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

Single Technology Appraisal

Pembrolizumab with platinum-based chemotherapy for untreated advanced oesophageal cancer [ID3741]

Committee Papers

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL

Pembrolizumab with platinum-based chemotherapy for untreated advanced oesophageal cancer [ID3741]

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 MSD

2. Clarification questions and company responses a. Appendices

3. Patient group, professional group and NHS organisation submissions from: a. Guts UK b. Heartburn Cancer UK

4. Evidence Review Group report prepared by PenTAG

5. Evidence Review Group report – factual accuracy check

6. Technical engagement response from company a. New evidence

7. Technical engagement responses and statements from experts: a. Dr Elizabeth Smyth, Consultant in Gastrointestinal Oncology – clinical expert, nominated by the Royal College of Physicians b. Prof. Wasat Mansoor, Consultant in Medical Oncology – clinical expert nominated by MSD

8. Technical engagement responses from consultees and commentators: a. NCRI-ACP-RCP-RCR

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

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma

[ID3741]

Document B Company evidence submission

8[th] February 2021

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Contents

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Tables and figures

Tables

Table 1 The decision problem ............................................................................................. 11 Table 2 Technology being appraised .................................................................................. 12 Table 3 Clinical effectiveness evidence ............................................................................... 18 Table 4 Trial Treatments ..................................................................................................... 23 Table 5: Summary of trial methodology ............................................................................... 28 Table 6: Subject Characteristics (ITT Population) – KEYNOTE-590 ................................... 30 Table 7: Statistical analysis plan summary .......................................................................... 32 Table 8: KEYNOTE-590– Analysis strategy for key efficacy endpoints ............................... 35 Table 9: Censoring Rules for Primary and Sensitivity Analyses of PFS ............................... 36 Table 10. Summary of statistical analyses .......................................................................... 39 Table 11. Quality assessment results for KEYNOTE-590 .................................................... 40 Table 12. Analysis of OS (ITT): IA July 2020 data cut ......................................................... 42 Table 13. Summary of OS Rate Over Time (ITT): IA July 2020 data cut ............................. 43 Table 14. Analysis of Overall Survival (Subjects with PD-L1 CPS >= 10, ITT Population) ... 45 Table 15. Overall Survival Rate (Subjects with PD-L1 CPS ≥ 10, ITT Population) ............... 46 Table 16. Analysis of Progression-Free Survival Based on Investigator Assessment per RECIST 1.1 (Primary Censoring Rule): IA July 2020 data cut ............................................. 48 Table 17. Summary of PFS Rate Over Time Based on Investigator Assessment per RECIST 1.1: IA July 2020 data cut .................................................................................................... 49 Table 18. Analysis of Progression-Free Survival Based on Investigator Assessment per RECIST 1.1 primary Censoring Rule) (Subjects with PD-L1 CPS >= 10, ITT Population) ... 50 Table 19. Summary of PFS Rate Over Time Based on Investigator Assessment per RECIST 1.1 (Subjects with PD-L1 CPS ≥ 10, ITT Population)........................................................... 50 Table 20. Analysis of Objective Response with Confirmation Based on Investigator Assessment per RECIST 1.1 (ITT); IA July 2020 data-cut ................................................... 52 Table 21. Summary of Best Overall Response Based on Investigator Assessment per RECIST 1.1 with Confirmation (ITT); IA July 2020 data-cut ................................................. 52 Table 22. Analysis of Objective Response with Confirmation Based on Investigator Assessment per RECIST 1.1 (Subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10), ITT Population) .......................................................................................................... 53 Table 23. Summary of Best Overall Response Based on Investigator Assessment per RECIST 1.1 with Confirmation (Subjects with PD-L1 CPS ≥ 10, ITT Population) ................ 53 Table 24. Analysis of Change from Baseline in EQ-5D VAS to Week 18; IA July 2020 datacut ....................................................................................................................................... 55 Table 25. Analysis of Change from Baseline in EQ-5D VAS to Week 18 (Subjects with PDL1 CPS ≥ 10, FAS Population) ............................................................................................ 55 Table 26. Summary of the trials of relevance identified through the SLR............................. 57 Table 27: Treatment characteristics of included trials .......................................................... 58 Table 28. Patient characteristics of trials identified through the SLR (age, sex, region, performance score) ............................................................................................................. 59 Table 29. Patient characteristics of trials identified through the SLR (disease stage, cancer type, histology) .................................................................................................................... 59 Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 30. Summary of trials included in the UK feasibility assessment ............................... 62 Table 31. Extent of Exposure - Summary of Duration on Therapy (ASaT Population) ......... 64 Table 32. Exposure by duration (ASaT population) ............................................................. 65 Table 33. Disposition of Subjects (ITT Population) .............................................................. 65 Table 34. Adverse Event Summary (ASaT Population) ....................................................... 66 Table 35: Exposure-Adjusted Adverse Event Summary (Including Multiple Occurrences of Events) (ASaT Population) .................................................................................................. 67 Table 36: Subjects With Adverse Events By Decreasing Incidence (Incidence ≥ 10% in One or More Treatment Groups) (ASaT Population) ................................................................... 68 Table 37: Exposure-Adjusted Adverse Events (Including Multiple Occurrences of Events) (Incidence ≥ 10% in One or More Treatment Groups) (ASaT Population) ........................... 70 Table 38: Subjects with Grade 3-5 Adverse Events by Decreasing Incidence (Incidence ≥ 5% in One or More Treatment Groups) (ASaT Population) ................................................. 71 Table 39: Subjects with drug-related grade 3-5 adverse events by decreasing incidence (Incidence ≥ 5% in One or More Treatment Groups) (ASaT Population) ............................. 73 Table 40. End-of-Life Criteria .............................................................................................. 78 Table 41. Baseline characteristics of patients included in the model ................................... 80 Table 42. Features of the economic analysis ...................................................................... 82 Table 43 Summary of NG83 Section 9.2.6 .......................................................................... 83 Table 44. Summary of goodness-of fit qualities of OS models for pembrolizumab in combination with chemotherapy and SOC .......................................................................... 91 Table 45. Long term Overall Survival estimates for SOC .................................................... 92 Table 46. Long term Overall Survival estimates for pembrolizumab in combination with chemotherapy ..................................................................................................................... 92 Table 47. Summary of goodness-of-fit qualities of the piecewise PFS models for pembrolizumab in combination with chemotherapy and chemotherapy ............................... 96 Table 48. Grade 3+ AE rates for AEs included in the model ................................................ 98 Table 49. Duration of Grade 3+ AEs in KEYNOTE-590 ....................................................... 99 Table 50. EQ-5D health utility scores by time-to-death ...................................................... 103 Table 51. EQ-5D health utility scores by progression status (pooled) ................................ 103 Table 52. Compliance of EQ-5D by visit and by treatment (FAS Population) .................... 103 Table 53. One-off QALY decrements due to Grade 3+ AEs by treatment arm ................... 106 Table 54. Summary of utility values for cost-effectiveness analysis................................... 107 Table 55. Acquisition costs for drugs considered in first-line therapy ................................. 109 Table 56. Relative dose intensities for treatments included in the economic model ........... 112 Table 57. Administration costs for regimens included in the economic model ................... 113 Table 58. Resource use and unit costs of progression-free, progressed health states within the model .......................................................................................................................... 114 Table 59. Unit costs of adverse events ............................................................................. 114 Table 60. One-off costs of AEs within the model ............................................................... 116 Table 61. Proportion of patients who receive each subsequent treatment (across all subsequent lines ≥5% in any arm) .................................................................................... 116 Table 62. Dosing schedules, drug acquisition costs, drug administration costs and treatment duration by subsequent therapy ........................................................................................ 117 Table 63 One-off subsequent costs applied in the model .................................................. 117 Table 64. Summary of variables applied in the economic model ....................................... 119

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 65. List of assumptions used in the economic model ............................................... 123 Table 66. Base-case results versus trial comparator SOC (discounted price) ................... 125 Table 67. Base-case results versus selected non-trial comparator (discounted price) ....... 126 Table 68. Base-case results versus non-trial blended comparator (discounted price) ....... 126 Table 69. Incremental cost-effectiveness results based on probabilistic sensitivity analysis versus trial comparator SOC (discounted price) ................................................................ 127 Table 70. Results from the scenario analyses versus trial comparator SoC (discounted price) ......................................................................................................................................... 132 Table 71. Incremental cost-effectiveness results for pembrolizumab in combination with chemotherapy vs. SOC for patients with CPS≥10 (discounted price) ................................ 135

Figures Figure 1: NICE pathway on locally advanced or metastatic oesophago-gastric cancer ....... 16 Figure 2 Schematic of KEYNOTE - 590 .............................................................................. 20 Figure 3. Maurer and Bretz multiplicity strategy approach used for hypothesis testing in KEYNOTE-590.................................................................................................................... 37 Figure 4: KM Estimates of OS (ITT); IA July 2020 data-cut ................................................. 44 Figure 5. Kaplan-Meier Estimates of Overall Survival (Subjects with PD-L1 CPS >= 10, ITT Population) .......................................................................................................................... 46 Figure 6. KM Estimates of PFS (Primary Censoring Rule) Based on BICR Assessment per RECIST 1.1 (ITT): IA July 2020 data-cut ............................................................................. 49 Figure 7. Kaplan-Meier Estimates of Progression-Free Survival Based on Investigator Assessment per RECIST 1.1 (Primary Censoring Rule) (Subjects with PD-L1 CPS >= 10, ITT Population) ................................................................................................................... 51 Figure 8. Network diagram of studies identified through SLR .............................................. 62 Figure 9: Between-Treatment Comparisons in Adverse Events Selected Adverse Events (>= 10% Incidence) and Sorted by Risk Difference (ASaT Population) ...................................... 69 Figure 10. Model structure .................................................................................................. 80 Figure 11. Partitioned survival model structure.................................................................... 81 Figure 12. Survival model selection process algorithm (adapted from TSD 14) ................... 86 Figure 13. Cumulative hazard plot of OS for pembrolizumab in combination with chemotherapy and SOC ...................................................................................................... 87 Figure 14. Log-cumulative hazard plot of OS for pembrolizumab + chemotherapy and SOC ........................................................................................................................................... 87 Figure 15. Plot of multiple Chow test statistics for OS in KEYNOTE-590: pembrolizumab in combination with chemotherapy, overall population ............................................................ 89 Figure 16. OS KM curve with fitted piecewise model, 40-week cut-off, for pembrolizumab + chemotherapy based on KEYNOTE-590 ............................................................................. 90 Figure 17. OS KM curve with fitted piecewise model, 40-week cut-off, for SOC based on KEYNOTE-590.................................................................................................................... 90 Figure 18. OS KM curves with fitted piecewise parametric models with a 40-week cut-off for the OS of pembrolizumab in combination with chemotherapy and SOC based on KEYNOTE590 over a 5-year period ..................................................................................................... 94

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 19. OS KM curves with fitted piecewise parametric models with a 40-week cut-off for the OS of pembrolizumab in combination with chemotherapy and SOC based on KEYNOTE590 over a lifetime horizon (20-year period) ........................................................................ 94 Figure 20. PFS KM curve vs. fitted piecewise parametric curves for pembrolizumab in combination with chemotherapy .......................................................................................... 95 Figure 21. PFS KM curve vs fitted piecewise parametric curves for chemotherapy ............. 96 Figure 22. PFS KM curves fitted with log-logistic curves from 10 weeks (base case) .......... 97 Figure 23. ToT KM curves for pembrolizumab in combination with chemotherapy based on KEYNOTE-590.................................................................................................................. 111 Figure 24. ToT KM curve for chemotherapy based on KEYNOTE-590 .............................. 111 Figure 25. Scatterplot of PSA results (1,000 simulations) versus trial comparator SOC (discounted price) ............................................................................................................. 128 Figure 26. Cost-effectiveness acceptability curve versus trial comparator SOC (discounted price) ................................................................................................................................. 128 Figure 27. Tornado diagram presenting the results of the deterministic sensitivity analysis for the 10 most sensitive variables versus trial comparator SOC (discounted price) .............. 130

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Abbreviations

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

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

B.1.1 Decision problem

The submission covers the technology’s full marketing authorisation for this indication: KEYTRUDA, in combination with platinum and fluoropyrimidine based chemotherapy, is indicated for the first-line treatment of patients with locally advanced unresectable or metastatic carcinoma of the oesophagus or HER-2 negative gastroesophageal junction adenocarcinoma in adults.

Please see Table 1 below for a summary of the National Institute for Health and Care Excellence (NICE) decision problem.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 1 The decision problem

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B.1.2 Description of the technology being appraised

The draft summary of product characteristics (SmPC) has been included in Appendix C; however, the European Public Assessment Report (EPAR) that includes the indication under assessment in this submission was not available at the time of the submission. The technology being appraised (pembrolizumab) is described in the Table 2 below.

Table 2 Technology being appraised

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

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

Oesophageal cancer is the eighth most common type (1) of cancer in the world. The UK has the highest age-standardised incidence of oesophageal cancer in Europe (2), with approximately 9,000 people diagnosed annually in the UK (3). 80% of oesophageal cancers develop in adults aged 60 or over (4), with a higher prevalence in males than females (5). The two main histology’s of oesophageal cancer are:

• Squamous cell carcinoma: This develops in the thin, flat cells of the mucosa, which line the oesophagus, and most commonly originates in the upper two-thirds of the oesophagus (3). Squamous cell carcinoma accounts for approximately a third of cases of oesophageal cancers in the UK(6).

