CADTH Reimbursement Review

Pembrolizumab (Keytruda)

Sponsor: Merck Canada Inc.

Therapeutic area: Triple-negative breast cancer

This multi-part report includes:

Clinical Review

Pharmacoeconomic Review

Stakeholder Input

Clinical Review

Executive Summary

An overview of the submission details for the drug under review is provided in Table 1.

Introduction

Triple-negative breast cancer (TNBC) is an invasive form of breast cancer affecting 10% to 20% of patients with breast cancer.^1,2 TNBC is distinguished by the absence of an estrogen receptor (ER) and progesterone receptor (PR) and little to no expression of the gene for human epidermal growth factor receptor 2 (HER2).^3-5 The disease is most common in females under 40 years of age, Black females, and females with a BRCA1 mutation.^2,3 Diagnosis of metastatic triple-negative breast cancer (mTNBC) is based on imaging tests (i.e., mammogram, breast ultrasound, and MRI), clinical symptoms, and biopsy to confirm breast cancer with negative results on all 3 tests (ER, PR, and HER2). Metastatic TNBC differs from other types of invasive breast cancer in that it tends to grow and spread faster, has fewer treatment options, and tends to have a worse prognosis. The predicted 5-year survival rate is 12% for mTNBC compared to 77% for all TNBC and 89% for all breast cancer.^6-8

The standard approach for mTNBC is treatment with single-drug chemotherapies such as taxanes and gemcitabine plus carboplatin.⁹ The median overall survival (OS) for patients with mTNBC treated with conventional chemotherapy is 9 to 13 months.^10-13 Combination chemotherapy may be used in patients with progressive or higher burden of disease, such as rapidly progressive visceral disease.⁹ No approved targeted treatments are currently available.⁹ Immunotherapy (in combination with chemotherapy) is considered a potential targeted treatment option for patients who have programmed cell death 1 ligand 1 (PD-L1) expression. According to the experts consulted by CADTH, patients who have a higher PD-L1–positive combined positive score (CPS) may be more likely to benefit from treatment regimens that use immunotherapies in combination with chemotherapy compared with chemotherapy alone.

Pembrolizumab is an immunoglobulin G4 monoclonal antibody against programmed cell death 1 protein (PD-1). By inhibiting the PD-1 receptor from binding to its ligands expressed on T cells, pembrolizumab restores cytotoxic T-cell effector function.¹⁴ Combinations of pembrolizumab and chemotherapy have been studied in an effort to enhance the antitumour activity of single drugs by creating a tumour microenvironment that stimulates responses to immunotherapy.¹⁴

The objective of this report is to perform a systematic review of the beneficial and harmful effects of pembrolizumab plus chemotherapy in adult patients with locally recurrent unresectable or mTNBC who have not received prior chemotherapy for metastatic disease and whose tumours express PD-L1 as determined by a validated test (CPS ≥ 10).

Stakeholder Perspectives

The information in this section is a summary of input provided by the patient groups who responded to CADTH’s call for patient input and clinician group input, and from clinical experts consulted by CADTH for the purpose of this review.

Patient Input

Two patient groups, the Canadian Breast Cancer Network (CBCN) and Rethink Breast Cancer (RBC) provided input for this review. The CBCN collected patient input via 2 online surveys in patients in Canada with mTNBC (survey 1: 2017, n = 14; survey 2: 2012, n = 87 [71 patients and 16 caregivers]) and a grey literature search. None of the respondents from the CBCN surveys had direct experience with pembrolizumab. The RBC input was based on general observations and insights gathered through various ongoing initiatives (including patients’ blogs, virtual support groups, working groups, patient advisory boards, peer-support networks, Instagram, and scientific advisory committees), as well as online surveys (N = 78), Zoom videoconference interviews (n = 7), one-on-one interviews (n = 2; 1 had experience with pembrolizumab), and 2 testimonials from patients in Canada with mTNBC.

In the CBCN submission, patients highlighted the negative impacts of metastatic breast cancer symptoms, such as fatigue (reported by 54%), insomnia (39%), and pain (37%). The majority of respondents experienced metastases to their lungs (N = 10), followed by metastases to other parts of their bodies (e.g., bones, liver, and brain). These symptoms and metastases impose a heavy physical, emotional, psychosocial, and financial toll and can negatively affect health-related quality of life (HRQoL). Respondents from the CBCN surveys acknowledged that currently available treatments for mTNBC have only been shown to prolong the progression-free period and highlighted the decreasing response rates in later lines of therapy; while the disease will eventually progress, patients seek to live their remaining months and years with the best possible HRQoL. The input provided by RBC indicated that, particularly for patients who are diagnosed at a young age, TNBC may have detrimental effects on patient well-being in terms of fertility, childcare, relationships, body image, social activities, employment, and mental health.

