITC analysis for rPFS:

• The hazard ratio for olaparib versus investigator’s choice of NHA in the PROfound Cohort A+B prior taxane subgroup was XXXXXXXXXXXXXX XXXX

• The hazard ratio for cabazitaxel versus investigator’s choice of NHA, generated from the recreated IPD data from the CARD study, was XXXXXXX XXXXXXXXXX.

• The hazard ratio for olaparib versus cabazitaxel was XXXXXXXXXXXXXX. XXX. for rPFS .

B.2.9.3.2 OS

For OS, the proportional hazards assumption was assessed using the same approach as for rPFS. The log-cumulative hazards and Schoenfeld plots for the PROfound and CARD studies are presented in Figure 21 and Figure 22, respectively.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Visual inspection of the log-cumulative hazards plots for OS indicates that proportional hazards assumption holds in both studies. This is further confirmed by the Schoenfeld individual tests, which resulted in p-values of 0.27 and 0.94 in the PROfound and CARD studies, respectively, indicating that there was no evidence against the null hypothesis of proportional hazards at the 95% significance level. Therefore, the evidence of proportional hazards across both studies and rPFS/OS endpoints supports the use of constant hazard ratios to generate comparative evidence for olaparib and cabazitaxel.

Figure 21. PROfound OS: Schoenfeld (left) and log-cumulative hazards (right) plots

==> picture [214 x 136] intentionally omitted <==

==> picture [235 x 134] intentionally omitted <==

Figure 22. CARD OS: Schoenfeld (left) and log-cumulative hazards (right) plots

==> picture [219 x 154] intentionally omitted <==

==> picture [228 x 153] intentionally omitted <==

In the ITC analysis for OS:

• The hazard ratio for olaparib compared with investigator’s choice of NHA in

the PROfound Cohort A+B prior taxane subgroup was XXXXXXXXXXXXXX. XXX..

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• The hazard ratio for cabazitaxel versus investigator’s choice of NHA, generated from the recreated IPD data from the CARD study, XXXXXXXXXX XXXXXXXXXX

• The hazard ratio for olaparib versus cabazitaxel was XXXXXXXXXXXXXX. XXX..

B.2.9.3.3 Summary of results

The results of the unadjusted Bucher ITC are summarised in Table 15 and were used to inform the economic evaluation described in Section B.3. The use of constant hazard ratios was considered appropriate since the proportional hazards assumption was found to hold across both studies and no effect modifiers were identified. The results show that, in patients who have received prior taxane and progressed on NHA, olaparib results in a XXXXXXXXXXXXX. in disease

progression, translating to a XXXXXXXXXX in mortality compared with cabazitaxel .

Table 15. Summary of results for rPFS and OS: Bucher anchored ITC olaparib versus cabazitaxel

CI, confidence interval; HR, hazard ratio; PFS, progression-free survival; OS, overall survival.

==> picture [39 x 11] intentionally omitted <==

Strengths and limitations

The anchored Bucher ITC described in this section is the optimal methodology for comparison using the available data, i.e. patient-level data from the PROfound study and aggregate data from the primary publication of the CARD study. Since no treatment effect modifiers were identified in the analysis, an unadjusted ITC

approach was deemed preferable to a PAIC and in accordance with NICE DSU guidance.

For OS, the unadjusted ITC approach represents the most parsimonious approach and closest to the pre-specified analyses from the original study and CSR, with no

loss of precision/sample size; it would be expected that the same principles would Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

hold for rPFS. Additionally, for OS the unadjusted ITC approach accounts for treatment switching without explicit loss of precision/sample size. The ITC utilised the switching-adjusted (counterfactual) investigators’ choice of NHA OS data from the prior taxane group of Cohort A+B of the PROfound study, which adjusted for the subsequent use of olaparib in the investigators’ choice of NHA arm. These data were deemed necessary to use as there was a clear confounding impact of subsequent use of olaparib on OS in the NHA arm of PROfound. Additionally, olaparib is not licensed or reimbursed in this treatment setting in the UK; therefore, adjusting for treatment switching (to olaparib) is a necessary step to better reflect outcomes on clinical practice (unlike cabazitaxel, which is a reimbursed treatment). Without adjusting for treatment-switching, the ITC analysis would not be informative for decision making purposes.

B.2.10 Adverse reactions

All patients who were randomised to the PROfound study and received at least one dose of randomised study treatment in Cohort A or B were included in the safety analysis set (SAS) in their respective cohorts.[61] Safety data captured on patients receiving investigators’ choice of NHA who subsequently switched to olaparib upon disease progression were summarised as per the treatment at the time of the onset of safety condition or lab result and reported in the safety switch analysis set (see Section 12, pages 237−280 of the PROfound Clinical Study Report, version 1, 23 October 2019).[61]

Overall, the safety and tolerability profile of olaparib in PROfound was consistent with the known safety and tolerability profile of olaparib and considered to be acceptable in this patient population. The most common AEs (reported by ≥20% of patients) in the olaparib arm (anaemia, nausea, decreased appetite, fatigue and diarrhoea) were known adverse drug reactions associated with olaparib and could generally be managed through dose modifications. No new safety signals were identified.

A summary of treatment exposure and adverse events reported in PROfound study is provided in the following sections; further details are available in the PROfound CSR (Section 12).[61] Collectively these data, in conjunction with the efficacy analysis

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

presented in Section B.2.6, support a positive benefit-risk profile for olaparib in patients with mCRPC who have failed prior treatment with an NHA and have HRR gene mutations. The safety profile in the subgroup of patients who had received treatment with a prior taxane was consistent with analyses in the Cohort A+B safety analysis set. This is as expected, since prior treatments received are not expected to impact upon a patient’s tolerability of study treatments.

A limitation of the data summarised below is that they do not provide comparative evidence versus cabazitaxel, the standard-of-care in England in this treatment setting. Since the mechanisms of action of olaparib and cabazitaxel are different, such analyses may be unreliable and inappropriate for use in decision-making. The most appropriate methodology for conducting an ITC (unadjusted comparison versus PAIC) is also difficult to determine since standard matching variables (such as patients characteristics) may not be relevant to the occurrence of AEs and other unidentified potentially effect modifying factors may exist. Although a formal comparator safety analysis of olaparib versus cabazitaxel was not conducted due to these factors, 6 UK clinical experts consulted to inform the company submission highlighted that, in their experience, olaparib has a manageable tolerability profile; in a minority of experts who compared their experience of olaparib tolerability with cabazitaxel (2 clinical experts), both confirmed that olaparib had a more manageable tolerability profile in real-world clinical practice.[102]

==> picture [44 x 10] intentionally omitted <==

Exposure to treatment, DCO1 (4[th] June 2019)

Exposure data reported in this section relates to length of time on treatment with the study drug (olaparib or investigators choice of NHA). In Cohort A+B, the median total duration of exposure to olaparib was ~1.9 times longer than in the investigators’ choice of NHA arm (7.5 months versus 3.9 months), consistent with the delayed time to radiological disease progression in the olaparib arm. A total of 20.3% of patients remained on treatment in the olaparib arm at 12 months.[61]

Any deviation from the planned bid dosing was captured as a dose reduction or interruption and a reason assigned. Dose reductions/interruptions included missed or forgotten doses, or modifications in response to an AE. In both arms of Cohort A+B, AEs were the most-common reason for dose interruption (occurring in 90 patients

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved Page 94 of 208

[35.2%] in the olaparib arm versus 11 [8.5%] patients in the investigators’ choice of NHA arm). Median relative dose intensity and percentage intended dose were >98% in both treatment arms, indicating that dose intensity was not affected by dose modifications.

Dose interruptions, reductions, or modifications were not separately analysed for the prior taxane subgroup of Cohort A+B; the total and actual treatment duration in the subgroup was similar to the overall Cohort A+B population. These data are summarised in Table 16.

Table 16. Summary of treatment exposure, dose interruptions, and dose modifications: Cohort A + B SAS and prior taxane subgroup, DCO1 (4[th] June 2019)

aTotal treatment duration = (last dose date – first dose date +1). Median days

bActual treatment duration = (last dose date – first dose date +1) excluding dose interruptions. Median days

Patient E7602055 (investigators choice of NHA) had treatment exposure 42 days longer than reported as discontinuation date was misreported. Due to this error, the dose durations are incorrectly derived. As this error was reported for only 1 patient, this would have had a very small impact on the calculations for dose durations; therefore, the reported dose durations are considered largely representative of their true values.

AE, adverse event; bid, twice daily; NHA new hormonal agent; SAS safety analysis set.[61]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

==> picture [46 x 10] intentionally omitted <==

Adverse events, DCO1 (4[th] June 2019)

The patients experiencing AEs in any category in Cohort A+B SAS and the prior taxane subgroup are summarised in Table 17. The majority of patients experienced 1 or more AEs during the course of the study. The incidence of AEs was similar in both treatment arms and across the full SAS and the prior taxane subgroup. The most common AEs in the olaparib arm (reported by ≥20% of patients) were anaemia, nausea, decreased appetite, fatigue and diarrhoea, which are known adverse drug reactions (ADRs) associated with olaparib treatment (data not shown; see Table 74 [Section 12.2.2.] of the CSR for details). These AEs were generally managed using olaparib dose modification; most AEs did not lead to treatment discontinuation (Table 17).