• Adenocarcinoma: This develops in the mucus-producing glandular cells of the submucosa, which line the oesophagus, and most commonly originates in the lower two-thirds of the oesophagus, including at the junction with the stomach (3). Adenocarcinoma accounts for approximately two-thirds of oesophageal cancers in the UK (6).

Lifestyle factors, such as obesity, alcohol consumption and smoking, are considered risk factors in 90% of oesophageal cancers (4). Obesity, defined as having a body mass index (BMI) ≥30, is linked to 25% of oesophageal cancers in men, and 10% in women (4). Smoking increases the risk of developing both the squamous cell carcinoma and adenocarcinoma forms of oesophageal cancer. Specific smoking related activities (smoking a pipe, chewing tobacco, using snuff, or using betel quid [paan / pan]) are also associated with an increased risk of oesophageal cancer (4). Alcohol consumption of more than 14 units of alcohol per week is linked to an increased risk of developing squamous cell oesophageal cancer (4).

Another risk factor is Barrett’s oesophagus; a condition characterised by abnormalities developing in the cells lining the oesophagus and caused by long term indigestion (gastrooesophageal reflux disease [GORD]). Barrett’s oesophagus can cause an increased risk of developing oesophageal cancer – between 5-10% of people with this condition will subsequently develop oesophageal cancer within 10-20 years (7).

Symptoms of oesophageal cancer include difficulty in swallowing (dysphagia), persistent indigestion/heartburn, bringing up food soon after eating, loss of appetite and weight loss and pain/discomfort in the upper stomach/chest. The most common method for diagnosing oesophageal cancer is via a specific type of endoscopy, called gastroscopy(3,8). Occasionally

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

a barium swallow may be used, followed by an x-ray, to identify any oesophageal blockages which could potentially be a sign of cancer(3,8).

The treatment for oesophageal cancer is largely dependent on the stage at which the cancer is diagnosed. Locally advanced oesophageal cancers are either stage 2 or stage 3, and are defined as cancer that has spread into the tissues around the oesophagus, but not spread to other organs. For stage 1-3 oesophageal cancer, surgical resection of the affected section of the oesophagus (oesophagectomy) is the usual course of treatment (9). Advanced, metastatic cancers are stage 4. Stage 4 oesophageal cancer is unlikely to be cured, however chemotherapy and radiotherapy can slow the cancer spreading, and provide relief from other symptoms (9). According to CRUK, 30% of patients present with stage 4 Oesophageal cancer.

Survival rates for individuals with advanced oesophageal cancer are very low: most individuals with advanced oesophageal cancer live for only 3-12 months following their cancer diagnosis, and only 4% live for 5 years or more (10). Oesophageal cancer is now the fourth highest cancer killer in the UK, responsible for 4,000 deaths annually (11).

There is a high unmet need for first-line treatment options for patient with advanced oesophageal cancer, meaning that currently, treatment is restricted to doublet and triplet palliative chemotherapy options.

In England, the NICE guideline (12) on oesophageal and gastric cancer states that for locally advanced or metastatic oesophago-gastric cancer, treatment is restricted to palliative care only, as depicted in Figure 1 . This figure also includes the proposed positioning of pembrolizumab as a first-line treatment option for adults with locally advanced or metastatic oesophageal cancer who have not been previously treated.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 1: NICE pathway on locally advanced or metastatic oesophago-gastric cancer

==> picture [408 x 325] intentionally omitted <==

*patients with human epidermal growth factor receptor 2 (HER2)-positive metastatic adenocarcinoma of the stomach or gastro-oesophageal junction who have not received prior treatment for their metastatic disease and have tumours expressing high levels of HER2 as defined by a positive immunohistochemistry score of 3 (IHC3 positive).

• $ patients with performance status 0 to 2 and no significant comorbidities

B.1.4 Equality considerations

MSD does not envisage any equality issues with the use of pembrolizumab in combination with platinum-based chemotherapy for the first-line treatment of patients with locally advanced unresectable or metastatic carcinoma of the oesophagus or HER-2 negative gastroesophageal junction adenocarcinoma in adults.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B.2. Clinical Effectiveness

B.2.1 Identification and selection of relevant studies

See appendix D for full details of the process and methods used to identify and select the clinical evidence relevant to the technology being appraised.

B.2.2 List of relevant clinical effectiveness evidence

A systematic literature review (SLR) was conducted to identify clinical studies relevant to this submission. The SLR was designed to identify randomised controlled trials (RCTs) relating to the efficacy and safety of pembrolizumab in combination with chemotherapy and relevant comparators (as per final scope described in Table 1) in patients with locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or adenocarcinoma of the EGJ.

The SLR was originally conducted in July 2020 and updated on 22 December 2020. As the manufacturer of the technology being appraised, MSD is aware of all relevant RCTs for pembrolizumab in combination with chemotherapy in this indication.

In total, 7 RCTs were identified (13–19): six trials reporting evidence for the relevant comparators and one reporting evidence for pembrolizumab in combination with chemotherapy: KEYNOTE-590 (20).

Please refer to Table 3 for a summary of the evidence coming from the pivotal clinical trial KEYNOTE-590 (20–22).

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 3 Clinical effectiveness evidence

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

B.2.3.1 KEYNOTE – 590 trial overview (21)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Trial design

KEYNOTE – 590 (21) is a randomised, double-blind, placebo-controlled multi-centre phase III trial to evaluate the efficacy and safety of pembrolizumab in combination with cisplatin and 5- fluorouracil (5-FU) versus placebo in combination with cisplatin and 5-FU as first-line treatment in subjects with locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the esophagogastric junction (EGJ) (20).

The enrolment period was divided into 2 periods: Global Cohort and China Extension Study. In the Global Cohort, 749 subjects were enrolled in the study. Subjects were required to have an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1 and to provide a tumour sample adequate for central laboratory analysis of biomarkers that may have been predictive of response to pembrolizumab.

Participants were randomised (1:1) to one of the following treatment arms, with allocation stratified by geographic region, histology, and ECOG performance score:

• Treatment arm 1: combination of pembrolizumab 200 mg administered intravenously (IV) every 3 weeks (Q3W) and cisplatin 80 mg/m[2] IV Q3W and 5-FU 800 mg/m[2] /day continuous IV infusion on each of days 1 to 5 Q3W (total of 4000 mg/m[2] per 3-week cycle)

or

• Treatment arm 2: placebo administered IV Q3W combined with cisplatin 80 mg/m[2] IV Q3W and 5-FU 800 mg/m[2] /day continuous IV infusion on each of days 1 to 5 Q3W (total of 4000 mg/m[2 ] per 3-week cycle).

Treatment continued until confirmed progressed disease (PD), unacceptable adverse events (AEs), intercurrent illness that prevented further administration of treatment, investigator’s decision to withdraw the participant, participant withdrew consent, pregnancy of the participant, noncompliance with study treatment or procedure requirements, completion of 35 administrations (approximately 2 years) of treatment with pembrolizumab or achievement of a CR, or administrative reasons. No crossover from placebo arm to pembrolizumab arm was allowed.

106 eligible subjects from China were enrolled in the China Cohort which included subjects enrolled in China during the Global enrolment period as well as the China Extension Study

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

enrolment period. The China Extension study is identical to the Global Cohort with respect to key study characteristics (inclusion and exclusion criteria, study endpoints, primary and secondary objectives, study procedures). The Global Cohort and China Extension study were merged for the primary analyses, and henceforth will be referred to as the Global Study population.

A schematic of the trial design is provided below in Figure 2.

Figure 2 Schematic of KEYNOTE – 590 (21)

==> picture [413 x 424] intentionally omitted <==

5-FU=5-fluorouracil; ECOG PS=Eastern Cooperative Oncology Group Performance Score; EGJ=Esophagogastric junction; Q3W=Every 3 weeks; Q9W=Every 9 weeks; RECIST 1.1=Response Evaluation Criteria in Solid Tumours Version 1.1.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Eligibility criteria

Male and female subjects with locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the EGJ of at least 18 years of age were enrolled in this trial.

Inclusion criteria

• Has histologically- or cytologically-confirmed diagnosis of locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the esophagogastric junction (EGJ)

• Has measurable disease per RECIST 1.1 as determined by the local site investigator/radiology assessment

• Eastern Cooperative Group (ECOG) performance status of 0 to 1

• Can provide either a newly obtained or archival tissue sample for PD-L1 by immunohistochemistry analysis

• Female participants of childbearing potential must have a negative urine or serum pregnancy test within 72 hours prior to randomization and be willing to use an adequate method of contraception (e.g. abstinence, intrauterine device, diaphragm with spermicide, etc.) for the course of the study through 120 days after the last dose of study treatment and up to 180 days after last dose of cisplatin

• Male participants of childbearing potential must agree to use an adequate method of contraception (e.g. abstinence, vasectomy, male condom, etc.) starting with the first dose of study treatment through 120 days after the last dose of study treatment and up to 180 days after last dose of cisplatin, and refrain from donating sperm during this period

• Has adequate organ function.

Exclusion criteria

• Has locally advanced oesophageal carcinoma that is resectable or potentially curable with radiation therapy (as determined by local investigator).

• Has had previous therapy for advanced/metastatic adenocarcinoma or squamous cell cancer of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the EGJ.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

• Has had major surgery, open biopsy, or significant traumatic injury within 28 days prior to randomization, or anticipation of the need for major surgery during the course of study treatment.

• Has a known additional malignancy that is progressing or requires active treatment. Exceptions include early-stage cancers (carcinoma in situ or Stage 1) treated with curative intent, basal cell carcinoma of the skin, squamous cell carcinoma of the skin, in situ cervical cancer, in situ breast cancer that has undergone potentially curative therapy, and in situ or intramucosal pharyngeal cancer.

• Has known active central nervous system metastases and/or carcinomatous meningitis.

• Has an active autoimmune disease that has required systemic treatment in past 2 years.

• Has a diagnosis of immunodeficiency or is receiving chronic systemic steroid therapy (in dosing exceeding 10 mg daily of prednisone equivalent) or any other form of immunosuppressive therapy within 7 days prior to the first dose of study treatment, or has a history of organ transplant, including allogeneic stem cell transplant.

• Has a history of (non-infectious) pneumonitis that required steroids or has current pneumonitis, or has an active infection requiring systemic therapy.

• Is pregnant or breastfeeding or expecting to conceive or father children within the projected duration of the study, starting with the screening visit through 120 days after the last dose of study medication and up to 180 days after last dose of cisplatin.

• Has received prior therapy with an anti-programmed cell death protein-1 (anti-PD-1), antiPD-L1, or anti-PD-L2 agent or with an agent directed to another co-inhibitory T-cell receptor or has previously participated in a pembrolizumab (MK-3475) clinical trial.

• Has severe hypersensitivity (≥ Grade 3) to any study treatment (pembrolizumab, cisplatin, or 5-FU) and/or any of its excipients.

• Has a known history of active tuberculosis (TB; Mycobacterium tuberculosis) or human immunodeficiency virus (HIV) infection.

• Has known history of or is positive for hepatitis B or hepatitis C.

• Has received a live vaccine within 30 days prior to the first dose of study treatment.

• Has had radiotherapy within 14 days of randomization. Participants who received radiotherapy >14 days prior to randomization must have completely recovered from any radiotherapy-related AEs/toxicities.

The full list of inclusion/exclusion criteria are provided in the study protocol (21).

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Settings and Locations where data were collected

The KEYNOTE-590 study was conducted at 168 centres in 26 countries: Argentina, Australia, Brazil, Canada, Chile, China, Colombia, Costa Rica, Denmark, France, Germany, Guatemala, Hong Kong, Japan, South Korea, Malaysia, Peru, Romania, Russia, South Africa, Spain, Taiwan, Thailand, Turkey, United Kingdom, and United States. 22 subjects from 3 UK centres participated in KEYNOTE 590 trial.

Trial drugs and concomitant medication

Trial drugs

Study medications used in this trial are outlined below.

Table 4 Trial Treatments

Trial treatment for cycle 1 should have begun within 3 days of randomisation. All trial treatments were administered on an outpatient basis. For 5-FU continuous infusion, use of a portable infusion pump is preferred; however, hospitalisation is acceptable if that is the standard procedure for the local site.

Study treatment in both arms begun on Day 1 of each 3-week dosing cycle.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Treatments were administered in the following order:

• Pembrolizumab or placebo infusion was administered first, followed by the cisplatin and 5- FU infusions. Administration of chemotherapy should have followed 1 to 2 days after pembrolizumab/placebo as needed per local standard of care. Treatment continued with pembrolizumab plus chemotherapy or placebo plus chemotherapy until documented confirmed PD, unacceptable AE(s), intercurrent illness that prevented further administration of treatment, investigator’s decision to discontinue treatment, subject withdrew consent, pregnancy of the subject, noncompliance with trial treatment or procedure requirements, subject received 35 administrations (approximately 2 years) of study medication, or administrative reasons requiring cessation of treatment. Regardless of clinical benefit, subjects only received 35 administrations (approximately 2 years) with pembrolizumab. Pembrolizumab 200-mg fixed dose was administered as a 30-minute IV infusion Q3W

• Placebo was normal saline solution prepared by the local pharmacist. Placebo was dosed and administered by blinded qualified trial site personnel in the same manner as pembrolizumab.