Patients from both groups expressed a need for new treatments that prolong survival, control disease, and maintain quality of life. The RBC submission suggested that patients value long-term health outcomes over immediate concerns such as reducing symptoms or managing side effects. CBCN respondents expected new treatments to improve disease control and quality of life, delay disease progression, have minimal side effects, and be affordable with easy access to treatment.

One patient with mTNBC from RBC who had direct experience with pembrolizumab treatment stated that the drug had helped control the disease, shrink tumour size, and improve HRQoL.

Clinician Input

Input From Clinical Experts Consulted by CADTH

According to the clinical experts, improvement in survival and quality of life (QoL) remain the most important unmet needs in patients with mTNBC. The clinical experts agreed that pembrolizumab would be used in the first-line setting for the treatment of mTNBC. In the opinion of the clinical experts, pembrolizumab with chemotherapy would be the new standard of care and would replace current treatment (e.g., chemotherapy alone), rather than being reserved for patients who are intolerant to existing treatments. The clinical experts advised that patients best suited for treatment with pembrolizumab combined with chemotherapy are those who meet the inclusion criteria of the KEYNOTE-355 study. In particular, those patients with a CPS of 10 or higher would be eligible for pembrolizumab. The clinical experts advised that patients least suitable for pembrolizumab included those who are unfit, frail, have a poor Eastern Cooperative Oncology Group Performance Status (ECOG PS), or have active autoimmune disease. The experts indicated that radiological and clinical investigations are used to evaluate response on a regular basis. The experts also noted that PD-L1 status has proven to be predictive of benefit in metastatic cases. The most common assessment modality is shrinkage of tumour on clinical and/or radiological tests. The clinical experts stated that disease progression (tumour enlargement unless pseudoprogression is suspected) and the occurrence of intolerable adverse events (AEs) would be considered when choosing treatment discontinuation. The clinical experts stated that prescription decisions depend on the prescribing centre’s comfort level with administration of the drug and ability to handle side effects.

Clinician Group Input

Input from 3 clinician groups was provided: Ontario Health – Cancer Care Ontario (OH-CCO) Breast Cancer Drug Advisory Committee (based on input from 1 clinician); breast cancer medical oncologists in Canada, led by The Ottawa Hospital Cancer Centre (TOHCC) (based on input from 6 medical oncologists); and the Provincial Breast Tumour Group (Alberta) (based on input from 2 medical oncologists). OH-CCO’s Drug Advisory Committee provides timely evidence-based clinical and health-system guidance on drug-related issues in support of Cancer Care Ontario’s mandate, including the Provincial Drug Reimbursement Programs and the Systemic Treatment Program. The TOHCC group consists of medical oncologists throughout Canada with an expertise in breast oncology and in particular systemic therapies in advanced TNBC. The Provincial Breast Tumour Group (Alberta) is responsible for treating patients with TNBC in Alberta.

The clinician groups indicated that the current first-line treatment for mTNBC, cytotoxic chemotherapy (taxane, platinum, and anthracycline), showed limited benefit in fulfilling the treatment goal of delaying disease progression and improving both the duration and quality of a patient’s life. The clinician groups identified an unmet need for more effective and tolerable treatment options in the metastatic setting. OH-CCO emphasized that the lack of CPS testing within the province may delay delivery of test results and the diagnosis of TNBC. The clinician groups all mentioned that pembrolizumab in combination with chemotherapy would fit in the first-line setting in treating patients with mTNBC. The clinician groups anticipated that pembrolizumab with chemotherapy could shift the current treatment paradigm by replacing chemotherapy alone. Patients suitable for pembrolizumab with chemotherapy identified by the clinician groups were those who expressed PD-L1 (CPS ≥ 10), with an ECOG PS of 0 to 2, and time from completion of adjuvant treatment greater than 6 months, which were aligned with the KEYNOTE 355 trial inclusion criteria. Patients least suitable were those who do not have a CPS of 10 or higher in other lines of therapy or had received adjuvant and/or neoadjuvant pembrolizumab for early-stage disease within 12 months, and those with a high risk of adverse toxicity related to immunotherapy.

Clinician groups agreed that treatment response should be measured using assessments based on symptoms, laboratory markers, radiographic scans, and tumour measurements. The appropriate interval for assessments is 3 months. Improvement of organ function (bone, liver, and lung) and severity of symptoms, maintenance or improvement of performance status, tumour radiographic response with either stabilization of disease or response as measured by Response Evaluation Criteria in Solid Tumours Version 1.1. (RECIST 1.1) were considered clinically meaningful responses. Disease progression, intolerable or dangerous toxicity (grade 3 or higher immune-mediated toxicity), and patient preference or refusal should be considered when deciding whether to discontinue treatment of pembrolizumab with chemotherapy. Clinician groups agreed that outpatient settings, such as hospitals or specialty clinics that have oncology specialists (medical oncologists, chemotherapy nurses, or oncology pharmacists) to administer systemic cancer therapies and monitor and manage treatment-related toxicities, are appropriate for treatment. In addition, the clinician groups pointed out that a combination of pembrolizumab and chemotherapy is now considered a new standard of care by internationally accepted guidelines and has been shown to be well tolerated with a manageable toxicity profile and to be highly accepted and valued by patients in Canadian clinical practice; it is therefore imperative that Canadians have access to this treatment.