Further information on AEs leading to dose interruptions, reductions, or discontinuation of study treatment, AE of CTCAE Grade 3 or above, SAEs, and fatal AEs in the Cohort A+B SAS are provided in Section 12.2 of the PROfound CSR. Briefly:

• AEs leading to dose interruption occurred in 44.9% of olaparib-treated patients; the most common AEs leading to dose interruption in the olaparib arm (reported in ≥5% of patients) were anaemia (25.0%) and thrombocytopenia (5.5%). The majority of AEs of anaemia or thrombocytopenia were managed with dose reductions or temporary dose interruptions. Reported events of thrombocytopenia rarely led to permanent discontinuation of study treatment in the olaparib arm (2% of patients); AE of anaemia led to discontinuation of study treatment in 7% of patients in the olaparib arm.

• AEs of CTCAE Grade ≥ 3 were reported in 50.8% of olaparib-treated patients. Anaemia was the only AE of CTCAE Grade ≥ 3 reported in ≥ 5% of patients. Anaemia was also the most-common SAE in the olaparib arm, reported in 22 (8.6%) patients.

• Just 7 patients in the olaparib arm had Grade 4 AEs (lung infection and septic shock [both in 1 patient], pulmonary embolism [1 patient], respiratory failure and sepsis [both in 1 patient] and thrombocytopenia [4 patients]); two of the patients also had Grade 5 AEs.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• There were 15 fatal AEs (10 patients [3.9%] in the olaparib arm and 5 patients [3.8%] in the investigators’ choice of NHA arm) during study treatment or 30-day follow-up. Two patients (1 patient in the olaparib arm [lung infection and neutropenia] and 1 in the investigators choice of NHA arm [pleural effusion]) had AEs with an outcome of death that were considered by the investigator to be causally related to study treatment.

The safety profiles of the prior taxane subgroup was comparable with the Cohort A+B SAS, with a similar proportion of Grade 3 and above AEs and SAEs in the two populations.

The frequency of AEs in the safety switch analysis set was similar to those randomised to receive olaparib (described above); these data are described in detail is Section 12.6 of the PROfound CSR.

Table 17. Adverse events in any category, DCO1 (4[th] June 2019) in Cohort A+B SAS/prior taxane subgroup.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Any AE relating to interruptions

115 (44.9) 24 (18.5) XXXXXX XXXXXX

a Patients with multiple events in the same category are counted only once in that category. Patients with events in more than one category were counted once in each of those categories. b As assessed by the investigator.

Includes AEs with an onset date on or after the date of first dose and up to and including 30 days following discontinuation of randomised treatment or the day before switching to olaparib. AE adverse event; bid twice daily; CTCAE Common Terminology Criteria for Adverse Events v4.03; DCO data cut-off; MedDRA Medical Dictionary for Regulatory Activities; NHA new hormonal agent; SAE serious adverse event; SAS safety analysis set.[61]

B.2.11 Ongoing studies

There are no ongoing studies relevant to the decision problem for this appraisal. The data cut-off for the final OS analysis of the PROfound study (DCO2) was on 20th March 2020. Top-line OS results from this data-cut are included in Sections B.2.6.3.1 and B.2.7.3;[31] further analyses including a treatment switch-adjusted analysis of OS and an anchored ITC versus cabazitaxel are currently underway and will be provided to NICE as soon as possible (as agreed during the decision-problem meeting on 17th March 2020).

B.2.12 Innovation

Olaparib is the only targeted therapy to have demonstrated a clinically-meaningful improvement in both rPFS and OS in a Phase III RCT versus investigators’ choice of NHA, in mCRPC patients with qualifying HRR gene mutations, whose disease has progressed after treatment with an NHA.

Key data from the interim analysis of the PROfound study (which forms the basis of the company submission) were presented at the Presidential Symposium of the 2019 European Society of Medical Oncology (ESMO) Annual Congress (Barcelona, Spain).[64] Professor Maha Hussain, who presented these data referred to the “ significant effect [of olaparib] on disease progression and other clinically relevant effects such as pain progression and objective response rate ” as being a “ remarkable achievement in such heavily pre-treated patients with prostate cancer ”. She added that “ prostate cancer has lagged behind all other common solid tumours in the use of molecularly targeted treatment ” and that it is “ very exciting that now we

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

[clinicians] can personalise an individual’s treatment based on specific genomic alterations in their cancer cells .”

Results of the anchored Bucher ITC described in Section B.2.9 show that treatment with olaparib is associated with a XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.. versus cabazitaxel, the current standard-of-care for patients with mCRPC who have received a prior taxane (for HSPC), and whose disease has progressed after treatment with an NHA (XXXXXXXXXXXXXX; Section B.2.9.3.1). Treatment with olaparib was also associated XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.. versus cabazitaxel (XXXXXXXXXXXXXXXXXXX;Section �).

22.3% of patients in the olaparib arm of Cohort A+B were still on study treatment at the time of DCO1 (4[th] June 2019), providing hope for a sustained long-term response to treatment for at least a subset of patients, thus addressing a key unmet need in this setting.

In addition to extending survival, treatment with olaparib is associated with relevant and meaningful patient benefits, such as delayed time to pain progression, a significant cause of morbidity in patients with mCRPC (Section B.1). Olaparib also represents an alternative to cytotoxic chemotherapy with cabazitaxel, an important consideration since many men are unable to access chemotherapy and/or unwilling to undergo further chemotherapy at this stage of their lives.[103] Olaparib is also an oral treatment that can be taken at home, thus negating the need for patients to travel to hospitals for their infusion (as with cabazitaxel) and freeing capacity/resources for the NHS. These important benefits of olaparib represent further value to patients and the healthcare system beyond that which is captured in the QALY.

Finally, it is worth mentioning that a significant part of olaparib’s clinical development programme was based in the UK and championed by UK scientists and clinicians. Phase II studies of olaparib in patients with mCRPC (TOPARP-A and TOPARP-B) were sponsored by The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust and funded by Cancer Research UK and AstraZeneca.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Professor Paul Workman, Chief Executive at The Institute of Cancer Research, said the following regarding results from the TOPARP-B study: “ Precision medicines targeted to specific genetic faults are transforming treatment for many different cancers, and with this new research it looks like we will soon be able to add prostate cancer to that list. It’s exciting to see a drug which the ICR helped pioneer having such widespread benefits for both women and men with cancer ”.

Olaparib was granted Breakthrough Therapy Designation by the US Food and Drug Administration (FDA) in January 2019 for the treatment of BRCA 1/2- or ATM genemutated mCRPC in patients who have received a prior taxane-based chemotherapy and at least one NHA (abiraterone or enzalutamide), based on the positive results of the TOPARP-A Phase II trial, which informed the PROfound trial. It was approved by the US FDA on the 19[th] of May 2020 as a treatment option “ for adult patients with deleterious or suspected deleterious germline or somatic HRR gene-mutated mCRPC, who have progressed following prior treatment with enzalutamide or abiraterone ” after being granted a priority review in January 2020.[104,105]

B.2.13 Interpretation of clinical effectiveness and safety evidence

==> picture [44 x 11] intentionally omitted <==

Principle findings from the evidence base

B.2.13.1.1 Efficacy and HRQoL

The PROfound study met its primary endpoint in the interim analysis (DCO1, 4th June 2019), demonstrating a statistically-significant and clinically-meaningful improvement in rPFS (by BICR) for olaparib versus investigators’ choice of NHA in patients with mCRPC and qualifying mutations in BRCA1 , BRCA2 , or ATM genes (Cohort A), who have who have failed prior treatment with an NHA (HR, 0.34, 95% CI, 0.25–0.47; p < 0.0001).

The study also met its secondary endpoint of rPFS (by BICR) in the overall population of patients with qualifying mutations in any of the 15 prespecified HRR genes (Cohort A+B), demonstrating a remarkable 51% reduction in the risk of radiological disease progression or death versus investigators’ choice of NHA (HR, 0.49, 95% CI, 0.38–0.63; p < 0.0001). An rPFS benefit in favour of olaparib (versus investigators’ choice of NHA) was observed across the pre-specified subgroup

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

analyses, which were conducted to assess the consistency of treatment effect across potential/expected prognostic baseline characteristics (such a previous taxane use or metastases at baseline [bone, visceral, other]), with olaparib treatment reducing the risk of radiological disease progression or death by 23% to 88% (Appendix E).

In the prior taxane subgroup, which is reflective of the real-world population of patients anticipated to receive olaparib in UK clinical practice,[c ] olaparib reduced the risk of radiological disease progression or death by 61% compared with the investigators’ choice of NHA (median rPFS, 5.8 months vs 2.6 months; Section B.2.7). Although the PROfound study was not powered to assess the efficacy of olaparib versus investigators’ choice of NHA in this population, prior taxane use was a stratification factor in the study, ensuring balanced distribution of patients between olaparib and investigators’ choice of NHA arms, and maintaining the robustness of this analysis.

rPFS data in the overall study population (Cohort A+B) and the prior taxane subgroup are also supported by the OS analysis, which showed a clinically meaningful reduction in the risk of death (HR, 0.67 [95% CI, 0.49–0.93) and HR, 0.61 [95% CI, 0.43-0.88], respectively), despite the majority of eligible patients (84.6% in Cohort A+B) crossing from investigators’ choice of NHA to olaparib upon BICR-assessed rPFS progression (interim OS analysis; DCO1). The use of olaparib after investigators’ choice of NHA confounds the OS analysis, biasing the results in favour of the comparator arm and underestimating the true OS benefit of olaparib treatment. A treatment-switching analyses using RPSFTMs were thus conducted in line with NICE DSU TSD 16 to adjust for treatment switching in Cohort A+B as well as the prior taxane subgroup. These analyses demonstrated an unprecedented OS benefit for olaparib versus investigators’ choice of NHA in both populations (HR, XX.. XXXXXXXXXXXXX and XXXXXXXXXXXXXXXX], respectively).