• Cisplatin 80 mg/m[2] was administered as a 60- or 120-minute IV infusion (or per site’s standard practice) Q3W on Day 1 of each treatment cycle and after pembrolizumab/placebo administration. Duration of cisplatin treatment was capped at 6 doses.

• 5-FU was administered as a continuous IV infusion of 800 mg/m[2] /day on each of Days 1 to 5 Q3W or per local standard for 5-FU administration duration as long as total dose of 4000 mg/m[2] per 3-week cycle was followed. 5-FU was administered after pembrolizumab/placebo administration. Duration of 5-FU treatment did not exceed 35 cycles.

Trial blinding

A double-blinding technique was used. Pembrolizumab and placebo were prepared and/or dispensed in a blinded fashion by an unblinded pharmacist or qualified trial site personnel. The subject and the investigator who was involved in the treatment or clinical evaluation of the subjects were unaware of the group assignments. The administration of pembrolizumab or placebo treatment was blinded to the subject, study site personnel, and sponsor personnel.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Acceptable Concomitant Medications

All treatments that the investigator considered necessary for a subject’s welfare were permitted to be administered at the discretion of the investigator in keeping with the community standards of medical care.

Prohibited concomitant medication

Subjects were prohibited from receiving the following therapies during screening to the end of treatment of this trial:

• Antineoplastic systemic chemotherapy or biological therapy

• Immunotherapy not specified in this protocol

• Chemotherapy not specified in this protocol

• Investigational agents other than pembrolizumab

• Radiation therapy

• Live vaccines within 30 days prior to the first dose of trial treatment and while participating in the trial.

• Systemic glucocorticoids for any purpose other than to modulate symptoms from an AE that is suspected to have an immunologic aetiology or for cisplatin or 5-FU supportive care. The use of physiologic doses of corticosteroids were approved after consultation with the Sponsor.

• Brivudine, sorivudine analogues, and other inhibitors of the enzyme dihydropyrimidine dehydrogenase were not to be administered with 5-FU therapy.

Subjects who, in the assessment of the investigator, required the use of any of the aforementioned treatments for clinical management were to be removed from the trial. Subjects may receive other medications that the investigator deems to be medically necessary.

Concomitant medications which were permitted to be used with caution

• Cimetidine, metronidazole, and interferons may increase levels of 5-FU.

• Phenytoin should not be started with cisplatin therapy.

• Subjects who were taking phenytoin in conjunction with 5-FU were to be examined regularly due to a potential elevation in phenytoin plasma levels.

• Hepatotoxic effects (i.e., rise in alkaline phosphatase, transaminase, or bilirubin levels) are commonly observed under the treatment with 5-FU and levamisole.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

• For 5-FU and cisplatin, protocol specified to refer to the product labels or local standards of care for further information regarding concomitant medications to be used with caution.

Subjects who, following the assessment by the investigator, required additional anti-cancer treatments were discontinued from study treatment but continued survival follow-up. Subjects who, following the assessment by the investigator, required any other prohibited medications for the assigned study treatment for long-term clinical management, were discontinued from trial treatment but continued disease assessments and survival follow-up.

The exclusion criteria describe other medications or vaccinations that were specifically prohibited in KEYNOTE-590.

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

primary outcome

KEYNOTE-590 (21) objectives were pre-specified. In male and female adult subjects (≥18 years of age) with locally advanced/metastatic RCC, the objectives were as follows:

Primary objective(s)

1. To compare overall survival (OS) between treatment arms in subjects with oesophageal squamous cell carcinoma (ESCC) whose tumours are PD-L1 biomarker-positive (CPS≥10).

2. To compare OS between treatment arms in subjects with ESCC.

3. To compare OS between treatment arms in subjects whose tumours are PD-L1 biomarkerpositive (CPS ≥10).

4. To compare OS between treatment arms in all subjects.

5. To compare progression free survival (PFS) per RECIST 1.1, as determined by investigator, in subjects with ESCC.

6. To compare PFS per RECIST 1.1, as determined by investigator, between treatment arms in subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10).

7. To compare PFS per RECIST 1.1, as determined by investigator, between treatment arms in all subjects.

OS was defined as the time from randomisation to death due to any cause. Subjects without documented death at the time of the final analysis were censored at the date of the last followup.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

PFS was defined as the time from randomisation to the first documented disease progression per RECIST 1.1 based on BICR or death due to any cause, whichever occurs first.

Secondary objective(s)

1. To evaluate objective response rate (ORR) per RECIST 1.1, as determined by investigator between treatment arms in all subjects.

2. Evaluate ORR per RECIST 1.1, as determined by investigator, between treatment arms in subjects with ESCC whose tumours are PD-L1 biomarker-positive (CPS ≥10), in subjects with ESCC, and in subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10).

3. Evaluate DOR per RECIST 1.1, as determined by investigator, between treatment arms in all subjects, in subjects with ESCC whose tumours are PD-L1 biomarkerpositive (CPS ≥10), in subjects with ESCC, and in subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10).

4. Evaluate the safety and tolerability profile.

5. To evaluate changes from baseline in health-related quality of life using the European Organization for the Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire C30 (QLQ-C30) and the EORTC Quality Of Life Questionnaire Oesophageal Module (QLQ-OES18) in all subjects, in subjects with ESCC whose tumours are PD-L1 biomarker-positive (CPS ≥10), in subjects with ESCC, and in subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10), treated with pembrolizumab plus chemotherapy compared with placebo plus chemotherapy.

ORR was defined as the proportion of the subjects in the analysis population who have a complete response (CR) or partial response (PR).

For subjects who demonstrated CR or PR, DOR is defined as the time from first documented evidence of CR or PR until disease progression per RECIST 1.1 based on assessments by BICR or death due to any cause, whichever occurs first.

Exploratory Objectives

To characterise patient reported outcome (PRO) utilities using EuroQoL 5-dimension 5-level (EQ- 5D-5L) questionnaire in all subjects, in subjects with SCC whose tumours are PD-L1 biomarker-positive (CPS ≥10), in SCC subjects, and in subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10) treated with pembrolizumab plus chemotherapy compared with placebo plus chemotherapy.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

1. Evaluate PFS per irRECIST as determined by investigator between treatment arms in subjects with ESCC whose tumours are PD-L1 biomarker-positive (CPS ≥10), in ESCC subjects, in subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10), and in all subjects.

2. To identify molecular (genomic, metabolic, and/or proteomic) biomarkers that may be indicative of clinical response/resistance, safety, pharmacodynamic activity, and/or the mechanism of action of pembrolizumab and other treatments. This could include the evaluation of microsatellite instability (MSI), whole exome sequencing (WES), and/or gene expression profiling (GEP) in available tumour tissue.

B 2.3.2 Comparative summary of the trial methodology

A summary of the trial methodology is present below in Table 5.

Table 5: Summary of trial methodology

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B 2.3.3 KEYNOTE-590: Participants baseline characteristics

Baseline characteristics are summarised in Table 6. The baseline demographic and disease characteristics of participants for the two groups were generally well balanced and representative of a patient population with metastatic oesophageal cancer. Most participants were male (83.4%) and had squamous cell carcinoma (73.2%). Among participants who either achieved an on-study CR or PR or were continuing in the study without PD per RECIST 1.1, 112 participants were evaluable for MSI status; none were MSI-H.

Table 6 : Subject Characteristics (ITT Population) – KEYNOTE-590 (22)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

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

This section reports the relevant statistical methodology of KEYNOTE-590 (21,22).

Table 7: Statistical analysis plan summary

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Abbreviations: n, sample size; IV, intravenously; Q3W, every 3 weeks; BID, twice daily; QD, daily; PFS, progression free survival; OS, overall survival; ORR, objective response rate; CI, confidence interval, ESCC, oesophageal squamous cell carcinoma.

Discontinuation of Treatment

Subjects were permitted to discontinue treatment at any time for any reason or be dropped from study treatment at the discretion of the investigator should any untoward effect occur. In addition, a subject may be discontinued from study treatment by the investigator or the study

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

sponsor if study treatment is inappropriate, the trial plan is violated, or for administrative and/or other safety reasons.

In the event of any of the following reasons, a subject must have been discontinued from the study treatment, but would continue to be monitored in the trial:

• The subject or subject’s legally acceptable representative requests to discontinue treatment.

• Confirmed radiographic PD as outlined in the trial protocol

• Unacceptable adverse experiences as described in the trial protocol

• Any progression or recurrence of any malignancy, or any occurrence of another malignancy that requires active treatment

• Intercurrent illness other than another malignancy as noted above that prevents further administration of treatment

• Recurrent Grade 2 pneumonitis

• A confirmed positive serum pregnancy test

• Investigator decision to discontinue treatment

• Completion of 35 treatments (approximately 2 years) with pembrolizumab/placebo

Discontinuation of Trial Treatment after CR:

Discontinuation of treatment could be considered for subjects who had attained a confirmed CR and had been treated for at least 8 cycles (at least 24 weeks), receiving at least 2 cycles of study treatment beyond the date when the initial CR was declared.

For subjects who were discontinued from treatment but continued to be monitored in the trial, all visits and procedures, as outlined in the study protocol, were to be completed. Discontinuation from treatment is considered “permanent.” Once a subject was discontinued, he/she was not allowed to restart treatment.

B 2.4.1 Statistical methods used to compare groups for primary and secondary outcomes and approach to missing data

The statistical methods and analysis strategy for the primary and secondary efficacy endpoints are summarised in the Table 8 below.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 8: KEYNOTE-590– Analysis strategy for key efficacy endpoints (21)

The non-parametric Kaplan Meier (KM) method was used to estimate the PFS and OS rates over time in each treatment group. The hypotheses of treatment differences in PFS and OS were assessed by the stratified log-rank test. A stratified Cox proportional hazard model with Efron’s method of tie handling was used to estimate the magnitude of the treatment difference (HR) between the treatment groups. The stratification factors used for the randomisation were applied to both the stratified log-rank test and the stratified Cox model.

Since PD was assessed periodically, PD could occur any time in the time interval between the last assessment where PD was not documented and the assessment when PD was documented. For the primary analysis, for the subjects who have PD, the true date of PD was approximated by the date of the first assessment at which PD was objectively documented per RECIST 1.1 by investigator. Death was always considered as a confirmed PD event. Subjects who did not experience a PFS event were censored at the last disease assessment.

In order to evaluate the robustness of the PFS endpoint per RECIST 1.1 by investigator, two sensitivity analyses with different sets of censoring rules were performed for comparison of PFS per RECIST 1.1 by investigator. The first sensitivity analysis followed the intention-totreat principle. That is, PDs/deaths were counted as events regardless of missed study visits or initiation of new anti-cancer therapy. The second sensitivity analysis considered discontinuation of treatment due to reasons other than complete response or initiation of new anti-cancer treatment, whichever occurred later, to be a PD event for subjects without documented PD or death. If a subject met multiple criteria for censoring, the censoring criterion Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

that occurred earliest was applied. The censoring rules for primary and sensitivity analyses are summarised in Table 9.

Subjects in the placebo plus chemotherapy arm were expected to discontinue treatment earlier compared with subjects in the pembrolizumab plus chemotherapy arm and may have switched to another anti PD-1 treatment following the verification of PD by the central imaging vendor. The study protocol specified that based on an examination of the appropriateness of the data to the assumptions required by recognised methods, exploratory analyses to adjust for the effect of crossover to other PD-1 therapies on OS may be performed based on recognised methods (e.g., the Rank Preserving Structural Failure Time model proposed by Robins and Tsiatis (25), 2-stage model, etc.,).

Table 9: Censoring Rules for Primary and Sensitivity Analyses of PFS

The proportional hazards assumption on PFS was examined using both graphical and analytical methods if warranted.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

IA was performed after enrolment completion, when a minimum of 460 PFS events had accrued in ESCC and 391 deaths have occurred in ESCC subjects and all participants were followed for at least 13 months after randomisation.

Multiplicity strategy for PFS, OS and ORR

The study used the graphical method of Maurer and Bretz (24) to provide strong multiplicity control for multiple hypotheses as well as interim analysis. According to this approach, study hypotheses might be tested more than once, and when a particular null hypothesis is rejected, the α allocated to that hypothesis can be reallocated to other hypothesis tests. Figure 3 shows the initial 1-sided α allocation for each hypothesis in the ellipse representing the hypothesis. The weights for re-allocation from each hypothesis to the others are shown in the boxes on the lines connecting hypotheses. The boundaries provided in this section are calculated based on the estimated number of events at each analysis, and the actual boundaries were determined from the actual number of events observed at the time of the analyses, using the spending functions specified. An assumption of 38% prevalence in ESCC subjects with PDL1 CPS ≥10, 51% prevalence in subjects with PD-L1 CPS ≥10, and 73% prevalence in ESCC subjects was made in the calculations. Details of multiplicity strategy for the primary and key secondary endpoints are provided in Appendix D.

Figure 3 . Maurer and Bretz multiplicity strategy approach used for hypothesis testing in KEYNOTE-590

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

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Abbreviations: ORR, objective response rate; OS, overall survival; PFS, progression-free survival, ESCC, oesophageal squamous cell carcinoma

B 2.4.2 Subgroup Analyses

The estimate of the between-group treatment effect (with a nominal 95% CI) for the dual primary endpoints were estimated and plotted within each category considered.