The OS benefit was maintained in the final OS analysis (DCO2), with olaparib reducing the overall risk of death versus investigators’ choice of NHA by XXXX

c Since the majority (~75%) of patients receive docetaxel (a taxane) in combination with ADT for HSPC, prior to receiving an NHA for mCRPC. Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

XXXXXX in the overall Cohort A+B population and the prior taxane subgroup,

respectively XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX despite

most patients in the investigators’ choice of NHA arm switching to olaparib treatment after BICR-confirmed disease progression. Further analyses of these data are currently underway.

In the anchored Bucher ITC described in Section B.2.9, treatment with olaparib XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. versus

cabazitaxel, the current standard-of-care for patients with mCRPC who have received a prior taxane (for HSPC), and whose disease has progressed after treatment with a NHA (XXXXXXXXXXXXXXXXXX, Section B.2.9.3.1). Treatment with olaparib was also associated XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX versus cabazitaxel (XXXXXXXXXXXXXXXXXXX.; Section �), supporting its positioning as a new standard-of-care for patients with mCRPC and qualifying HRR gene mutations, whose disease has progressed after treatment with prior taxane and NHAd.

Importantly, XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. Instead,

treatment with olaparib resulted in meaningful benefits to patients (versus investigators’ choice of NHA) in the form of delayed time to deterioration in FACT-P Total and all subscale scores in the overall study population (Cohort A+B), as well as delayed time to pain progression (HR, 0.64, 95% CI, 0.35–1.21), delayed time to first opiate use XXXXXXXXXXXXXXXXXXXX, and delayed time to first symptomatic

skeletal-related event XXXXXXXXXXXXXXXXXXXX, which are significant causes of morbidity in patients with mCRPC. In the subgroup of patients with bone metastases only, olaparib reduced the risk of radiographic disease progression or death by 43% (see Appendix E for details).

d It is worth reiterating that while the prior taxane subgroup is the focus of this submission (aligned to the anticipated positioning of olaparib for the majority of mCRPC patients, who currently receive a taxane [docetaxel] earlier in the treatment pathway, prior to NHA), treatment with olaparib is also effective in those patients who have not received a prior taxane (rPFS HR = 0.77, 95% CI, 0.50-1.22 Cohort A+B).

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

B.2.13.1.2 Safety and tolerability

The median total duration of exposure to olaparib (7.5 months) was consistent with the median duration of BICR-assessed rPFS in the olaparib arm (Cohort A+B SAS). The high relative dose intensity and percentage intended dose of > 98% indicated that most patients were able to take the full dose of olaparib.

Overall, the safety and tolerability profile of olaparib in PROfound was consistent with the known safety and tolerability profile of olaparib. The most commonlyreported AEs in the olaparib arm (i.e. anaemia, nausea, decreased appetite, fatigue, and diarrhoea) were known ADRs for olaparib and could generally be managed through dose modifications. No new safety signals were identified.

Overall, the safety analyses showed that treatment with olaparib was well tolerated in patients with mCRPC. This is further corroborated by patient reported outcome (PRO) data, which show that treatment with olaparib had no detrimental impact on patients’ HRQoL (relative to investigators’ choice of NHA). Taken in the context of the substantial and sustained efficacy of olaparib in this setting, these data support a favourable risk-benefit profile for the use of olaparib in patients with mCRPC with qualifying HRR gene mutations, who have failed on previous NHA treatment.

==> picture [46 x 10] intentionally omitted <==

Strengths and limitations of the evidence base

Strengths of the evidence base:

PROfound was a large, multicentre, randomised, prospective, Phase III, open-label study that provided comparative evidence for olaparib versus investigators’ choice of NHA, in mCRPC patients with qualifying HRR gene mutations, who have previously failed treatment with NHA.[30]

The PROfound study was approved by the independent Institutional Review Board (IRB)/Independent Ethics Committee (IEC) associated with each study centre. It was performed in accordance with the ethical principles that have their origin in the Declaration of Helsinki and that are consistent with International Council for Harmonisation (ICH)/Good Clinical Practice (GCP), applicable regulatory requirements, and the AstraZeneca policy on Bioethics. Quality of data was assured through monitoring of investigational sites, appropriate training for study personnel,

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

and use of data management procedures. In addition, an independent data monitoring committee was created to assess the safety of the study on a regular basis.

The primary endpoint of rPFS is widely-used in prostate cancer trials, and a clinically-relevant in prostate cancer. Owing to the open-label design of PROfound, rigorous methodology was employed to ensure robustness of the primary endpoint assessment, with the primary analysis of rPFS based on BICR-assessed of all radiological scans. The rPFS data are supported by a range of secondary endpoints, including PFS2 and OS, which consistently show a compelling clinical benefit in favour of olaparib (versus investigators’ choice of NHA), despite the majority of patients switching over to olaparib after disease progression on NHA (see below).

Limitations of the evidence base:

In addition to the considerations involving comparators specified in the final NICE scope (Section B.1.4) and the need for treatment switching analysis (to adjust for patients switching from investigators’ choice of NHA to olaparib confounding the OS analysis; Sections B.2.6.3.1 and B.2.7.2), the maturity of the interim OS analysis is also worth noting in relation to the clinical effectiveness evidence. At the time of the interim OS analysis (DCO1), OS data in the overall HRRm population of PROfound (i.e. Cohort A+B) was 41% mature. OS data maturity in the prior taxane subgroup was 48.0%.

The data-cut for the final OS analysis was on the 20 March 2020. OS data in the overall population (Cohort A+B) was 64% mature at the time of this analysis; further analyses of these data are currently underway and will be provided to NICE as soon as possible.

==> picture [46 x 10] intentionally omitted <==

End of Life considerations

mCRPC represents an area of significant unmet need, with patients typically surviving less than two years from starting active treatment.

• In the COU-AA-301 study of abiraterone (plus prednisone) in patients with mCRPC whose disease had progressed after prior docetaxel treatment, a median OS of just 15.8 months was achieved with abiraterone treatment.[106]

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

• Similarly, in the AFFIRM trial of enzalutamide in mCRPC patients whose disease had progressed after prior docetaxel treatment, median OS of just 18.4 months was reported with enzalutamide treatment.[107]

The addition of cabazitaxel improved outcomes for those patients whose disease had progressed after receiving treatment with taxane and NHA (abiraterone, or enzalutamide), with a median OS of 13.6 months from the initiation of cabazitaxel treatment; however, life expectancy remains suboptimal with a need for new, life extending treatment options.

Data from the PROfound study directly show a clinically-meaningful OS benefit for olaparib versus investigators’ choice of NHA, in mCRPC patients whose disease has progressed after treatment with a taxane and NHA, with a median XXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX..After adjusting for treatment switching, a survival gain of XXXXXXXXXXXXXXXXXXXXXXXXXXXXX was achieved for olaparib versus investigators’ choice of NHA (using the RPSFTM preferred analysis).

In the anchored Bucher ITC versus cabazitaxel - the current standard-of-care in England and the relevant comparator for olaparib in mCRPC patients whose disease has progressed after treatment with taxane and NHA – olaparib was associated with XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

In summary, treatment with olaparib offers a clinically-meaningful survival benefit (in excess of 3 months) versus the current standard-of-care, in a setting where usual life expectancy is less than 24 months or 2 years, thereby meeting the end-of-life criteria specified by NICE. These data are summarised in

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 18 below.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

Table 18. End-of-life criteria

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640]

© AstraZeneca 2020 All rights reserved

B.3 Cost effectiveness

B.3.1 Published cost-effectiveness studies

A total of four abstracts[108-111] were identified through systematic searching that reported on four independent cost-effectiveness analyses that estimated the relative cost of different sequences of taxanes and NHA in patients with mCRPC in Egypt,[108] Japan,[110] Russia[109] and Spain.[111] Full details of the SLR methodology and a summary of the literature identified are given in Appendix G.

Hand searches for previous HTAs and relevant economic assessments were also conducted. Based on a patient population with mCRPC who have experienced progression following an NHA no relevant evidence was identified. Despite this, previous NICE technology appraisals in the mCRPC indication were still considered, where relevant, to inform the approach taken in the cost-effectiveness analysis for olaparib. Based on the final NICE scope the following technology appraisals were considered potentially relevant (summarised in Table 20): TA391 (cabazitaxel),[41] TA412 and TA376 (radium-223 dichloride),[112] TA387[37] and TA259 (abiraterone),[113] TA377[38] and TA316 (enzalutamide),[114] TA101 (docetaxel).[115]

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

B.3.2 Economic analysis

As none of the publications identified by the SLR reported evidence specific to the UK, they were not considered relevant to this submission. Previous NICE technology appraisals in the broader mCRPC indication were still considered, where relevant, to inform the approach taken in the cost-effectiveness analysis for olaparib, even though they were not conducted specifically within the post-NHA setting.

==> picture [38 x 11] intentionally omitted <==

Patient population

As described in section B.1.1, the anticipated EMA license for olaparib in this indication is XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.

In line with the decision problem and data limitations outlined in section 0 and section B.2, the cost-effectiveness analysis focused on the pre-specified prior taxane subgroup of the overall HRRm population (i.e., Cohort A+B) of PROfound. This subgroup aligns with the patient population with mCRPC who would be eligible to receive cabazitaxel in England and Wales, as cabazitaxel is only reimbursed for those patients whose disease has progressed during or after prior docetaxel treatment.[41] The prior taxane subgroup of PROfound included those patients who previously received at least one prior taxane-based treatment, and comprised approximately XX% of the overall trial population. Prior taxane use (yes/no) was a stratification factor in the PROfound study, this ensuring balance between the treatment and comparator arms of the trial.

Table 19. Description of modelled patient population

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

received cabazitaxel only; this is not expected to have any impact on the analysis or the interpretation of the results. Further details are available in Section B.2.3.7.

HRR(m), homologous recombination repair (mutation); NHA, new hormonal agent; mCRPC, metastatic castration-resistant prostate cancer.

==> picture [39 x 10] intentionally omitted <==

Model structure

A de novo economic model was developed in Microsoft Excel[®] to evaluate the cost-

effectiveness of olaparib in the mCRPC setting. The cohort model follows a partitioned survival (or ‘area-under the curve’) approach with three health states (Figure 23):

• Stable disease (progression-free)

• Progressed disease

• Death

The partitioned survival model structure is a widely accepted approach that has been used in previous NICE health technology assessments across many oncology indications. The structure reflects the likely disease history of the patient population and is able to capture the key determinants of health and cost outcomes in a clear and simple manner. The model structure is flexible and is able to adequately quantify the primary objectives of treating patients with mCRPC: extending survival, delaying progression, and improving quality of life. The partitioned survival approach relies on the use of key endpoints (OS, rPFS) reported in clinical trials to estimate health and cost outcomes as described below.

Patients enter the model in the stable disease state and are assumed to be on treatment. At each model cycle, the number of patients in each independent and mutually exclusive health state is updated.

The stable disease state includes patients who are alive and whose disease has not yet progressed. Patients can either remain in the stable disease health state, progress, or die. At any model cycle the proportion of patients who are progressionfree is represented directly from the rPFS curve for each intervention. Treatmentrelated costs, drug acquisition, drug administration and AE costs for each comparator are accrued based on the rPFS curve for each intervention (see Section

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

B.3.3.2 for further details). Monitoring costs associated with patients on treatment are also accrued.

The progressed disease state includes patients who are alive but whose disease has progressed. It is assumed that once patients have progressed they cannot return to the stable disease state; they can only transition into the death state. At any model cycle the proportion of patients with progressed disease is calculated as the difference between OS and rPFS (all patients who are alive who, but not

progression-free). After progression, patients could receive subsequent treatment or best supportive care (BSC). These costs are accrued; however, no additional adjustment on survival is required as any impact of subsequent treatment is implicit in the OS data. Monitoring costs associated with patients who are alive but have discontinued treatment are accrued.

In the model, death is an absorbing state calculated as 1-OS; that is, all patients who are not alive. A one-off cost for end-of-life care is applied to patients who die at each model cycle.

Additionally, in line with standard practice for developing partitioned survival models in oncology, the following constraints are applied in the model to ensure logical patient flow at each cycle:

• The risk of death in the modelled population cannot be lower than the all-cause mortality of the UK general population at each model cycle, determined by published life tables.[116] This ensures that at any given cycle, the mortality risk of the modelled population is equal to or greater than that of the general population (matched on age [age at start of model equivalent to patients’ mean age at baseline in the prior taxane subgroup of PROfound, 63.7 years] and gender).

• rPFS is constrained by OS, such that the number of patients who are progressionfree cannot exceed the total number of patients alive.

Full details regarding modelling PFS, TTD and OS are presented in section B.3.3. Details regarding costs are described in section B.3.5.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Figure 23. Structure of the cost-effectiveness model

==> picture [463 x 122] intentionally omitted <==

==> picture [463 x 122] intentionally omitted <==

a Proportion of patients who have progressed calculated as the residual of OS and rPFS. Health outcomes defined by rPFS and OS; in the base case analysis, cost outcomes are aligned with rPFS for patients who are on treatment and progression free, and PD (OS-rPFS) for patients who are alive but have progressed in the model.

1-OS, all patients that are not alive; OS, overall survival; rPFS, radiographic progression-free survival.

B.3.2.2.1 Justification of the chosen structure

The strengths of the partitioned survival approach are well-documented (NICE DSU TSD19).[117] As mentioned above, this approach is flexible, and is able to adequately quantify the primary objectives of treating patients with mCRPC, particularly as it is not necessary to model multiple lines of subsequent therapy given the limited treatment options for patients in the post-NHA setting. It directly uses trial-based time-to-event endpoints (OS, rPFS) and it is simple to incorporate indirect comparisons in the form of hazard ratios, where head-to-head trial data is not available.

The selected approach is consistent with previous appraisals relevant to this submission, as summarized in Table 20. The models developed in TA391 (cabazitaxel),[41] TA316[114] and TA377[38] (enzalutamide) and TA101 (docetaxel)[115] are believed to be incorrectly described as Markov models in the manufacturers’ submissions; descriptions of the methods show that state occupancy in the

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

manufacturers’ models were based on survival curves, following a partitioned survival model approach.[117]

Alternative model structures were considered; however, they were ultimately deemed less appropriate for addressing the decision problem, as discussed below:

• A patient-level discrete event simulation model may capture detailed changes along the clinical pathway more accurately (such as treatment pre- and postdocetaxel, and sequencing of subsequent lines of treatment), as in TA387 (abiraterone).[37] However, there were concerns with using the same model structure in the context of this appraisal. Patient-level simulations are more complex in nature and have additional data needs, often requiring access to individual patient-level data. Based on the limited amount of data available for the comparator, it would not be possible to simulate events other than first progression and death without introducing a significant amount of uncertainty into the model. Furthermore, no external data was identified to sufficiently validate patient-level model outcomes, other than rPFS/OS benefit, which are already modelled in a partitioned survival structure.

• Markov models require estimates of transition probabilities between health states. This process involves competing risks and multi-state modelling, consideration of selection effects and dependent censoring, and careful validation. In TA412[112] and TA376[118] (radium-223), the decision to develop a 5-health state Markov model was based on the need to explicitly model symptomatic skeletal-related event outcomes. This is an important measure for radium-223, due to the specific mechanism of action and population that radium-223 is indicated for (i.e. men with CRPC, symptomatic bone metastases and no known visceral metastases). Although the use of olaparib is not limited to patients with symptomatic bone metastases (efficacy has been observed regardless of the site of metastases at baseline), the importance of SREs for mCRPC patients with bone metastases in particular is acknowledged[119] and as discussed in section B.3.3.5. SREs can be considered in a partitioned survival structure in a clear and transparent way, without necessitating the use of a complex Markov structure.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

==> picture [39 x 10] intentionally omitted <==

Intervention technology and comparators

B.3.2.3.1 Intervention

The olaparib dosage is implemented in the economic analysis according to the anticipated European Marketing Authorisation for this indication, and the treatment regimen in the PROfound trial.[61]

The intervention is the tablet formulation of olaparib at the dose of 300 mg (2 x 150 mg tablets) taken twice daily, equivalent to a total daily dose of 600 mg, until disease progression or unacceptable toxicity (whichever occurred first).[61]

B.3.2.3.2 Comparators

In line with the decision problem stated in section 0 (Table 2), the most commonlyused and appropriate comparator for olaparib in current clinical practice (where the majority [~75%] of patients receive docetaxel with ADT for HSPC prior to an NHA, and radium-223 dichloride is received for later lines of treatment). This also aligns with where the evidence base allows for a robust ITC between the intervention and comparator of interest, thus minimising uncertainty and providing a meaningful analysis for decision-making purposes.

Cabazitaxel is administered at a licensed dose of 25 mg/m[2 ] every three weeks in combination with prednisolone 10 mg/day,[120] and for up to a maximum of 10 treatment cycles according to NICE guidelines (TA391).[41] The decision to limit treatment duration to a maximum of 10 cycles was based on the TROPIC study, a ‐ Phase III randomised open label multicentre trial that compared cabazitaxel with mitoxantrone in men with mCRPC whose disease had progressed on or after treatment with docetaxel.

The key trial relevant to the current submission (i.e., in the post-NHA mCRPC setting) is the more recently conducted CARD study, which assessed cabazitaxel versus NHA after disease progression on NHA. It is worth noting that in CARD, cabazitaxel was administered until radiographic disease progression, unacceptable toxicity or patient’s refusal of further study treatment. The range of treatment cycles received was 1 to 29 cycles, with a median of 7 cycles received.[67] The implications of various treatment duration assumptions is discussed further in Section B.3.3.3. Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Cabazitaxel is administered with a specified premedication regimen per the SmPC. Clinical guidelines also recommend the concomitant use of primary prophylactic G- CSF to prevent neutropenia-related complications.[120] Cabazitaxel was administered in the CARD trial in accordance with this.

==> picture [39 x 10] intentionally omitted <==

Perspective of the analysis

The economic evaluation takes an NHS/Personal Social Services (PSS) perspective, as per the NICE reference case. This includes drug acquisition and drug administration costs, costs associated with disease monitoring and resource use, adverse events, skeletal-related events, subsequent treatment and end-of-life care.