Please refer to Section 2.7 for details on statistical tests used in the primary analysis of the subgroups and results.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B 2.4.3 Statistical analysis

Table 10. Summary of statistical analyses

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B 2.4.5 Participant flow in the relevant randomised controlled trials

Details of the participant flow in KEYNOTE-590 (22) are provided in Appendix D.

B.2.5 Quality assessment of the relevant clinical effectiveness evidence

2.5.1 & 2 Summary of quality assessment

Quality Assessment of KEYNOTE-590 (21,22) was conducted using the Cochrane risk of bias tool. Based on this analysis, the study was determined to be at ‘low risk’ across all six key domains. The complete quality assessment is included in Appendix D1.4. A tabulated summary of the quality assessment results is presented below in Table 11.

Table 11. Quality assessment results for KEYNOTE-590 (21,22)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B.2.6 Clinical effectiveness results of the relevant trials

B 2.6.1. KEYNOTE-590 results

Early results are presented from the KEYNOTE-590 (22) study, based on the interim analysis (IA), which had a data cut-off date of 02 July 2020. Part of this study was conducted during the COVID-19 pandemic. The trial SOPs for study conduct, monitoring, and oversight during the pandemic were continuously followed and a risk-based approach to assess and mitigate impact on study conduct was employed. Efficacy analyses were conducted using the intentionto-treat (ITT) population. The study enrolment period was divided into 2 periods: Global Cohort and China Extension Study. The Global Cohort and China extension study were merged for the primary analyses and henceforth referred to as the Global Study population.

Interim analysis – data cut-off 02 July 2020

IA occurred after meeting the cut-off criteria of accruing at least 448 PFS events and with a minimum follow up of 13 months.

A total of 1020 participants were screened and 749 were randomised across 168 global study sites in 26 countries. 22 patients were recruited across 3 sites in the UK. A total of 740 randomised participants received at least 1 dose of study medication (pembrolizumab plus chemotherapy: 370; chemotherapy: 370). The participant flow and subject disposition from KEYNOTE-590 (22) are provided in Appendix D.

Primary efficacy endpoints: clinical outcome measures included within the health economic model

At IA, KEYNOTE-590 (22) had met both of its primary endpoints of PFS and OS, with p-values below the prespecified boundary for statistical significance of 0.02477 and 0.01421, respectively. As of the data cut-off date (02 July 2020) for IA, the median duration of follow up was 12.6 months (0.1 to 33.6 months) in the pembrolizumab plus chemotherapy group and 9.8 months (0.1 to 33.6 months) in the chemotherapy group.

There are different populations covered by the co-primary endpoints in the KEYNOTE-590 study:

• participants with ESCC whose tumours are PD-L1 biomarker-positive (CPS ≥10)

• • participants with ESCC

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

• participants whose tumours are PD-L1 biomarker-positive (CPS ≥10)

• all participants

The focus of this submission (base case) is the population covering all study participants. In addition, cost-effectiveness analyses are presented for the sub-populations of participants whose tumours are PD-L1 biomarker-positive (CPS ≥10). The clinical efficacy and safety results relating to these sub-populations are also presented in the main body of the submission, with the results for the remaining sub-population assessed as a co-primary endpoint (participants with ESCC, participants with ESCC whose tumours are PD-L1 biomarker-positive (CPS ≥10) covered within the appendices).

OS: IA 02 July 2020 data-cut

All participants (ITT)

Pembrolizumab in combination with chemotherapy demonstrated a statistically significant and clinically meaningful improvement in OS compared with chemotherapy for the 1L treatment of patients with locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the EJG. The OS HR of 0.73 (95% CI: 0.62, 0.0.86; p<0.0001), represents a 27% reduction in the risk of death for participants in the pembrolizumab plus chemotherapy group compared with the chemotherapy group (Table 12).

The median OS was 12.4 months (95% CI: 10.5, 14.0) for the pembrolizumab plus chemotherapy group and 9.8 months (95% CI: 8.8, 10.8) for the chemotherapy group. OS results have demonstrated that pembrolizumab plus chemotherapy is superior to chemotherapy in all participants (Table 12). The KM curves for the OS separated early and remained separated throughout the evaluation period in favour of pembrolizumab plus chemotherapy (Figure 4).

The OS rates were higher in the pembrolizumab plus chemotherapy group at 12 (50.6% versus 9.4%) and 24 months (27.7% vs 16.3%) compared with the chemotherapy group (Table 13).

Table 12. Analysis of OS (ITT): IA July 2020 data cut

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 13. Summary of OS Rate Over Time (ITT): IA July 2020 data cut

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 4: KM Estimates of OS (ITT); IA July 2020 data-cut

==> picture [392 x 277] intentionally omitted <==

PD-L1 biomarker-positive (CPS ≥10)

Pembrolizumab plus chemotherapy provided a statistically significant and clinically meaningful improvement in OS compared with chemotherapy. The OS HR was 0.62 (95% CI: 0.49, 0.78; p<0.0001) represents a 38% reduction in the risk of death (Table 14). The median OS was 13.5 months (95% CI: 11.1, 15.6) for the pembrolizumab plus chemotherapy group and 9.4 months (95% CI: 8.0, 10.7) for the chemotherapy group (Figure 5). OS results have demonstrated that pembrolizumab plus chemotherapy is superior to chemotherapy in participants whose tumours express PD-L1 CPS ≥10.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 14. Analysis of Overall Survival (Subjects with PD-L1 CPS >= 10, ITT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 15. Overall Survival Rate (Subjects with PD-L1 CPS ≥ 10, ITT Population)

Figure 5. Kaplan-Meier Estimates of Overall Survival (Subjects with PD-L1 CPS >= 10, ITT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

==> picture [452 x 328] intentionally omitted <==

PFS: IA 02 July 2020 data-cut

All participants (ITT)

Pembrolizumab in combination with chemotherapy demonstrated a statistically significant and clinically meaningful improvement in PFS based on investigator assessment per RECIST 1.1 compared with chemotherapy for the 1L treatment of patients locally advanced

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the esophagogastric junction. The PFS HR of 0.65 (95% CI: 0.55, 0.76; p <0.0001) represents a 35% reduction in the risk of progression or death for participants in the pembrolizumab plus chemotherapy group compared with the chemotherapy group (Table 16).

The median PFS was 6.3 months (95% CI: 6.2, 6.9) for the pembrolizumab plus chemotherapy group and 5.8 months (95% CI: 5.0, 6.0) for the chemotherapy group. PFS results have demonstrated that pembrolizumab plus chemotherapy is superior to chemotherapy in all participants.

In the pembrolizumab plus chemotherapy group, the PFS rates at 12 and 18 months were higher compared with the chemotherapy group (

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 17). The KM curves separated early and remained separated for the duration of the evaluation period (Figure 6).

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 16. Analysis of Progression-Free Survival Based on Investigator Assessment per RECIST 1.1 (Primary Censoring Rule): IA July 2020 data cut

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 17. Summary of PFS Rate Over Time Based on Investigator Assessment per RECIST 1.1: IA July 2020 data cut

Figure 6. KM Estimates of PFS (Primary Censoring Rule) Based on BICR Assessment per RECIST 1.1 (ITT): IA July 2020 data-cut

==> picture [411 x 279] intentionally omitted <==

PD-L1 biomarker-positive (CPS ≥10) (ITT)

Pembrolizumab plus chemotherapy provided a statistically significant and clinically meaningful improvement in PFS based on investigator assessment per RECIST 1.1 compared with chemotherapy. The PFS HR was 0.51 (95% CI: 0.41, 0.65; p<0.0001) in favour of pembrolizumab plus chemotherapy compared to chemotherapy, representing a 49% reduction in the risk of death and disease progression (Table 18). The median PFS was 7.5 months (95% CI: 6.2, 8.2) for the pembrolizumab plus chemotherapy group and 5.5 months (95% CI:

4.3, 6.0) for the chemotherapy group. PFS results have demonstrated that pembrolizumab Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

plus chemotherapy is superior to chemotherapy in participants whose tumours express PDL1 CPS ≥10.

Table 18. Analysis of Progression-Free Survival Based on Investigator Assessment per RECIST 1.1 primary Censoring Rule) (Subjects with PD-L1 CPS >= 10, ITT Population)

Table 19. Summary of PFS Rate Over Time Based on Investigator Assessment per RECIST 1.1 (Subjects with PD-L1 CPS ≥ 10, ITT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 7. Kaplan-Meier Estimates of Progression-Free Survival Based on Investigator Assessment per RECIST 1.1 (Primary Censoring Rule) (Subjects with PD-L1 CPS >= 10, ITT Population)

==> picture [452 x 306] intentionally omitted <==

Secondary Efficacy Endpoints

The results of key secondary endpoints ORR and DOR in all participants, subjects with ESCC and subjects whose tumours are PD-L1 biomarker-positive [CPS ≥10]) are presented below. The results of the key secondary endpoints in the subgroups and subpopulations (subjects with ESCC, subjects with ESCC whose tumours are PD-L1 biomarker-positive [CPS ≥10]; are presented in Appendix L.

ORR: IA 02 July 2020 data-cut

All participants (ITT)

Pembrolizumab with chemotherapy demonstrated a statistically significant and clinically meaningful improvement in ORR compared with SOC for the 1L treatment of participants with untreated, unresectable locally advanced or metastatic oesophageal cancer or GEJ adenocarcinoma. The confirmed ORR (based on investigator assessment per RECIST 1.1) was substantially higher with pembrolizumab plus chemotherapy than SOC (45.0% vs 29.3%),

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

reflecting a clinically meaningful and statistically significant 15.8% difference ( p <0.0001) (Table 20).

A total of 24 (6.4%) participants treated with pembrolizumab plus chemotherapy achieved a CR (investigator-assessed), while 9 (2.4%) participants treated with SOC achieved a CR (investigator-assessed) (Table 21). Participants receiving pembrolizumab plus chemotherapy were more likely to experience a reduction in tumour size than those receiving SOC alone.

Table 20. Analysis of Objective Response with Confirmation Based on Investigator Assessment per RECIST 1.1 (ITT); IA July 2020 data-cut

Table 21. Summary of Best Overall Response Based on Investigator Assessment per RECIST 1.1 with Confirmation (ITT); IA July 2020 data-cut

PD-L1 biomarker-positive (CPS ≥10)

In participants whose tumours express PD-L1 CPS ≥10, the confirmed ORR based on

investigator assessment per RESCIST 1.1 was substantially higher with pembrolizumab plus Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

chemotherapy than chemotherapy (51.1% v 26.9%), reflecting a clinically meaningful 24.0% improvement relative to chemotherapy (Table 22).

Table 22. Analysis of Objective Response with Confirmation Based on Investigator Assessment per RECIST 1.1 (Subjects whose tumours are PD-L1 biomarker-positive (CPS ≥10), ITT Population)

Table 23. Summary of Best Overall Response Based on Investigator Assessment per RECIST 1.1 with Confirmation (Subjects with PD-L1 CPS ≥ 10, ITT Population)

DOR: IA July 2020 data-cut

All participants

The responses in the pembrolizumab plus chemotherapy group for the untreated,

unresectable locally advanced or metastatic oesophageal cancer or GEJ adenocarcinoma

were more durable compared with the chemotherapy group (Appendix L), based on data from Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

the July 2020 data cut-off. The median DOR in the pembrolizumab plus chemotherapy group was numerically longer (8.3 months; range: (1.2+ - 31.0+) chemotherapy group (6.0 months; 1.5+ to 25.0+) (Appendix L) In the pembrolizumab plus chemotherapy group, a 3-fold higher percentage of participants had extended responses for ≥24 months by KM estimation (18.1% versus 6.1%) (Appendix L).

PD-L1 biomarker-positive (CPS ≥10)

In participants whose tumours express PD-L1 CPS ≥10, responses were more durable in the pembrolizumab plus chemotherapy group compared with the chemotherapy group. Median DOR in the pembrolizumab plus chemotherapy group was numerically longer (10.4 months; range: 1.9 to 28.9+) compared with the chemotherapy group (5.6 months; 1.5+ to 25.0+) (Appendix L). The median time to response was the same in the pembrolizumab plus chemotherapy group and chemotherapy group.

Exploratory objectives

PRO Compliance Rate and Completion Rate – IA July 2020 data-cut

PROs were analysed in the PRO FAS population (n=730), which consisted of participants who received at least 1 dose of study medication and completed at least 1 PRO assessment. All PROs for both arms were performed at Cycles 1 to 9. After Cycle 9 (Week 24), PROs were administered every 3 cycles. Compliance rates for all the PROs were high at baseline and Week 18 in both treatment groups (Appendix L). As expected, completion rates generally decreased at each time point as more participants discontinued the study treatment.

EQ-5D-VAS Health Status/Quality of Life change from baseline to Week 30: IA July 2020 data-cut

All participants

In the PRO FAS population, the compliance and completion for the EQ-5D at baseline was 98.1% for both the pembrolizumab plus chemotherapy and the chemotherapy groups. Compliance at Week 18 was 90.4% for the pembrolizumab plus chemotherapy group and 92.7% for the chemotherapy group. Completion at Week 18 was 61.6% for the pembrolizumab plus chemotherapy group and 56.7% for the chemotherapy group. There were no clinically meaningful differences from baseline to week 18 in the EQ-5D-VAS health status/QoL score for participants in both the pembrolizumab plus chemotherapy group and the chemotherapy group based on data from the July 2020 data-cut (Table 24). Changes from baseline to week

18 were generally similar between the treatment groups at week 18 (Table 24).