==> picture [39 x 11] intentionally omitted <==

Time horizon

A lifetime horizon has been used, consistent with the NICE reference case. This is assumed to be 15 years given that the average age of patients in the prior taxane subgroup of PROfound was approximately 67 years at baseline,[61] (i.e. patients can live up to a maximum of 82 years of age). The time horizon is long enough to capture all important differences in costs or outcomes accrued over the lifetime of a patient with mCRPC while considering that a small minority of patients may respond exceptionally well to treatment in this setting, in line with clinical expert opinion (discussed in Section B.3.2.2).

==> picture [39 x 11] intentionally omitted <==

Cycle length, half-cycle correction and discounting

The model cycle length is 1 month. This is short enough to accurately capture differences in cost or health effects between cycles. Half-cycle correction was applied to prevent under- or over-estimation of costs and quality-adjusted life years (QALYs). Half-cycle correction was not applied to direct drug acquisition and administration costs, since treatments are administered at the start of each cycle and costs would therefore be incurred at the start of each cycle regardless of the patient’s movement thereafter, in line with previous submissions.[41]

The discount rate used for both costs and outcomes is 3.5% per annum, consistent with the NICE reference case.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Table 20. Features of the economic analysis

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca (2020). All rights reserved Page 116 of 208

BNF, British National Formulary; Cab, cabazitaxel; Ctx, chemotherapy; DES, discrete event simulation; EAP, early access programme; ed, edition; EQ-5D, EuroQol 5-Dimension; eMIT, Drugs and pharmaceutical electronic market information tool; HS, health state; mCRPC, metastatic castration-resistant prostate cancer; PFS, progression-free survival; PPRS, PSA, prostate-specific antigen; PSSRU, Personal Social Services Research Unit; NA, not applicable; NICE, National Institute of Health and Care Excellence; NHS, National Health Service; NR, not reported; QALY, quality-adjusted life years; rPFS; radiologically confirmed progression-free survival; TTD, time to treatment discontinuation; UK, United Kingdom

a Based on the Assessment Group’s report, as details of the manufacturer’s submission are not available.115

b Incorrectly described as Markov models in the submission; the description of methods are actually consistent with a partitioned survival model approach. c Assumed based on TA376 Committee Papers.118

d PFS as measured by PSA was also included however ALP- PFS was deemed the most appropriate measure of PFS for the decision problem.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca (2020). All rights reserved Page 117 of 208

B.3.3 Clinical parameters and variables

The clinical outcomes included in the economic analysis are listed below.

• Overall survival

• Radiographic progression-free survival

• Time to treatment discontinuation (scenario analysis only)

• Adverse events

• Skeletal-related events

• Health-related quality of life

==> picture [38 x 10] intentionally omitted <==

Key efficacy data sources

A summary of the main efficacy data sources, analysed population and a description of assumptions needed to conduct the analysis is provided in Table 21. In general, safety outcomes were obtained from the same data source as the efficacy outcomes; however, in some cases (such as for specific AE- and SRE-inputs) it was necessary to obtain data from the literature. These are described in detail in the following sections of the submission.

B.3.3.1.1 Olaparib (PROfound)

The clinical outcomes used to inform the cost-effectiveness analysis were based on patient-level data from the Phase III pivotal study for olaparib, PROfound, at DCO1 (14[th] June 2019). The data used in the cost-effectiveness analysis for olaparib is based on the HRRm (Cohort A+B) prior taxane subgroup of PROfound as described in Section B.1.1, Section B.2.7 and Section B.3.2.1.

B.3.3.1.2 Cabazitaxel (CARD)

Clinical outcomes for cabazitaxel were based on published data as patient-level data from clinical trials were not available. The CARD[67] study was identified as the most relevant source of data for cabazitaxel, as discussed in section B.2.9. The efficacy outcomes (OS, rPFS) for cabazitaxel were estimated based on the anchored ITC results for olaparib versus cabazitaxel, in the absence of head-to-head trial data (as described in section B.2.9.). Safety outcomes were mainly sourced from the CARD study.[67,124]

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Table 21. Summary of main clinical efficacy data sources for each intervention in the economic analysis.

HRRm, homologous recombination repair mutation; ITC, indirect treatment comparison; ITT, intentionto-treat; mCRPC, metastatic castration-resistant prostate cancer; NHA, new hormonal agent; OS, overall survival; (r)PFS, (radiographic) progression-free survival.

==> picture [39 x 10] intentionally omitted <==

Efficacy outcomes

Efficacy outcomes for olaparib were modelled based on time to event analysis of the patient level data from the PROfound trial. Outcomes for the cabazitaxel were modelled by applying the anchored ITC hazard ratios to the olaparib curves as the

reference arm. Details regarding the ITC including an assessment of the proportional hazards assumptions are provided in section B.2.9 and Appendix D.

Methods for rPFS and OS are discussed first, followed by treatment duration (available for scenario analysis) in section B.3.3.2.

B.3.3.2.1 PROfound time to event analysis – olaparib

Given that the median duration of follow-up in the PROfound study was XXX months

(olaparib arm)/XXX months (control arm),[61] and it is necessary to assess the cost-

effectiveness of olaparib over a lifetime horizon, parametric survival analysis was Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

undertaken to extrapolate rPFS (section B.3.3.2.1.1), OS (section B.3.3.2.1.2) and TTD (scenario analysis; section B.3.3.3) to inform the cost-effectiveness model beyond the trial period. Outcomes were analysed based on patient-level data from the HRRm (Cohort A+B) prior taxane subgroup of PROfound.

Six standard parametric models were fitted to rPFS, OS and TTD data from PROfound (exponential, Weibull, log-normal, log-logistic, Gompertz, and generalised gamma).

The methods used to extrapolate outcomes followed the guidance outlined in NICE Decision Support Unit (DSU) Technical Support Documents (TSDs) 14[125] and 18.[96] However, the guidance focuses on situations where patient-level data are available for all interventions and economic analyses are undertaken on a single relevant trial. This is not applicable in this submission as the comparator arm of PROfound (NHA rechallenge) is not reflective of current UK practice for patients previously treated with NHA (Section 0) and was thus not included in this economic analysis.[114,126]

Based on this, it is appropriate to extrapolate survival outcomes with olaparib based on the separately-fitted curves. The direct use of the separately-fitted curves to the olaparib arm data from PROfound represents the best use of the patient-level-data available for olaparib.

For each outcome, an assessment of the fitted models was conducted to determine which parametric survival models were most appropriate. The following factors were considered:

• Statistical goodness of fit (Akaike Information Criterion [AIC]/Bayesian Information Criterion [BIC])

  • The statistical fit of each curve was assessed by considering the total AIC and the BIC values.

• Visual fit to Kaplan–Meier plots

  • The goodness of fit of the parametric curves to the Kaplan–Meier data for olaparib was visually assessed, with consideration given to the entire trial period for which data were available.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

• Clinical plausibility of model extrapolations for OS

  • External validation is greatly important in understanding the suitability of the extrapolated curves.[125] The plausibility of modelled overall survival estimates was validated against UK clinical expert opinion and published literature. This exercise was helpful for understanding the range of plausible outcomes that could be expected under the current standard of care (i.e. cabazitaxel), and to validate the survival extrapolations for olaparib.

Relevant and clinically plausible best fitting models were selected for the base case. Alternative models were considered in sensitivity analysis.

B.3.3.2.1.1 Radiographic progression-free survival (rPFS) B.3.3.2.1.1.1 Olaparib (PROfound, DCO1 4[th] June 2019); HRRm (Cohort A+B) prior taxane subgroup

At DCO1, the rPFS data for the prior taxane subgroup of the overall HRRm (Cohort A+B) population in PROfound were relatively mature although not all patients had experienced an event (72.9% maturity, 124 events in 170 patients). These data are shown in Table 22.

Table 22. Number of events in the HRRm (Cohort A+B) prior taxane subgroup of PROfound.

HRRm, homologous recombination repair; rPFS, radiographic progression-free survival.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Figure 24. Olaparib BICR rPFS, Kaplan–Meier plot (HRRm [Cohort A+B] – Prior taxane).

==> picture [452 x 246] intentionally omitted <==

bid, twice daily NHA, new hormonal agent.

Figure 25. Modelled rPFS for olaparib based on PROfound (HRRm [Cohort A+B] – prior taxane).

==> picture [453 x 128] intentionally omitted <==

==> picture [453 x 127] intentionally omitted <==

Gen, generalised; KM, Kaplan–Meier; rPFS, radiographic progression-free survival.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

The curves for the six parametric models fitted to the olaparib rPFS data (Cohort A+B, prior taxane subgroup) are shown in Figure 25. Based on a visual assessment of the separately-fitted parametric models, all distributions fitted reasonably well to the observed data, which is expected given the relative maturity of the Kaplan–Meier data for rPFS. Although there is some variation in the extrapolated outcomes, all distributions produced similar long-term outcomes for rPFS where <5% of patients are predicted to remain progression-free at 3 years.

According to the total AIC/BIC statistics (Table 23), the Weibull distribution was the best-fitting curve, and was thus used in the base case analysis. The generalised gamma distribution was the next best-fitting curve based on total AIC/BIC and was tested in scenario analysis.

Table 23. AIC and BIC values for parametric models for rPFS (HRRm [Cohort A+B] – prior taxane).

AIC, Akaike information criterion; BIC, Bayesian information criterion; bid, twice daily; Gen, generalised; HRRm, homologous recombination repair; rPFS, radiographic progression-free survival.

B.3.3.2.1.1.2 Cabazitaxel (anchored ITC HR applied to olaparib rPFS curve

In the absence of head-to-head trial data comparing olaparib with cabazitaxel, an anchored ITC was conducted to estimate the relative effectiveness of treatments (Section B.2.9 and Appendix D).

As described in Section B.2.9, treatment with olaparib reduced the risk of radiographic disease progression or death by X..% versus cabazitaxel in the ITC analysis (XXXXXXXXXXXXXXXXXX..). rPFS in the cabazitaxel arm was modelled by applying the reciprocal of the estimates of relative effectiveness from the ITC to

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

the olaparib rPFS curve as the reference arm. The extrapolated curves for cabazitaxel are shown in Figure 26.

Figure 26. Modelled rPFS for cabazitaxel based on ITC HR vs olaparib as reference curve.

==> picture [453 x 127] intentionally omitted <==

==> picture [453 x 127] intentionally omitted <==

Gen, generalised; HR, hazard ratio; ITC, indirect treatment comparison; (r)PFS, radiographic progression-free survival.

B.3.3.2.1.2 Overall survival (OS)

B.3.3.2.1.2.1 Olaparib (PROfound, DCO1 4[th] June 2019); HRRm (Cohort A+B) prior taxane subgroup

Data presented in this section are based directly on the prior taxane subgroup of the PROfound study. At DCO1, OS data for olaparib in the prior taxane subgroup of the overall HRRm (Cohort A+B) population of PROfound were relatively immature (42.9% maturity, 73 events in 170 patients), Table 24. Median OS was 15.8 months in the olaparib arm.

Figure 27 shows the Kaplan–Meier curves for the olaparib arm in the prior taxane subgroup of PROfound.[94] An overlay of Kaplan–Meier curves with parametric models fitted to the patient-level data are shown in Figure 28, with the respective AIC/BIC values for each parametric model presented in Table 25.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Table 24. Number of events in the HRRm (Cohort A+B) prior taxane subgroup of PROfound.

bid, twice daily; OS, overall survival.

Figure 27. OS Kaplan–Meier plot (HRRm [Cohort A+B] – prior taxane subgroup, olaparib)

==> picture [452 x 242] intentionally omitted <==

bid, twice daily; HRRm, homologous recombination repair mutation; OS, overall survival.

Based on a visual assessment of the separately-fitted parametric models for the olaparib arm (Figure 28), the exponential and log-normal distributions generated poor visual fits against the Kaplan–Meier curves for olaparib within the trial period. The Gompertz, generalised gamma and Weibull models were more pessimistic within the observed period, while the exponential and log-normal models were more optimistic. The log-logistic model produced extrapolations in between the optimistic and pessimistic curves.

The total AIC/BIC statistics were similar across distributions, except for the exponential and log-normal curves, which show poor statistical fit to the observed data (Table 25).

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Figure 28. Modelled OS based on PROfound (HRRm [Cohort A+B] – prior taxane subgroup, olaparib).

==> picture [453 x 127] intentionally omitted <==

==> picture [453 x 127] intentionally omitted <==

KM, Kaplan–Meier; OLA, olaparib; OS, overall survival.

Table 25. AIC and BIC values for parametric models for OS (HRRm [Cohort A+B] – prior taxane, olaparib).

AIC, Akaike information criteria; BIC, Bayesian information criteria; OS, overall survival.

PROfound is the first and only Phase III RCT to assess treatment of patients with mCRPC with HRRm whose disease has progressed after treatment with a taxane and NHA. All parametric models predicted a similar median OS estimate for olaparib compared with the observed median OS from PROfound at DCO1 (15.8 months).

Given the lack of other published data on long-term survival with PARP inhibitors in mCRPC, clinical experts experienced in using these treatments in mCRPC clinical

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

trials were consulted to understand which of the OS extrapolations best captured patient outcomes that could be realised in real-world clinical practice.

Based on UK clinical expert opinion,[58] long-term survival with olaparib is expected to be better than that achieved with the current standard-of-care in the real world setting, where a small number of patients remain alive 5 to 10 years after starting treatment in a post-NHA setting (described further in Section B.3.3.2.1.2.2). Based on an average of responses, approximately XX% of patients and X% of patients who have previously received docetaxel and who have progressed on a prior NHA could remain alive 5 and 10 years after starting treatment with olaparib. Since the Weibull, Gompertz and generalised gamma distributions predicted no long-term survivorship at these timepoints for olaparib or with cabazitaxel (Section B.3.3.2.1.2.2), these distributions were considered clinically implausible and inappropriate to inform decision making.

The only remaining clinically plausible models are the exponential, lognormal and log-logistic curves. Of these curves, the lognormal and log-logistic distributions provided estimates of 5 and 10 year survival that were closest to those predicted by UK clinical experts. Although the exponential curve predicts a plausible OS rate at 5 years, it underestimates 10-year survivorship and showed poor visual fit to observed data, and was thus discarded as an option to model OS.

Of the lognormal and log-logistic distributions, the log-logistic model provided better statistical and visual fit to data. The log-logistic distribution was therefore used in the base-case analysis, despite producing more conservative estimates of long-term survival than that predicted by clinical experts. The lognormal distribution produced long-term survival estimates that most-closely reflected clinical-expert estimates, and was therefore tested in scenario analysis.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Table 26 . OS estimates for olaparib (HRRm [Cohort A+B] - prior taxane, olaparib)

Gen, generalised; OS, overall survival.

a Yes = 5- and/or 10-year survival do not contradict estimates provided by clinical experts (i.e. long-term survivorship is non-zero); No = 5- and 10-year survival estimates contradict estimates provided by clinical experts.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca (2020). All rights reserved Page 128 of 208

B.3.3.2.1.2.2 Cabazitaxel (anchored ITC applied to olaparib OS curve)

As described in Section B.2.9, treatment with olaparib reduced the overall risk of death by X..% versus cabazitaxel in the anchored ITC (OS HR, XXXXXXXXXXXX), using data from the CARD[67] and PROfound studies.[30,61] The approach to modelling OS with cabazitaxel is the same as with rPFS, with the resulting curves shown in Figure 29.

Figure 29. Modelled OS for cabazitaxel based on ITC HR vs olaparib as reference curve.

==> picture [478 x 91] intentionally omitted <==

==> picture [478 x 91] intentionally omitted <==

==> picture [478 x 91] intentionally omitted <==

Gen, generalised; HR, hazard ratio; ITC, indirect treatment comparison; OS, overall survival.

Median OS for cabazitaxel projected by all six distributions were similar to the median OS of 13.6 months observed in the cabazitaxel arm of the CARD trial.[67] In the absence of alternative sources of published literature on long-term survival outcomes achieved with cabazitaxel in a post-NHA setting (Section B.2), UK clinical expert opinion was sought to understand real-world survival outcomes on cabazitaxel and to validate the choice of distribution used in the base case analysis.

Based on UK clinical expert opinion, approximately XXXX. of patients could survive 5 years from the start of cabazitaxel in a post-NHA mCRPC (Table 27).[58] Clinicians

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

had experience with a small number of long-term survivors and all respondents expected some patients to survive 10 years from the start of treatment. In the average of responses, XXX. of patients could survive 10 years from the start of cabazitaxel in the post-NHA mCRPC setting.

As with the comparison of survival estimates for olaparib, the Weibull, Gompertz and generalised gamma distributions drastically underestimated survival with cabazitaxel, and produce clinically implausible long-term OS outcomes that would not be appropriate to use for the purposes of decision making.

Of the remaining distributions, the log-logistic distribution produced the closest estimate for 3-year survival compared with the observed data at 2.8 years in the CARD study publication, and was clinically plausible albeit slightly conservative at 10 years, confirming that the use of the log-logistic distribution is appropriate and valid in terms of outcomes with cabazitaxel. The lognormal distribution produced longterm survival estimates for cabazitaxel that most closely matched estimates from UK clinician experts. There is consistency in the conclusions regarding the log-logistic and lognormal distributions for both cabazitaxel and olaparib, which provides support for validity of using these models in the base case and scenario analysis.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640] © AstraZeneca (2020). All rights reserved Page 130 of 208

Table 27. OS for cabazitaxel (predicted via ITC)

Gen, generalised; ITC, indirect treatment comparison; OS, overall survival

• Last time point available based on data published by de Wit 2019.[67]

a Yes = 5- and/or 10-year survival do not contradict estimates provided by clinical experts (i.e. long-term survivorship is non-zero); No = 5- and 10-year survival estimates contradict estimates provided by clinical experts.

Company evidence submission template for olaparib for previously treated hormone-relapsed metastatic prostate cancer [ID1640] © AstraZeneca (2020). All rights reserved Page 131 of 208

B.3.3.2.1.3 Summary of rPFS and OS distributions included in analysis

The choice of rPFS and OS curves used in the base case and sensitivity analyses is summarised in Table 28 along with a brief description of the rationale explained in Section B.3.3.2.1.1 and Section B.3.3.2.1.2. The base case rPFS and OS distributions for olaparib and cabazitaxel are shown in Figure 30 and Figure 31.