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 24. Analysis of Change from Baseline in EQ-5D VAS to Week 18; IA July 2020 data-cut

PD-L1 biomarker-positive (CPS ≥10) (ITT)

Table 25. Analysis of Change from Baseline in EQ-5D VAS to Week 18 (Subjects with PD-L1 CPS ≥ 10, FAS Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B.2.7 Subgroup analysis

B 2.7.1. Subgroup analyses carried out

Subgroup analyses were pre-specified in the KEYNOTE-590 study protocol to determine whether the treatment effect was consistent across subgroups. The estimate of the betweengroup treatment effect (with a nominal 95% CI) for the primary endpoints were estimated and plotted within each category of the following classification variables:

• Stratification factor: histology (adenocarcinoma versus squamous cell carcinoma)

• Stratification factor: geographic region (Asia versus Rest of World)

• Stratification factor: ECOG performance scale (0 versus 1)

• Disease status (locally advanced versus metastatic)

• Age category (< 65 versus ≥ 65)

• Sex (male versus female)

The results of subgroup analyses for all sub-populations are presented in Appendix E.

OS by Subgroup: IA: July 2020 data-cut

The improvement in OS for pembrolizumab plus chemotherapy compared with SOC in all subjects (based on the July 2020 data-cut) was consistent across all subgroups and subpopulations analysed (Appendix E). Subgroup analyses of OS for the sub-populations covered by the other co-primary endpoints of KEYNOTE-590 (subjects with ESCC whose tumours are PD-L1 biomarker-positive [CPS ≥10]; subjects with ESCC; subjects whose tumours are PDL1 biomarker-positive [CPS ≥10]) are presented in Appendix E.

PFS by Subgroup: IA: July 2020 data-cut

The improvement in PFS for pembrolizumab plus chemotherapy compared with chemotherapy (based on the July 2020 data-cut) was observed across all subgroups and subpopulations analysed (Appendix E.) Subgroup analyses of PFS for the sub-populations covered by the other co-primary endpoints of KEYNOTE-590 (subjects with ESCC whose tumours are PD-L1 biomarker-positive [CPS ≥10]) are presented in Appendix E.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B.2.8 Meta-analysis

Based on the SLR results, there is only one phase III randomised, controlled trial of pembrolizumab plus chemotherapy compared with a relevant comparator, in our specific population of interest (patients with patients with oesophageal cancer): KEYNOTE-590 (20– 22). Therefore, it was not possible to conduct a meta-analysis.

B.2.9 Indirect and mixed treatment comparisons

Please refer to Appendix D for full details of the methodology used for the SLR.

Given that pembrolizumab in combination with cisplatin and 5-FU have only been compared head-to-head to placebo plus cisplatin and 5-FU, an indirect treatment comparison is needed to obtain estimates of the relative efficacy and safety of pembrolizumab versus other regimens relevant to the UK context, including capecitabine plus cisplatin and epirubicin with cisplatin and 5-FU; however, because these interventions have generally only been evaluated in noncomparative studies, performing an anchored network meta-analysis (NMA) of pembrolizumab plus chemotherapy versus competing interventions was not feasible.

Further details are provided in the following sections:

B 2.9.1 Summary of trials identified following systematic literature review (SLR)

Trials which are relevant for the generation of comparative effectiveness data were identified through the SLR and are presented in Table 26. An overview of the patients’ characteristics in all included studies is provided in Table 27, Table 28 and Table 29.

Table 26. Summary of the trials of relevance identified through the SLR

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 27: Treatment characteristics of included trials

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 28. Patient characteristics of trials identified through the SLR (age, sex, region, performance score)

Notes: * Only mean age was reported and presented here; ** Ross 2002 had a mixed cancer population of esophageal, EGJ, and gastric cancer patients with no subgroup characteristics by cancer type; patient characteristics of the overall population is presented. Abbreviations: 5-FU, fluorouracil; ECOG, Eastern Cooperative Oncology Group.

Table 29. Patient characteristics of trials identified through the SLR (disease stage, cancer type, histology)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Note: * Ross 2002 had a mixed cancer population of esophageal, EGJ, and gastric cancer patients with no subgroup characteristics by cancer type; patient characteristics of the overall population is presented. Abbreviations: 5-FU, fluorouracil; EGJ, esophagogastric junction

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Feasibility assessment

A feasibility assessment for an NMA is a well-established process and typically involves determining whether the evidence from RCTs for the interventions of interest form a network, followed by an assessment of the differences in study, treatment, patient, and outcomes characteristics within or between treatment comparisons. As illustrated in Error! Reference source not found., there was a lack of network connectivity between the three studies included in feasibility assessment. Hence the NMA feasibility assessment process was adapted to the context of an unanchored MAIC as summarised in Appendix D. In the context of an MAIC, the validity of estimates depends on the degree of overlap in the study populations and the extent that it is possible to up or down-weight patients to achieve an appropriate match to the external trial. Beyond exploring the between-study differences based on the study-level data, descriptive statistics based on the index trial can be performed, and the extent of the overlap in the study populations can be assessed based on diagnostic criteria of the MAIC.

Studies included in feasibility assessment

Of the seven trials identified in the SLR and deemed relevant to the UK context, four were not considered for the feasibility assessment. Three of these trials Lorenzen 2009 (15), POWER (16,17), and Wang 2017 (19) were excluded for using cisplatin with 5-FU, which is already captured in the population of interest in the index trial, KEYNOTE-590. One additional trial Ross 2002 (18) was excluded due to a lack of reported patient characteristics for the population of interest. As a result, three trials were included in the feasibility assessment: KEYNOTE-590, an RCT comparing pembrolizumab plus cisplatin and 5-FU to placebo plus cisplatin and 5-FU and two trials that administered capecitabine plus cisplatin 2008 and Lee et al 2015 (13,14). A summary of each included study is presented in Table 30.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 30. Summary of trials included in the UK feasibility assessment

Abbreviations : 5-FU, fluorouracil; CR, complete response; ECOG, Eastern Cooperative Oncology Group; EGJ, esophagogastric junction; HRQoL, health-related quality of life; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; PR, partial response; RECIST, Response Evaluation Criteria in Solid Tumours; TTP, time-toprogression; WHO, World Health Organization.

The network of evidence identified in the SLR is depicted in Figure 8 – as can be seen, it was not possible to form a connected network.

Figure 8. Network diagram of studies identified through SLR

==> picture [425 x 150] intentionally omitted <==

Recommendation

Upon reviewing the inclusion criteria and patient characteristics for the three trials included in the feasibility assessment, key differences were observed between the patient populations of Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

KEYNOTE-590 (21,22) and the two external trials as described in Table 30. Unlike KEYNOTE590, which was carried out in multiple centres internationally, Lee 2008 (13) and Lee 2015 (14) were both conducted in a single centre in South Korea. Lee 2008 and Lee 2015 only enrolled ESCC patients whereas KEYNOTE-590 enrolled ESCC, oesophageal adenocarcinoma, and EGJ Siewert type I adenocarcinoma patients. By virtue of these differences, non-Asian patients as well as those Asian patients without ESCC will automatically receive a weight of 0 within an MAIC, which will likely to lead to large reductions in effective sample size (ESS) and a high degree of uncertainty around the estimated treatment effects due to these being heavily influenced by relatively few individuals. In addition, patients with locally advanced disease would also receive a null weight in a comparison with Lee 2015 (14), further reducing the ESS and increasing the associated uncertainty.

A major source of potential bias in MAIC estimates stems from imbalances in prognostic factors/effect modifiers post-matching. As IPD are not available from Lee 2008 (13) and Lee 2015 (14), it is only possible to adjust for those factors reported in the associated publications. A targeted literature review was performed to identify studies of prognostic factors in oesophageal cancer (see details in Appendix D). Although some of the prognostic factors identified in the review are reported in Lee 2008 (13) and Lee 2015 (14), other factors such as tumour size/length and weight loss/BMI are not. The following potential effect modifiers were also identified by examining result of RCTs within the target population: sex/gender, performance status, disease stage (metastatic versus locally advanced), tumour location (oesophagus versus GEJ), histology (squamous cell carcinoma versus adenocarcinoma), ethnicity (Asian versus non-Asian), PD-L1 status (relevant for immunotherapies only). While the majority of these can be accounted for, performance status cannot as only Lee 2008 (13) reported this characteristics but not with sufficient granularity to allow for adjustment; the split reported in the study is between ECOG 0-1 and 2, but all patients in KEYNOTE 590 fell into the former category. The presence of ECOG 2 patients, which although only reported to be a small proportion of the Lee 2008 (13) trial (8.9%), would further introduce bias into any comparisons.

Even if a sufficiently large ESS could be achieved and the bias due to potential residual imbalances between populations were assumed to be minimal, there are also questions over the generalisability of results from a comparison of KEYNOTE-590 with the Lee 2008 (13) and Lee 2015 (14) trials to the population relevant to the decision problem. In an MAIC, treatment effects are estimated for the population as specified in the external comparator trials, which in

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

this context would be patients of Asian ethnicity with (metastatic) ESCC. Ethnicity, histology, and tumour location have all been identified as effect modifiers, although the role of ethnicity is subject to some controversy (26), meaning these may not hold within the more broader population specified in the decision problem. This along with the other issues noted above leads to the recommendation that indirect comparisons are not conducted.

B.2.10 Adverse reactions

The primary safety analyses of IA were based on data from the ASaT population of 740 participants as of the cut-off date of 02 July 2020. In all tables, individuals are counted only once for a specific AE term by the worst severity recorded.

Please refer to Appendix E for information related to the following:

• Drug Related AEs

• Grade 3-5 AEs

• Serious AEs

• Death to AEs

• Discontinuation due to AEs

• Interruptions due to AEs

• AEs of special interest

IA: July 2020 data-cut

The median exposure to study drug was similar between the pembrolizumab plus chemotherapy and the chemotherapy groups (Table 31). However, the mean exposure and mean number of cycles received was higher in the pembrolizumab plus chemotherapy group compared with the chemotherapy group. Participants in the pembrolizumab plus chemotherapy group remained on treatment longer as compared with the chemotherapy group. The rate of drug-related AEs was similar between the groups.

Table 31. Extent of Exposure - Summary of Duration on Therapy (ASaT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

In the pembrolizumab plus chemotherapy group, more participants had a duration of exposure of ≥6, ≥12, ≥18 and ≥24 months compared with participants in the chemotherapy group (Table 32).

Table 32. Exposure by duration (ASaT population)

Adverse events

The overall incidence of AEs was similar in the two groups. The incidences of AEs, drugrelated AEs, Grade 3 to 5 AEs, drug-related Grade 3 to 5 AEs, SAEs, drug-related SAEs, discontinuation due to AEs, and discontinuation due to SAEs were similar between treatment groups. The number of reported deaths due to AE was similar between the 2 treatment groups (pembrolizumab plus chemotherapy: 7.6%; chemotherapy: 10.3%). One death from the pembrolizumab plus chemotherapy group was due to COVID-19. There were no trends identified in the overall incidences of AEs by age, sex, race, baseline ECOG PS, and geographic region (Table 33).

Table 33. Disposition of Subjects (ITT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 34. Adverse Event Summary (ASaT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 35: Exposure-Adjusted Adverse Event Summary (Including Multiple Occurrences of Events) (ASaT Population)

Overall AEs

The overall incidence of AEs was similar in the pembrolizumab plus chemotherapy (100%)

and chemotherapy (99.5%) arms (Table 36). The most frequently reported AEs (those reported in ≥40% of participants in either treatment group) were nausea, anaemia, decreased appetite, fatigue, and constipation (Table 36).

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 36: Subjects With Adverse Events By Decreasing Incidence (Incidence ≥ 10% in One or More Treatment Groups) (ASaT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

MedDRA preferred terms "Neoplasm Progression", "Malignant Neoplasm Progression" and "Disease Progression" not related to the drug are excluded.

A rainfall plot comparing commonly reported AEs (incidence ≥10% in either treatment group) showed higher rates of WBC count decreased, neutrophil count decreased, rash, and hypothyroidism in the pembrolizumab plus chemotherapy group than in the chemotherapy group (Figure 9). These events were predominantly Grades 1 to 3 and manageable.

After adjustment for exposure, decreased WBC count, rash, and hypothyroidism remained higher in the pembrolizumab plus chemotherapy group compared with the chemotherapy group (Table 37). However, neutrophil count decreased was lower in the pembrolizumab plus chemotherapy group than the chemotherapy group after adjustment for exposure.

Figure 9: Between-Treatment Comparisons in Adverse Events Selected Adverse Events (>= 10% Incidence) and Sorted by Risk Difference (ASaT Population)

==> picture [480 x 253] intentionally omitted <==

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

==> picture [482 x 194] intentionally omitted <==

Table 37: Exposure-Adjusted Adverse Events (Including Multiple Occurrences of Events) (Incidence ≥ 10% in One or More Treatment Groups) (ASaT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Grade 3-5 AEs

The overall incidence of Grade 3 to 5 AEs was similar for pembrolizumab plus chemotherapy (85.9%) compared with chemotherapy (83.2%) (Table 38). The most frequently reported Grade 3 to 5 AEs (incidence ≥5%) for both treatment arms were: decreased neutrophil count, anaemia, neutropenia, hyponatremia, pneumonia, white blood cell count decrease.