Table 28. Final rPFS and OS models selected for economic evaluation (HRRm [Cohort A+B] – prior taxane)

AIC, Akaike information criteria; BIC, Bayesian information criteria; HRRm, homologous recombination repair mutation; OS, overall survival; rPFS, radiographic progression-free survival.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Figure 30. Selected distribution for extrapolating OS for olaparib (PROfound, Kaplan–Meier estimate and loglogistic model; prior taxane subgroup) and cabazitaxel (via ITC).[a, b]

==> picture [453 x 247] intentionally omitted <==

a Unadjusted Kaplan–Meier estimates, based on PROfound (prior taxane subgroup) for olaparib; anchored ITC HR (section B.2.9) for cabazitaxel vs olaparib applied to the modelled olaparib curve (half-cycle correction applied).

b Best fitting curve, log-logistic, based on combined assessment of visual and statistical fit to the Kaplan–Meier data, comparison of median OS estimates and clinical plausibility based on feedback from UK clinical experts.

HR, hazard ratio; KM, Kaplan–Meier; ITC, indirect treatment comparison; OS, overall survival.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Figure 31. Selected distribution for extrapolating rPFS for olaparib (PROfound, Kaplan–Meier estimate and Weibull model; prior taxane subgroup) and cabazitaxel (via ITC).[a,b ]

==> picture [453 x 247] intentionally omitted <==

a Unadjusted Kaplan–Meier estimates, based on PROfound for olaparib; anchored ITC HR for cabazitaxel vs olaparib applied to the modelled olaparib curve (half-cycle correction applied). b Best fitting curve, Weibull, based on combined assessment of visual and statistical fit to the Kaplan– Meier data

HR, hazard ratio; KM, Kaplan–Meier; ITC, indirect treatment comparison; rPFS, radiographic progression-free survival.

==> picture [39 x 11] intentionally omitted <==

Treatment duration

Treatment duration was explicitly modelled to accurately estimate treatment costs associated with each intervention. Several options were implemented to project a treatment duration curve over the time horizon, depending on the type of data that was available for each intervention, as outlined in Table 29.

Additionally, a maximum treatment duration of 10 treatment cycles was applied for cabazitaxel to ensure that costs reflected the duration of treatment administration in England, per NICE recommendations for cabazitaxel based on the TROPIC trial (TA391).[41] It should be noted that efficacy data from the CARD study, which was used in the ITC analysis to estimate the relative efficacy of olaparib versus cabazitaxel, reflects the administration of cabazitaxel until radiographic disease progression, unacceptable toxicity or patient’s refusal of further study treatment

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

(median number of treatment cycles received, 7; range 1 to 29[67] ). Therefore, the approach of restricting the costs of cabazitaxel to a maximum of 10 treatment cycles without adjusting the efficacy estimates for the duration of treatment may lead to conservative estimates of cost-effectiveness for olaparib versus cabazitaxel. The impact of removing the maximum treatment duration for cabazitaxel on the results is tested separately in a scenario.

Table 29. Summary of TTD approach selected for base case and scenario analysis

TTD, time to treatment discontinuation.

B.3.3.3.1 Treat to progression (base case)

The treat to progression option assumes that patients receive treatment up until the point of progression, according to the rPFS curves defined in Section B.3.3.2.1.1.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

This treatment duration rule accurately aligns with the expected use of olaparib based on its anticipated marketing authorisation for this indication, and the use of cabazitaxel (which is limited to a maximum of 10 treatment cycles by NICE).[41]

B.3.3.3.2 Median duration

The median duration of treatment is a commonly-reported outcome in clinical trials as a measure of the average time that patients received active treatment. As a median value was available for olaparib treatment in the PROfound study[61] , and for cabazitaxel treatment in CARD,[67] it was possible to utilise this data as a common method of modelling treatment duration for each intervention (XX.and 5.1 months, respectively).

In the economic model, the median duration of time on treatment was used to predict the probability of treatment discontinuation at each model cycle using the exponential distribution:

Given the median treatment duration of X, the probability of remaining on on treatment at cycle t is:

exp((ln(0.5)/X)* t )

B.3.3.3.3 Parametric time to treatment discontinuation (TTD) curve

In this option, time on treatment was explicitly modelled using parametric curves fitted to time-to-treatment-discontinuation patient-level data following methods described in section B.3.3.2.1 and as described below. In the model, the TTD curves are constrained by rPFS to ensure logical patient flow according to the anticipated Marketing Authorisation for olaparib. That is, once a patient has confirmed evidence of progression, they are assumed to also discontinue treatment. This option was only available for olaparib as no other published time to treatment discontinuation data, either in the form of Kaplan–Meier data or parametric models, were available for cabazitaxel.

In the prior taxane subgroup of the overall HRRm (Cohort A+B) population of PROfound, XX. out of 170 patients had discontinued treatment in the olaparib arm; the median duration of treatment was XX months (excluding dose interruptions).[61]

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Figure 32 shows the Kaplan–Meier curves for time to treatment discontinuation, whilst Figure 33 shows an overlay of the Kaplan–Meier and modelled curves. As with rPFS and OS extrapolations, TTD was modelled based on separately-fitted parametric curves to the olaparib arm, since treatment with NHA (as the comparator arm of PROfound) is not a relevant comparator in this appraisal. Table 30 presents the AIC/BIC values for separately-fitted TTD curves to the olaparib data.

Based on an assessment of the visual and statistical fit of the curves to the Kaplan– Meier data, the Weibull was determined to be the best-fitting TTD curve and was used in a scenario analysis.

– – Figure 32. TTD, Kaplan Meier plot (HRRm prior taxane)

==> picture [453 x 247] intentionally omitted <==

bid, twice daily; BICR, blinded independent committee review; HRRm, homologous recombination repair mutation; TTD, time to treatment discontinuation.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

– Figure 33. Modelled treatment duration based on PROfound (HRRm prior taxane, olaparib).

==> picture [453 x 246] intentionally omitted <==

Gen gamma, generalised gamma; HRRm, homologous recombination repair; KM, Kaplan–Meier; TTD, time to treatment discontinuation

Table 30. AIC and BIC values for parametric models for TTD (HRRm – prior taxane, olaparib)

AIC, Akaike information criterion; BIC, Bayesian information criterion; HRRm, homologous recombination repair; TTD, time to treatment discontinuation.

==> picture [39 x 10] intentionally omitted <==

Adverse events

Adverse events have a quality of life and resource impact. There are clinically meaningful differences in the AE profiles for olaparib and cabazitaxel.

For each treatment, it was considered preferable to derive safety data from the same study that was used to determine relative efficacy. This ensures that the AEs

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

accurately reflect those that are relevant to the treatment, as observed in the safety and efficacy assessment in clinical trials. Furthermore, using the same data source for safety and efficacy inputs avoids introducing uncertainty related to cross-study differences (e.g. differences in trial populations or drug administration).

Treatment-related AE rates for olaparib were based on the PROfound trial.[61] AEs reported within the prior taxane subgroup of the overall HRRm population of PROfound were used in line with the modelled population. AE data for cabazitaxel were obtained from the CARD[67] study, with the exception of leukopenia and neutropenia, as described below:

• There is uncertainty around the AE incidence rates associated with cabazitaxel for leukopenia and neutropenia, since the de Wit et al , 2019 publication reported Grade 3 and above “laboratory abnormalities” in 32% and 45% of patients, respectively.[67] It was stated that these were based on systematic analysis of blood samples, which “may not have been reported as an AE”. Clinical experts consulted by AstraZeneca on this topic confirmed that these are purely laboratory test results and would not necessarily reflect the actual incidence of complications causing a quality of life or resource use impact (such as infections and febrile neutropenia).

• To avoid overestimating the incidence of leukopenia and neutropenia associated with cabazitaxel treatment these data were obtained directly from UK clinical experts (the lowest of estimates from the responses were assumed: 5% leukopenia, 5% neutropenia), in the context of patients receiving primary prophylactic G-CSF with cabazitaxel (i.e., in line with the administration of cabazitaxel in the CARD study). This is an appropriate approach because it aligns with current expectations in clinical practice in England and Wales

The economic model included Grade 3 and above AEs occurring in at least 5% of patients in the olaparib arm of PROfound (prior taxane subgroup)[61] or cabazitaxel from CARD.[67] This is a commonly accepted approach as Grade 3 and above AEs reflect events that are likely to require hospitalization; therefore, having the greatest burden on resources and quality of life.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

In addition, the following Grade 3 and above AEs were included, based on advice from UK clinical experts, even though they may occur in <5% of patients:

• Febrile neutropenia was included based on its relevance to this indication as a key concern with chemotherapy.

• Diarrhoea was also included due to the potential to lead to infections at the Grade 3 and above level.

• Fatigue/asthenia and musculoskeletal pain or discomfort was included due to the impact on quality of life, a particularly important consideration for patients on chemotherapy.

• Thrombocytopenia was included due to its severity and potential impact on resource use.

The AEs included in the economic model are summarised in Table 31. Based on a naïve side-by-side comparison, olaparib had a more tolerable and manageable safety profile compared with cabazitaxel, with either a lower of very similar incidence of Grade 3 or above AEs occurring in ≥5% of patients with the exception of anaemia.

Table 31. Grade 3 and above AEs affecting at least 5% of patients included in base case analysis.

a Described in de Wit et al . 201967 as including back pain, flank pain, musculoskeletal discomfort and/or pain, neck pain, or pain in extremities. No related events were reported in PROfound. b Input values based on clinical expert advice on the incidence of leukopenia/neutropenia (Grade 3 and above) that would require hospitalisation (data on file).[102]

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

c Laboratory abnormalities reported in de Wit et al . 201967 may not have been reported as an adverse event in CARD although the values were used as clinical experts confirmed that this reflected what they would expect in clinical practice.[102]

e Occurred in fewer than 5% of patients in PROfound/CARD, but added to the list of AEs (validated by UK clinical experts).[61,67]

==> picture [39 x 10] intentionally omitted <==

Skeletal-related events (SREs)

B.3.3.5.1 Overall occurrence of SREs

SREs are a key clinical aspect of mCRPC due to the high propensity for prostate cancer to metastasise to bone tissue. These events were therefore included in the economic model as a one-off cost and SRE-specific utility decrement for patients upon progression. The definition of SREs in the PROfound trial protocol[30,87] was consistent with the clinical literature and those used in other mCRPC trial protocols (for example, CARD [in the post-NHA mCRPC setting relevant to this appraisal],[67] and also ALSYMPCA [for radium-223 dichloride in patients with bone metastases only],[127] AFFIRM [for enzalutamide in a post-docetaxel population][107] and COU-AA301 [abiraterone in a post-docetaxel population][128] ). Based on data for the prior taxane subgroup of the overall HRRm population of PROfound, at least one SRE occurred in XXXX of patients in the olaparib arm.[61] In the CARD study publication, SREs occurred in 18.6% of patients in the cabazitaxel arm, which is included in the analysis.[67]

B.3.3.5.2 Distribution of interventions used to manage SREs

In an attempt to more accurately capture the true impact of each SRE, following the approach used in previous NICE submissions in prostate cancer such as TA580,[57] TA316,[114] TA376,[118] and TA412,[43] the distribution of SRE therapies were incorporated into the economic model and used to weight SRE-related costs.

However, the distribution of each SRE was not analysed in PROfound and data were not available from the CARD study (as described above).[61,67] A targeted literature search was therefore conducted to identify relevant data that could inform the economic model.

The search identified a detailed breakdown of SREs from the ALSYMPCA (radium-

223 dichloride versus placebo/best supportive care),[118] COU-AA-301 (abiraterone Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

plus prednisone versus placebo plus prednisone in a post-docetaxel setting)[129] and AFFIRM studies (enzalutamide versus placebo in a post-docetaxel setting).[63,114] The average value across trials was used in the economic model. The distribution of SRE therapies were assumed to be the same for olaparib and cabazitaxel, as presented in Table 32. UK clinical experts consulted by AstraZeneca confirmed that this is an appropriate method, and that the average values shown in Table 32 are generalisable to UK clinical practice in a post-NHA setting.

Table 32. Distribution of SREs identified from mCRPC clinical trials.

a AFFIRM pooled rates for radiation and surgery to the bone from Saad 201763 adjusted proportionally with ratio of events reported in NICE TA316 (AFFIRM interim analysis).[114]

The data used in the base case analysis are summarized in Table 33. The impact of uncertainty and various limitations of SRE assumptions in the economic model were tested extensively in scenario analyses and had minimal impact on the results compared with the base case analysis.

Table 33. Probability of SREs occurring with each intervention in the base case analysis.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

a Reported as the number of patients experiencing at least one SRE. b AFFIRM pooled rates for radiation and surgery to the bone from Saad 201763 adjusted proportionally with ratio of events reported in NICE TA316 (AFFIRM interim analysis).[114]

B.3.4 Measurement and valuation of health effects

The economic model uses HRQoL data collected from the PROfound study as detailed in the following sections (Section B.3.4.1 to Section B.3.4.3). Mean baseline health state utility values were analysed and applied to patients in the progressionfree and progressed-disease health states in the model, assumed to be equal across interventions.

Additionally, the following adjustments were incorporated to capture important differences in the impact of treatment with olaparib versus cabazitaxel on patients’ HRQoL; as the data required for these adjustments were not available from PROfound, inputs were sourced from previous NICE technology appraisals in prostate cancer and supplemented by targeted literature reviews:

• Utility decrements to AEs (Section B.3.4.4) due were incorporated to reflect differences in the safety profiles of olaparib and cabazitaxel (Section B.3.3.4);

• Utility decrements due to SREs (Section B.3.4.5) for the occurrence of SREs with olaparib and cabazitaxel (Section B.3.3.5);

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

• A modality-specific utility decrement was applied for the duration of treatment with cabazitaxel to reflect the negative impact of the intravenous administration of chemotherapy on patients’ quality of life, and the benefit of olaparib taken orally (Section B.3.4.6).

==> picture [38 x 11] intentionally omitted <==

Health-related quality-of-life data from clinical trials

To investigate the impact of treatment and the disease on patients’ HRQoL, EQ-5D5L data were collected in PROfound as a predefined exploratory endpoint. EQ-5D-5L assessments were carried out at baseline (day 1 on study treatment), and then every 8 weeks (+/- 7 days) until discontinuation of study treatment, and on the last dose of study drug (see section B.2.6.6 for details).[61] EQ-5D-5L was also collected every 8 weeks for up to 24 weeks after discontinuation of study treatment. EQ-5D data (to week 64) are presented in section B.2.6.4.2.

Patients were classified as being in the following health states:

• Progressed disease (PD) state: Any EQ-5D collected after the date of progression was classified as being observed while PD

• Progression-free (PF) state: Any EQ-5D collected prior to the date of progression or censoring for progression was classified as being observed while PF

• Any EQ-5D collected after the date of censoring for progression was classified as “PF/PD unknown.”

Given the NICE position statement regarding the use of EQ-5D in technology appraisals,[130,131] the EQ-5D-5L measurements collected in the PROfound trial were analysed and mapped onto EQ-5D-3L, as described in section B.3.4.2.

==> picture [39 x 11] intentionally omitted <==

Mapping

EQ-5D-5L measurements collected in the PROfound trial were mapped onto the EQ5D-3L. HSUV were provided based on the societal preferences of the general population in the UK using the value sets developed (Dolan et al , 1993)[132] for the mapping (“crosswalk”) approach to the three-level version of the EQ-5D (EQ-5D-3L) published by van Hout et al . 2012.[132]

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Full details of the cross-walk methods and key model results are provided in Appendix H.

The impact of treatment (olaparib vs NHA) and disease progression (PF vs PD) on HSUVs were formally assessed using a linear mixed effects models fitted to observed data in the full analysis set of the HRRm population. These models give appropriate estimates of health state utility under the assumption that any missing data are missing at random. The best fitting model was chosen based on AIC/BIC criteria and the rest of the mixed effect modelling analysis was performed using that model. The regression showed a significant association between HSUV and progression and an improvement in EQ-5D-5L for olaparib vs NHA.

See Appendix H for more details of the outputs from these analyses.

The mean EQ-5D-3L HSUVs derived for PF and PD are shown in Table 34, and were assumed to be the same for olaparib and cabazitaxel.

Table 34. Mean (LSM) HSUV estimates by progression status (HRRm population)

Note: These are the estimates for PF/PD for olaparib, if estimated by the model with treatment and progression as covariate. We have conservatively assumed the same utilities by treatment arm in the model (e.g. cabazitaxel and olaparib).

BICR, blinded independent central review; CI, confidence intervals; HRRm, homologous recombination repair mutation; LSM, least squares mean; PF, progression-free; PD, progressive disease; SE, standard error.

==> picture [39 x 10] intentionally omitted <==

Health-related quality-of-life studies

Appendix H provides details of the SLR search strategy for identifying health-related quality-of-life studies in the published literature. Two studies were identified specific to the patient population of interest in a post-NHA setting, however in both studies

the size of the evaluated cohort was small (< 50 patients), and neither reported health state utility values. Therefore, the EQ-5D analysis from the PROfound study

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

was deemed to be the most relevant source of data for inclusion in the economic model.

Given the lack of published evidence in the post-NHA setting, a targeted literature review not restricted to post-NHA was conducted in an attempt to identify any studies that could be helpful for validating HSUV assumptions for olaparib or cabazitaxel in the appraisal (Table 35). Only one study was identified as appropriate: the UK early access programme (EAP) for cabazitaxel, which was an observational study conducted as part of an international phase IIIB/IV study within the bounds of clinical practice.[133] The primary objective of the early access programme was to allow early access to cabazitaxel for patients in clinical practice and similar to those evaluated in the TROPIC trial (i.e., patients with mCRPC who have received prior docetaxel); with the secondary objective being to document the safety of cabazitaxel in these patients.[133][41] , which is unique to the administration of cabazitaxel. The UK EAP included a total of 112 patients mCRPC with disease progression during or after docetaxel, across 12 UK cancer centres. Patients received cabazitaxel 25 mg/m[2] every 3 weeks with prednisolone 10 mg daily for up to 10 cycles in line with the SmPC. Full details are provided in Bahl 2015 and summarised in TA391.[41,133] The reported mean HSUV for PF was 0.7533 (Bahl, 2015)[133] or 0.7370 based on the weighted average of utility values across cycles (as reported in TA391), and 0.6266 for PD.[41,133] These values are similar to the EQ-5D analysis from PROfound, and were therefore tested in scenario analyses to understand the impact of utility assumptions on the results.

Company evidence submission template for olaparib for previously treated hormonerelapsed metastatic prostate cancer [ID1640]

© AstraZeneca (2020). All rights reserved

Table 35. Utility values based on targeted literature search in mCRPC setting (not specific to post-NHA population).