Table 38: Subjects with Grade 3-5 Adverse Events by Decreasing Incidence (Incidence ≥ 5% in One or More Treatment Groups) (ASaT Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Drug-related Grade 3-5 AEs

The types and frequencies of drug-related Grade 3 to 5 AEs reported in the pembrolizumab plus chemotherapy group were generally consistent with the known safety profiles of pembrolizumab monotherapy and 5-FU/cisplatin chemotherapy.

The overall incidence of drug-related Grade 3 to 5 AEs was similar between pembrolizumab plus chemotherapy (related to at least one drug or both) (71.9%) and chemotherapy (67.6%) (Table 39). The most frequently reported drug-related Grade 3 to 5 AEs (incidence ≥5%) in both treatment groups were: Neutrophil count decrease, neutropenia, anaemia, white blood cell count decrease.

The largest between-treatment difference was noted in the incidence of neutrophil count decreased in the pembrolizumab plus chemotherapy group (22.7%) compared with the chemotherapy group (16.8%).

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 39: Subjects with drug-related grade 3-5 adverse events by decreasing incidence (Incidence ≥ 5% in One or More Treatment Groups) (ASaT Population)

Drug-related Serious AEs

The overall incidence of drug-related SAEs was similar between pembrolizumab plus chemotherapy (31.6%) and chemotherapy alone (26.21%). The most frequently reported drug-related SAEs were pneumonia (pembrolizumab plus chemotherapy: 3.5%; chemotherapy: 0.8%), pneumonitis (pembrolizumab plus chemotherapy: 3.2%; chemotherapy: 0.0%), and febrile neutropenia (pembrolizumab plus chemotherapy: 2.4%; chemotherapy: 3.2%).

B.2.11 Ongoing studies

The KEYNOTE-590 (21,22) study is ongoing, with an estimated study completion date of May 2023. There are no other ongoing clinical trials for pembrolizumab in this indication other than KEYNOTE-590.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B.2.12 Innovation

Currently in the UK, there is no innovative immuno-oncology treatment available for the firstline treatment of patients with untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma. Data from KEYNOTE590 (22) show that pembrolizumab in combination with chemotherapy is a promising treatment option which has demonstrated efficacy, including significant survival benefits, in all oesophageal patients, and has an acceptable tolerability profile.

Pembrolizumab, a monoclonal antibody, directly blocks the interaction of PD-1 and its ligands PD-L1 and PD-L2 enabling the immune response of both tumour-specific cytotoxic T lymphocytes in the tumour microenvironment and anti-tumour immunity. As evident by clinical and safety data presented, pembrolizumab offers a durable and well tolerated treatment option for patients with oesophageal cancer.

For decades, cytotoxic chemotherapies have remained the main treatment options for metastatic oesophageal cancer. For patients who have not received previous treatment, combination chemotherapies are routinely used, as palliative treatment options. Various palliative chemotherapy regimens have been investigated in oesophageal cancer studies and have been shown to have at least some activity in the first-line setting, with response rates ranging from 19% to 52% and 5-year survival rates of approximately 5%, with significant toxicity rates (27). Current National Comprehensive Cancer Network (NCCN) (28), European Society for Medical Oncology (ESMO) (29), and NICE NG83 (12) guidelines recommend the combination of a fluoropyrimidine (5-FU or capecitabine) with platinum agents (cisplatin, oxaliplatin, or carboplatin), either alone or in combination with a third drug such as epirubicin or a taxane, as the most effective first-line treatment option.

In 2018, the results of the interim analysis of KEYNOTE-180 demonstrated that pembrolizumab showed clinically meaningful anti-cancer activity as a third-line therapy in patients with oesophageal cancer, with an ORR of 9.9% (95% CI: 5.2, 16.7) (30). In addition, pembrolizumab monotherapy was approved for patients with ESCC whose tumours express PD-L1 CPS ≥10 in the US, Japan, China, and other countries based on KEYNOTE- 181, which used pembrolizumab as second line therapy in participants with advanced metastatic oesophageal cancer. Median OS for patients with ESCC whose tumours express PD-L1 CPS ≥10 was 10.3 months compared with 6.7 months with chemotherapy, 12-month OS was 48% versus 23%, respectively (31). As a result of the findings of KEYNOTE-180 and KEYNOTE181, pembrolizumab was granted FDA approval as monotherapy for the treatment of recurrent

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

locally advanced, metastatic, ESCC whose tumours express PD-L1 CPS ≥10 with disease progression after 1 or more systemic treatments.

The majority of patients are diagnosed with advanced/metastatic cancer, and in this setting, response to chemotherapeutic agents is poor. Given the high incidence and mortality worldwide and lack of effective therapeutic options, oesophageal cancer patients represent a patient population with a high unmet need for drug development.

B.2.13 Interpretation of clinical effectiveness and safety evidence

KEYNOTE-590 (22) met the predefined criteria for statistical significance for both of its primary endpoints of OS and PFS, as well as its key secondary endpoint of ORR. The results from the interim analysis of KEYNOTE-590 (22) provide evidence that treatment with pembrolizumab plus chemotherapy is superior to SOC alone for patients with untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma and provides a clinically meaningful improvement in OS, PFS and ORR. Results for OS, PFS, and ORR showed consistent benefit of pembrolizumab plus chemotherapy across all subpopulations analysed under the co-primary endpoints, and for all protocol specified subgroups that were considered. Pembrolizumab in combination with chemotherapy has generally acceptable safety profile.

In the all-comer population (the population reflected within the base-case of this submission), OS HR was 0.73 (95% CI: 0.62, 0.86; p <0.0001) in favour of pembrolizumab plus chemotherapy, reflecting a 27% reduction in the risk of death. The median OS was 12.4 months (95% CI: 10.5, 14.0) for the pembrolizumab plus chemotherapy group compared to 9.8 months (95% CI: 8.8, 10.8) for the placebo plus chemotherapy group.

PFS HR in all participants was 0.65 (95% CI: 0.55, 0.76; p< 0.0001) in favour of pembrolizumab plus chemotherapy, reflecting a 35% reduction in the risk of death or disease progression based on the investigator assessment per RECIST 1.1. The median PFS was 6.3 months (95% CI: 6.2, 6.9) for the pembrolizumab plus chemotherapy group and 5.8 months (95% CI: 5.0, 6.0) for the chemotherapy group.

Confirmed ORR in all participants was substantially higher in the pembrolizumab plus chemotherapy group than the chemotherapy group (45.0% vs 29.3%) reflecting a clinically meaningful 15.8% difference ( p <0.0001), based on investigator assessment per RECIST 1.1.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Median DOR in all participants in the pembrolizumab plus chemotherapy group was numerically longer (8.3 months; range: 1.2 to 31.0+) compared with the chemotherapy group (6.0 months; 1.5+ to 25.0+) based on investigator assessment per RECIST 1.1.

In all participants and other subpopulations, the differences in LS means from global health status/QoL were similar between the both treatment groups.

Incidences of AEs, drug-related AEs, Grade 3 to 5 AEs, drug-related Grade 3 to 5 AEs, SAEs, drug-related SAEs, discontinuation due to AEs, and discontinuation due to SAEs were similar between treatment groups. Incidences of discontinuation of any drug within the treatment regimen due to drug- related AEs and drug-related SAEs were higher in the pembrolizumab plus chemotherapy group than in the chemotherapy group. The incidence of AEOSI was higher in the pembrolizumab plus chemotherapy group compared with the chemotherapy group. The most common AEOSI categories in the pembrolizumab plus chemotherapy group were hypothyroidism (10.8%), pneumonitis (6.2%), and hyperthyroidism (5.7%).

It was not feasible to conduct an indirect treatment comparison between pembrolizumab and other treatments relevant to the UK population (capecitabine plus cisplatin and epirubicin with cisplatin and 5-FU) due to studies differences and lack of connected network between the studies identified in the SLR.

Internal validity

KEYNOTE-590 (21,22) is a robust, multi-centre, randomised, double-blind, placebo controlled phase III trial of pembrolizumab plus chemotherapy versus chemotherapy in patients with locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the EJG who have not received prior therapy. Prior to randomisation, eligible subjects were first stratified by histology (adenocarcinoma vs squamous cell carcinoma), geographic region (Asia versus Rest of the world) and ECOG PS, (0 vs 1).

The primary endpoints were to compare OS and PFS (per RECIST 1.1 as assessed by investigator) in subjects treated with pembrolizumab plus chemotherapy versus chemotherapy. OS is a clinically relevant endpoint, that was directly referenced in the final scope for this appraisal and the decision problem. This selected endpoint is consistent with that used in studies of other therapeutic agents in the population of locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the EJG. The definition of Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

progression when evaluating PFS in KEYNOTE-590 (21,22) followed an established response evaluation criterion (RECIST 1.1), in line with European Guidance (32).

HRQoL was explored under exploratory endpoints in the KEYNOTE-590 (21,22) study, with changes from baseline in patients treated with pembrolizumab plus chemotherapy compared to chemotherapy recorded using both the preferred measure of EQ-5D according to the NICE reference case, in addition to the cancer specific EORTC QLQ-C30 and EORTC QLQ-OES18. KEYNOTE-590 (21,22) is a double-blind study, with study sponsor, investigator and participant not aware of the treatment administered. This ensures the absence of bias in study results and the credibility of study conclusions.

External validity

KEYNOTE-590 (21,22) is a global study conducted in 168 centres in 26 countries, including 29 sites in Europe. Of the patients participating in the study, 162 were enrolled at sites in Europe, including 22 from the UK.

Baseline characteristics of patients enrolled in KEYNOTE-590 (22) were as expected for patients with locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the oesophagus or advanced/metastatic Siewert type 1 adenocarcinoma of the EJG. Most patients were male, 68.9 % of participants were Asian and 21.7% white, median age of the participants was 63 years. Subgroup analyses confirm the benefit of pembrolizumab plus chemotherapy versus chemotherapy in patients of all histologies (please see section 2.6.1 and Appendix E for more detailed discussion). The treatment arms were generally well balanced by all baseline characteristics.

The observed safety profile of pembrolizumab plus chemotherapy in KEYNOTE-590 (22) reflects the known safety profiles of the components i.e. generally well-tolerated. The types and severity of adverse events observed in the pembrolizumab plus chemotherapy group were generally consistent with the established pembrolizumab safety profile. No new safety signal was identified.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 40. End-of-Life Criteria

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

B.3 Cost effectiveness

B.3.1 Published cost-effectiveness studies

An SLR was conducted in two phases; an original search and a subsequent update, to identify relevant cost-effectiveness studies from the published literature. The initial search was conducted on 30[th] April 2020. An updated search employing the same search strategy of all previously searched bibliographic databases and grey literature was conducted on 24[th] November 2020.

The search did not identify any cost-effectiveness studies evaluating pembrolizumab in combination with chemotherapy in the specified population. Full details of the SLR search strategy, study selection process and results are presented in Appendix G.

B.3.2 Economic analysis

A cost-effectiveness study that met the relevant inclusion criteria for this appraisal was not identified, therefore a de novo cost-effectiveness model was built to assess the costeffectiveness of pembrolizumab in combination with chemotherapy compared with the relevant comparators.

Patient population

The patient population included in the economic evaluation consisted of patients with untreated, unresectable locally advanced or metastatic oesophageal cancer or HER-2 negative gastroesophageal junction adenocarcinoma. This is in line with the anticipated licence and with the NICE final scope. The patient characteristics were based on the European patients from the KEYNOTE-590 trial and are presented in Table 41, below.

As outlined in Section 2.6, MSD consider the CPS≥10 sub-population to be of particular clinical significance. The base-case population is reflective of the anticipated marketing authorisation, however analyses in the CPS≥10 sub-population is presented within section B.3.9, with the assumptions used within the economic model outlined within Appendix M.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Table 41. Baseline characteristics of patients included in the model

*These values refer to patients recruited from European sites participating in KEYNOTE-590

Model structure

The modelling approach for this appraisal is consistent with economic models developed for recent NICE oncology submissions in an advanced cancer setting (33–35), as well as advanced oesophageal cancer in the second-line setting (36). A partitioned survival cohort simulation model was developed to estimate health outcomes and costs for pembrolizumab in combination with chemotherapy and comparator regimens in the target patient population. The transition diagram of the cohort simulation model is presented in Figure 10, below.

Figure 10. Model structure

==> picture [251 x 181] intentionally omitted <==

----- Start of picture text -----
Progression
Free
Death
Progressive
Disease
----- End of picture text -----

There are three mutually exclusive health states in the model:

• Pre-progression, which is the starting health state, with patients staying in this state until disease progression or death

• Post-progression, which encompasses patients alive after progression and before death

• Death, which is an absorbing health state

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Partitioned survival modelling uses an overall survival curve to estimate the proportion of people alive over time directly- either from a parametric distribution or directly from KM trial data (37). The area under the (extrapolated) OS curve provides an estimate of mean life expectancy. OS may be further partitioned into different health states to allow these health states to have different HRQoL and cost implications (37). The model used requires two survival curves to estimate state membership for the model; the area underneath the OS curve represents the proportion of patients that were still alive (both in pre-progression and postprogression) at different points in time, while the proportion of patients in the pre-progression state were identified by the patients located underneath the PFS curve; where progression is defined by the primary censoring rule in KEYNOTE-590 trial (21), i.e. assessment per RECIST 1.1 assessed by investigator (21). Hence, the area between the PFS and the OS represents the proportion of post-progression patients, i.e. those who were in the ‘post-progression’ health state. Please see Figure 11 below.

Figure 11. Partitioned survival model structure

==> picture [343 x 186] intentionally omitted <==

Patients enter the model in the pre-progression health state. At the end of each weekly cycle, patients may remain in the state, transition to the post-progression health state or to death; patients who are in the post-progression state may remain in that state or die at the end of each cycle. Patients cannot transition to an improved health state (i.e. from post-progression to pre-progression). The partitioned survival model is unlike a Markov model, in which transition probabilities between health states are needed, as the proportions of patients in each health state at each time point is directly estimated.

For each health state, a specific cost and quality-of-life adjustment weight (i.e. utility) is assigned within each time period for calculating the cumulative costs and cumulative QALYs Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

over the modelled time horizon. Costs and QALYs are discounted with an annual rate of 3.5% in line with NICE reference case (38).

Please see Table 42 below comparing features of economic analysis between this appraisal and previous appraisals. Please note that NICE ID1249 (36) (Nivolumab for previously treated unresectable, advanced oesophageal cancer) is within a different patient population to that of this appraisal and has only been referred to due to the sparsity of Technology Appraisals conducted within an untreated advanced oesophageal population.

Table 42. Features of the economic analysis

Intervention technology and comparators

The intervention (pembrolizumab in combination with chemotherapy) was included in the

model as per the proposed licensed dosing regimen (i.e. pembrolizumab administered Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

intravenously at a fixed dose of 200 mg over 30 minutes up to 35 administrations combined with cisplatin administered intravenously at a dose of 80 mg/m[2] every 3 weeks (Q3W) for 6 doses and 5-FU administered as continuous IV infusion on days 1 to 5 at a dose of 800mg/m[2] /day (4000 mg/m[2] total per cycle) Q3W, up to 35 administrations).

The proposed licence states that pembrolizumab has to be administered until PD or unacceptable toxicities or for a maximum of 35 doses (2 years).

The decision problem outlined within the final scope issued by NICE determined the following to be comparators of relevance:

Platinum-based chemotherapy without pembrolizumab, such as:

• doublet treatment with fluorouracil or capecitabine plus cisplatin or oxaliplatin

• triplet treatment with fluorouracil or capecitabine plus cisplatin or oxaliplatin epirubicin

This includes the KEYNOTE-590 trial comparator, cisplatin in combination with 5-FU.

The NICE Guideline in the assessment and management of Oesophago-gastric cancer in adults (NG83)(12) outlines the difference in evidence between the different combinations (see Section 9.2.6). Comparisons 2, 7 and 8 are the most relevant for this appraisal:

Table 43 Summary of NG83 Section 9.2.6

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

As evidenced by NG83, comparison 2 shows the addition of an anthracycline (e.g. epirubicin) is not associated with a clinically significant difference in OS. Comparison 7 suggests that there is no clinically significant difference in overall survival in groups treated with oxaliplatin combinations compared with cisplatin combination. Furthermore, comparison 8 suggests that treatment with 5-FU combinations indicated a clinically significant benefit in OS when compared with groups treated with non-5-FU based combinations. The REAL-2 study further evidenced that capecitabine is similar to flurouracil, and oxaliplatin is similar to cisplatin, in terms of efficacy.

MSD sought clinical expert opinion to gain a greater understanding of the relative efficacies of the different combination therapies recommended by NICE. The clinical experts agreed that there was no difference in terms of efficacy between the doublet therapies. Furthermore, the addition of epirubicin (an anthracycline) added little efficacy benefit, with a considerable impact on toxicity. This was further verified at an advisory board, held virtually on the 29[th] January 2021.

As stated in section B.2.9, due to the difficulties in conducting an NMA, the approach undertaken to compare versus non-trial comparators which are used within UK clinical practice, is to assume clinical equivalency with the trial comparator (cisplatin in combination with 5-FU). This simplifying assumption is supported by NG83 and clinical expert opinion, given the clinician advice, the impact of this assumption is anticipated to be minimal and is investigated within scenario analyses. The base case comparison will be conducted versus the trial comparator, with comparisons versus each therapy as outlined by the NICE Final

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Scope (45), and a blended comparator (assuming equal distribution) investigated within scenario analyses.

B.3.3 Clinical parameters and variables

Method of modelling effectiveness

The clinical effectiveness parameters for pembrolizumab in combination with chemotherapy and SOC in the cost-effectiveness model were estimated from the KEYNOTE-590 patientlevel data on OS, PFS and adverse event rates.

The follow-up period of KEYNOTE-590 was shorter than the time horizon of the economic model, therefore, extrapolation of the OS and PFS curves was required. Parametric models were fitted to the KEYNOTE-590 KM data. The survival curve fitting was carried out in line with NICE Decision Support Unit (DSU) guidelines outlined in Technical Support Document 14 (46). In summary, the steps that were followed are presented in Figure 12 below.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 12. Survival model selection process algorithm (adapted from TSD 14) (46)

==> picture [381 x 369] intentionally omitted <==

AFT: Accelerated failure time; AIC: Akaike information criterion; BIC Bayesian information criterion; PH: Proportional hazards

Modelling OS

As KEYNOTE-590 (21) is a comparative phase III trial, patient-level data are available for both arms of the study. The cumulative and log cumulative hazard plots can be found in Figure 13 and Figure 14, respectively.

The log-cumulative hazard plots allow an assessment of whether the proportional hazards assumption is reasonable (37). As seen in Figure 14, the plots of the two arms are not parallel, as the plots cross, suggesting the proportional hazard assumption does not hold; hence a pooled parametric curve was deemed inappropriate. Furthermore, given the availability of IPD and the different mechanisms of action of pembrolizumab in combination with chemotherapy and SOC, parametric survival models were fitted separately to each treatment arm, as this approach required fewer assumptions than jointly fitted models.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 13. Cumulative hazard plot of OS for pembrolizumab in combination with chemotherapy and SOC

==> picture [291 x 274] intentionally omitted <==

Key: Pembrolizumab + Chemotherapy red; SOC blue

Figure 14. Log-cumulative hazard plot of OS for pembrolizumab + chemotherapy and SOC

==> picture [290 x 273] intentionally omitted <==

Key: Pembrolizumab + Chemotherapy red; SOC blue

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Two modelling approaches were used to estimate OS. The fully fitted modelling approach fits parametric models to the entirety of the KM data. Conversely, in the piecewise modelling approach, KM data is used directly within the model up to the point of the cut-off (week 32 or week 40), after which a parametric curve is fitted to the remaining KM data and used to estimate OS beyond the trial period. Both approaches utilised IPD from KEYNOTE-590.

The fully fitted parametric models have poor visual fit versus piecewise models. All fully fitted parametric models, including the statistically best-fitting log-logistic curve based on AIC/BIC statistics, underestimated the observed OS KM data in the first 8 months for both arms. Literature suggests that 5-year overall survival for stage IV oesophageal cancer is around 5% (47,48). The best-fitting one-piece log-logistic distribution underestimated 5-year survival for the trial comparator arm.

The two cut-offs at week 32 and week 40 for the piecewise models were identified based on structural changes observed by the Chow tests (see Figure 15), with higher Chow test statistics indicating a higher likelihood of structural change (49,50). Please note the Chow test statistics for the SOC arm proved inconclusive for determining an appropriate cut-off. The peak for the Chow test statistic was observed around week 40, the statistics are observed a smaller peak at around week 32. The piecewise model using the week 40 cut-off was selected as the base case given it is associated with the highest Chow test statistics. The Week 32 cutoff was explored in the scenario analysis.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 15. Plot of multiple Chow test statistics for OS in KEYNOTE-590: pembrolizumab in combination with chemotherapy, overall population

==> picture [342 x 307] intentionally omitted <==

For consistency, the Week 40 cut-off was selected for both the pembrolizumab in combination with chemotherapy and the SOC arms. Parametric survival extrapolations and the observed Kaplan–Meier data within the trial period are presented in

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 16 for the pembrolizumab in combination with chemotherapy arm and in Figure 17 for the SOC arm.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 16. OS KM curve with fitted piecewise model, 40-week cut-off, for pembrolizumab + chemotherapy based on KEYNOTE-590

==> picture [418 x 274] intentionally omitted <==

Figure 17. OS KM curve with fitted piecewise model, 40-week cut-off, for SOC based on KEYNOTE-590

==> picture [418 x 274] intentionally omitted <==

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Statistical tests using the Akaike information criterion (AIC) and the Bayesian information criterion (BIC), combined with visual inspection were used to help select the best-fitting parametric distribution based on internal validity. Table 44 below details the AIC/BIC statistics for both pembrolizumab in combination with chemotherapy and SOC.

Table 44. Summary of goodness-of fit qualities of OS models for pembrolizumab in combination with chemotherapy and SOC

The AIC/BIC statistics suggested that for pembrolizumab in combination with chemotherapy the best fitting distribution is the log-logistic function. For the SOC arm the best fitting distribution is the log-normal function. Notably, the gompertz distribution leads to clinically implausible outcomes.

However, as the modelled period is longer than the length of KM data, the external validity is also important for considering parametric curve selection. To help with the selection and validation of base case parametric survival models for extrapolation, a targeted literature search was conducted to identify studies reporting long-term OS for advanced and metastatic oesophageal cancer (47,51):

• A pan-European study based on oesophageal patients diagnosed between 1995 and 1999 and followed up to 2003 in 66 cancer registries in 24 European countries, including the UK, reported a 5-year survival rate of 3.8% for patients with distant stage oesophageal cancer (52).

• A 5-year survival rate of 5% for distant oesophageal cancer was reported by the American Cancer Society based on the US Surveillance, Epidemiology, and End Results (SEER) database for people diagnosed with oesophageal cancer between

2009 and 2015 (47,51). Similar estimates were reported by other publications using Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

stage IV oesophageal patients in the SEER database between 2010 and 2014 (53). Based on metastatic oesophageal patients in the SEER database between 1988 and 2012, Wu et al. reported 5-year and 10-year survival rates of 5.4% and 3.5% (54).

• In a small single-site retrospective study in Japan of 80 ESCC patients with distant organ metastasis, the 5-year survival rate of less than 5% was reported (55).

Table 45. Long term Overall Survival estimates for SOC

Table 46. Long term Overall Survival estimates for pembrolizumab in combination with chemotherapy

==> picture [452 x 150] intentionally omitted <==

----- Start of picture text -----
Log- Log- Generalized
KM Exponential Weibull Gompertz
normal logistic Gamma
3 month ***** ***** ***** ***** ***** ***** *****
6 month ***** ***** ***** ***** ***** ***** *****
1 year ***** ***** ***** ***** ***** ***** *****
2 year ***** ***** ***** ***** ***** ***** *****
5 year * **** **** ***** ***** ***** ****
10 year * **** **** **** **** ***** ****
20 year **** **** **** **** ***** ****
----- End of picture text -----

For the pembrolizumab in combination with chemotherapy arm, the log-logistic model has the best AIC and BIC and good visual fit within the trial period when compared with the observed Kaplan–Meier data. For the SOC arm, the log-logistic model has the second best-fitting AIC and BIC (with a difference of less than 1 versus the log-normal model that has the best-fitting AIC and BIC) and good visual fit. The log-logistic curve’s tail was considered by clinical experts to be more credible based on the expectation that a percentage of patients would derive a

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

long-term survival benefit from treatment with pembrolizumab in combination with chemotherapy.

The log-logistic curve with a cut-off at 40 weeks provides the most clinically plausible prediction for a survival rate of 4.8% and 2.0% at 5-years and 10-years for the SOC arm compared with available external data (see Table 45)(47,48). The long-term extrapolation for the pembrolizumab in combination with chemotherapy arm (5-year OS of 11.4% see Table 46) was also considered as clinically plausible by clinical experts, based on the mechanism of action for immunotherapy where a subgroup of patients are expected to receive long-term survival benefit.

Taking all of these factors into consideration; visual fit, statistical fit, clinical plausibility of longterm OS estimates, the piecewise log-logistic model with a cut-off at week 40 was used to model OS for both the intervention and comparator arms.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 18 and Figure 19 below show the extrapolated OS over a 5-year period and a lifetime horizon of 20 years. Please see Appendix M for a description of modelling OS in the CPS≥10 sub-population.

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 18. OS KM curves with fitted piecewise parametric models with a 40-week cut-off for the OS of pembrolizumab in combination with chemotherapy and SOC based on KEYNOTE-590 over a 5-year period

==> picture [418 x 273] intentionally omitted <==

Figure 19. OS KM curves with fitted piecewise parametric models with a 40-week cut-off for the OS of pembrolizumab in combination with chemotherapy and SOC based on KEYNOTE-590 over a lifetime horizon (20-year period)

==> picture [418 x 273] intentionally omitted <==

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Modelling PFS

PFS data from KEYNOTE-590 was relatively complete with over 90% of patients in each arm having reach the PFS endpoint. Based on the KEYNOTE-590 trial protocol (21), the first postrandomisation imaging was conducted at week 9 (± 1 week) with subsequent imaging being performed every 9 weeks (± 1 week) or more frequently if clinically indicated. Visual inspection of the KM PFS curves revealed a steep drop at around week 9 in both arms of KEYNOTE-590 data (22), reflecting the first protocol-scheduled tumour imaging assessment at 9 weeks (± 1 week). Prior to the first imaging only a small proportion of patients enrolled in KEYNOTE-590 reached the PFS endpoint, therefore, piecewise models with 10-week cut-off were selected as the base case approach for extrapolating PFS. In this approach the hazards rates for PFS failure are obtained directly from the KM curve up to week 10 followed by parametric models fitted to the post- week 10 data. To identify the most plausible PFS curve extrapolation among the standard parametric curves, the guidance from the NICE DSU was followed (46). Figure 20 and Figure 21 below display the PFS 10-week piecewise models for pembrolizumab in combination with chemotherapy and SOC.

Figure 20. PFS KM curve vs. fitted piecewise parametric curves for pembrolizumab in combination with chemotherapy

==> picture [419 x 273] intentionally omitted <==

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Figure 21. PFS KM curve vs fitted piecewise parametric curves for chemotherapy

==> picture [418 x 273] intentionally omitted <==

Statistical tests based on the AIC and the BIC, combined with visual inspection were used to help select the best-fitted parametric distribution based on internal validity. Table 47 reports the AIC/BIC statistics for the PFS curve extrapolations for both pembrolizumab in combination with chemotherapy and chemotherapy alone.

Table 47. Summary of goodness-of-fit qualities of the piecewise PFS models for pembrolizumab in combination with chemotherapy and chemotherapy

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

The best statistical fit according to AIC/BIC criterion for both treatment therapies is the loglogistic distribution (see Table 47), the second-best fitting curve for both arms being the generalised gamma distribution. Given that the KEYNOTE-590 PFS data is sufficiently mature, the primary method for the validation of the KM curve extrapolation is visual inspection and the AIC/BIC statistical tests, therefore the curves with the lowest scores were considered.

The base case for modelling PFS was a piecewise modelling approach; using KM data up to a cut-off of 10 weeks, followed by a log-logistic distribution. From Figure 20 and Figure 21 above, it is evident that the different extrapolation methods yield similar results due to the maturity of the data, as a result, the progressed proportion of patients is relatively insensitive to the choice of distributions used to extrapolate the KM data.

The modelled PFS curves used in the base-case analysis described above are presented in Figure 22 below. Please see Appendix M for a description of modelling PFS in the CPS≥10 sub-population.

Figure 22. PFS KM curves fitted with log-logistic curves from 10 weeks (base case)

==> picture [418 x 273] intentionally omitted <==

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Adverse Events

The AEs considered in the economic model include grade 3+ all-cause adverse events occurring in at least 5% of patients. The approach used to identify the relevant AEs to be included in the economic model has been previously validated by clinical experts.

The incidence of AEs was ascertained from the KEYNOTE-590 trial (22). Table 48 below presents the AEs included in the model and the proportion of patients who experience them. The AE profile for the blended chemotherapy arm was assumed to be equivalent to that of the comparator arm in the KEYNOTE-590 trial. MSD considers this to be a conservative approach given that clinical expert opinion suggests that triplet therapies (a proportion of which makes up the blended comparator arm) are associated with an increased incidence of adverse events. The unit costs and disutilities associated with each of the individual AEs were assumed to be the same across the different treatment arms; therefore, the difference in costs and disutilities associated with AEs in the model is driven by the AE rates displayed in Table 48 below. This approach is consistent with methods used in previous oncology submissions (33,34) and ensures that the full cost and HRQoL impact associated with AEs are captured within the model for all treatment arms without discounting.

In the base case, the impact of AEs was incorporated by estimating weighted average costs per patient, applied as a one-off cost. These were then applied in the first cycle of the model for each treatment arm, the same approach was taken for disutility associated with AEs.

Table 48. Grade 3+ AE rates for AEs included in the model

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Mean durations of the all-cause AEs were based on KEYNOTE-590 and were used together with AE rates and AE disutility to estimate the QALY decrements of each modeled arm due to AEs. The all-cause AE durations are assumed to be the same for all modelled arms and are presented in Table 49 below.

Table 49. Duration of Grade 3+ AEs in KEYNOTE-590

B.3.4 Measurement and valuation of health effects

Health-related quality-of-life (HRQoL) data from clinical trials

HRQoL of patients with advanced oesophageal cancer

The health-related quality-of-life of patients with untreated, unresectable locally advanced or metastatic oesophageal cancer or HER-2 negative gastroesophageal junction adenocarcinoma is likely to be heavily impaired due to the debilitating nature of the disease. Often patients have difficulty eating, and swallowing can also become difficult- leading to weight loss (59). Furthermore, as the cancer grows it can block or partially block the oesophagus, preventing food entry through the gut and hence the absorption of nutrients and calories. If patients are not able to eat and drink, they become more susceptible to other problems such as infection. Patients can also often feel fatigued and lacking in energy, with the emotional and physical changes affecting patients’ relationships (59). These factors exacerbate as a patient comes closer to death, alongside other physical changes such as being semi-conscious, loss of bladder and bowel control, restlessness, changes in breathing and confusion (60,61).

Evaluating HRQoL in KEYNOTE-590

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

HRQoL was evaluated in the KEYNOTE-590 trial using EuroQoL EQ-5D-5L. The latest position statement on the use of EQ-5D-5L value set for England by the National Institute for Health and Care Excellence (NICE) does not recommend using the EQ-5D-5L value set for English technology appraisals (62), and prefers utility values to be calculated by mapping the EQ-5D-5L descriptive system data onto the EQ-5D-3L value set using the mapping function developed by van Hout et al. (2012) in the reference (39). This approach was taken, and the results of this analysis have been used to inform inputs for the cost-effectiveness model.

In KEYNOTE-590, for both arms, the EQ-5D-5L questionnaire was administered on day 1 of every cycle from cycles 1 to 9, after which it was administered every 3 cycles for up to 1 year or until the end of treatment, whichever occurred first (21). The EQ-5D-5L data were also collected at time of discontinuation, and at the 30-day post-treatment discontinuation followup visit. A visit window of ±7 days were applied to EQ-5D-5L visit assessments.

The analyses of the EQ-5D-5L utilities were based on the Full Analysis Set (FAS) population for EQ-5D-5L questionnaire. The FAS population comprised of subjects who were randomized, received a study treatment, and completed at least one EQ-5D-5L questionnaire. Subjects were analysed in the treatment group allocated at randomization. The EQ-5D-5L FAS population included a total of 713 subjects.

HRQoL Utility approaches

The base-case used within the economic model is to apply time-to-death utilities. This approach is the most valid in such a rapidly progressing cancer, where patients deteriorate quickly as they get closer to death, hence a single utility value to homogeneously represent the post-progression period is less appropriate than in a cancer with a slower deterioration We recognise that health state based utilities according to progression status will also be of interest to the NICE committee; these have been included in a scenario analysis. The methodology employed is described below.

• Estimation of utilities based on time-to-death

This approach reflects the known decline in cancer patients’ quality of life during the terminal phase of the disease. This approach to define health state utilities based on time to death was developed by Batty et al. (2011) (63) and Hatswell et al. (2014) (64), which reflects the decline in the quality of life for patients with advanced or metastatic cancer as they approach death. It has previously been used in the estimation of

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

HRQoL in patients with advanced NSCLC who had previously received platinumbased chemotherapy or palliative radiotherapy (65–68) and in advanced melanoma (63,64) patients.

A time-to-death approach more accurately capturing the decrease in health-related quality of life over time (versus standard progression-based utilities) for patients with advanced oesophageal cancer. Hatswell et al (64) noted that disease progression may not fully capture all predictive factors of patient utility and time-to-death provides a good fit to patient data.

The time to death utility approach was accepted and deemed appropriate in multiple metastatic cancer HTA submissions to NICE and in a number of other pembrolizumab submissions to NICE in other metastatic cancer settings (e.g. TA531 (69), originally TA447 (70)).

Moreover, the utility estimate for progressed disease in the standard progressionbased approach is limited in reflecting the patient experience. EQ-5D utility data collected were collected within KEYNOTE-590 for up to one year, or until the end of treatment, whichever came first; the EQ-5D data were also collected at time of discontinuation, and at the 30-day post-treatment discontinuation follow-up visit. Therefore, data were collected for newly progressed patients but not for those whose condition had deteriorated. The application of time-to-death utilities mitigates this issue by basing utility valuations on time-to-death (regardless of whether death arises from a progression-free or progressive disease state) rather than by progression status.

An important limitation of the time-to-death approach is that the records measured within 360 days from OS censoring date cannot be assigned to a time-to-death category due to uncertain death date. However, for KEYNOTE-590 (22), by the data cut-off date of July 2, 2020, 571 of 749 patients (76.2%) in the intention-to-treat population had a known death date. Amongst all 5744 EQ-5D-5L measures, only 318 (5.54%) had unknown time-to-death category. For this analysis, the uncertainty in the utility by time-to-death approach due to unknown death dates is relatively low and hence the results can be deemed to be largely representative of the patient population.

In this analysis, the linear mixed-effects regression model included indicators for time to death (i.e. 0-29, 30-89, 90-179, 180-359, or ≥ 360 days until death) and the

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

presence/absence of any Grade 3+ AEs, as well as patient-level random effects to account for correlation between repeated measurements of the same patient.

In the model, utilities were applied based on the distribution of patients across different categorizations of time to death in each weekly cycle. In a given weekly cycle, the proportion of patients within each time to death category was estimated based on the modelled OS within each treatment arm.

• Estimation of utilities based on progression-free and progressed disease states.

This approach, commonly seen in previous oncology economic modelling literature, defines health states based on time relative to disease progression (33,71). This approach generates results to be used in the economic model by health state. As previously mentioned, a limitation of this approach is attributed to the distribution and collection of EQ-5D questionnaires within KEYNOTE-590. EQ-5D utility data postprogression were only collected at the point of treatment discontinuation and at the 30day post-treatment discontinuation follow-up visit. Therefore, post-progression utility data is limited, and unlikely to accurately reflect the rapid deterioration of patients’ quality of life in this cancer. Due to the paucity of data within this disease area, it is not possible to substitute utility values from the literature to alleviate this issue. The reference case asserts a preference for using utility data ascertained from the relevant clinical trial, hence using utility values sourced from the literature, as a substitute, would require substantial justification as utility data from KEYNOTE-590 is available. Due to this limitation, the post-progression utility is not representative of the lived experience of patients in their terminal month; hence utilities reported by the time-todeath approach are considered a more accurate measure of HRQoL.

The date of progression was determined via RECIST 1.1 per investigator assessment (21). To estimate utilities:

• for the progression-free health state, EQ-5D scores collected at all visits before the progression date were used.

• for the progressive disease health state, EQ-5D scores collected at all visits after the progression date were used.

Utility in both the progression-free state and the progressive disease were estimated by the linear mixed-effects model. The model included indicators for progressionbased health states (progression-free vs progressive disease) and the

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

presence/absence of any Grade 3+ AEs. The same utilities were applied to all treatment arms in the model and also for all subgroups.

Please see the results of both methods in Table 50 and Table 51 below.

Table 50. EQ-5D health utility scores by time-to-death

Table 51. EQ-5D health utility scores by progression status (pooled)

Collection of EQ-5D questionnaires

For each of the utility approaches, mean EQ-5D utility scores by health status were estimated per treatment arm (pembrolizumab in combination with chemotherapy and SOC arms), and pooled for both arms. In addition, 95% CIs were obtained for each estimated EQ-5D utility and the statistical significance of the differences between treatment arms was tested. The level of EQ-5D compliance through time is presented in Table 52.

Table 52. Compliance of EQ-5D by visit and by treatment (FAS Population)

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

Mapping

In KEYNOTE-590 (21), the EQ-5D-5L (72) instrument was used to measure generic health status. It contains five health state dimensions: mobility, self-care, usual activity, pain/discomfort, and anxiety/depression. Each dimension is rates on a 5-point scale from 1 (no problems) to 5 (extreme problems).

The latest position statement on the use of EQ-5D-5L value set for England by NICE does not recommend using the EQ-5D-5L value set for England. The EQ-5D-5L data collected in the trial can be converted to population-based utility valuations using published algorithms, and the NICE position statement recommends calculating utility values by mapping the EQ-5D-5L descriptive system data onto the EQ-5D-3L value set using the mapping function developed by van Hout et al. (2012). In this analysis, health state utilities were estimated based on

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

mapping EQ-5D-5L data collected in KEYNOTE-590 to EQ-5D-3L value set using the mapping function from van Hout et al. (2012), as per NICE’s recommendation (39,62).

Health-related quality-of-life studies

Please see Appendix H for a list of the studies identified through the SLR.

Adverse reactions

AE-related disutility was applied as a one-time QALY decrement in the first model cycle (i.e. Week 0). Disutility associated with AEs per patient was calculated in each treatment arm as a function of: the rates of included AEs in the treatment arm (Table 48), the mean duration of AEs (Table 49), and the estimated disutility associated with Grade 3+ AE based on the timeto-death approach.

Table 53. One-off QALY decrements due to Grade 3+ AEs by treatment arm

Age-related disutility

The model considered additional age-related utility decrements as the modelled population ages over the modelled time horizon. The age decrements were calculated as the relative change of the general population utility at the modelled age compared with the utility at the starting age (Table 41). The calculated age adjustments were then applied to the estimated time to deaths or progression-based utility values in the model. The UK age-related general population utility were derived from the literature (73).

Pembrolizumab with platinum-based chemotherapy for untreated, unresectable locally advanced or metastatic oesophageal cancer or gastroesophageal junction adenocarcinoma [ID3741]

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