Overall, the views of the clinician groups were consistent with those of the clinical experts consulted by CADTH.

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may affect their ability to implement a recommendation. This includes considerations for initiation, discontinuation, and prescribing of therapy, and generalizability. The implementation questions and corresponding responses from the clinical experts consulted by CADTH are summarized in Table 4.

Clinical Evidence

Pivotal Study

Description of Study

The KEYNOTE-355 study is an ongoing, phase III, randomized, multicentre, double-blind, 2-part, placebo-controlled trial. The primary objective of the trial is to evaluate the efficacy and safety of pembrolizumab plus chemotherapy versus placebo plus chemotherapy for patients with locally recurrent inoperable or metastatic TNBC that has not been previously treated with chemotherapy for metastatic disease. The KEYNOTE-355 trial was initiated in August 2016 and has 251 participating centres across 29 countries in North America (5 sites in Canada, N = 34), South America, Europe, Asia, and Australia. The study was conducted in 2 parts. Part 1 was the safety run-in (N = 30), and part 2 was the efficacy evaluation (N = 847). Patients from part 1 were not included in part 2. For the purpose of this CADTH review, only part 2 was evaluated. The 2 primary efficacy outcomes were OS and progression-free survival (PFS) in all patients and patients with PD-L1–positive tumours. Secondary outcomes included objective response rate (ORR), duration of response (DOR), disease control rate, and HRQoL in all patients and patients with PD-L1–positive tumours. The study is considered to have met its primary objective if the combination of pembrolizumab and chemotherapy was superior to placebo and chemotherapy in either PFS or OS in either all patients or in patients with PD-L1–positive tumours (CPS ≥ 1 or CPS ≥ 10) at either an interim analysis or the final analysis (OS only). Given that the indication for the sponsor’s reimbursement request is for patients with a PD-L1 CPS of 10 or higher, this review focuses on patients with a PD-L1 CPS of 10 or higher and all patients for comparison. In total, 1,372 patients were screened, of which 847 were randomized via an interactive voice recognition system in the intention-to-treat (ITT) population at a 2:1 ratio between pembrolizumab plus chemotherapy (N = 566) and placebo plus chemotherapy (N = 281). Randomization was stratified based on 3 stratification factors:

• type of chemotherapy on study (paclitaxel or nab-paclitaxel or gemcitabine and carboplatin)

• PD-L1 expression at baseline (CPS ≥ 1 or < 1)

• prior treatment with the same class of chemotherapy in the neoadjuvant or adjuvant setting (yes or no).

Treatment was to continue until confirmation of progressive disease or death. Both patients and investigators were blinded to study treatments administered during the trial. In total, 75.1% and 38.1% had a tumour-tissue PD-L1 CPS of 1 or higher and 10 or higher, respectively. All patients enrolled were female. The majority of patients were younger than 65 years of age, white, postmenopausal, and had an ECOG PS of 0. There were 3 planned efficacy interim analyses and a final analysis for part 2 of the KEYNOTE-355 trial. The final analysis of the study (data cut-off date of June 15, 2021) was event- and follow-up-time–driven and was to be conducted after approximately 500 OS events had been observed among all patients or after approximately 23 months since the last patient was randomized, whichever was later. The final analysis of ORR results was conducted at interim analysis 1 (IA1) and the final analysis of PFS was conducted at interim analysis 2 (IA2).

The focus of this CADTH review is the final analysis and its consistency with results from IA1 and IA2.

Efficacy

Table 2 presents a summary of efficacy results from the KEYNOTE-355 trial. All efficacy results are reported for the subset of patients with PD-L1–positive tumours (CPS ≥ 10).

Overall Survival

At the final analysis data cut-off (June 15, 2021), the KEYNOTE-355 trial met the success criterion for the primary hypothesis of OS in patients with locally recurrent inoperable or metastatic TNBC and PD-L1–positive tumours (CPS ≥ 10). The median OS was 23.0 months (95% CI, 19.0 to 26.3) in the pembrolizumab plus chemotherapy group and 16.1 months (95% CI, 12.6 to 18.8) in the placebo plus chemotherapy group. The hazard ratio (HR) obtained between pembrolizumab plus chemotherapy versus placebo plus chemotherapy was 0.73 (95% CI, 0.55 to 0.95; P = 0.0093; multiplicity-adjusted, 1-sided nominal alpha = 0.0113). Overall, pembrolizumab plus chemotherapy demonstrated a statistically significant and clinically meaningful improvement in OS that represented a 27% reduction in the risk of death for patients with PD-L1–positive tumours (CPS ≥ 10).

Health-Related Quality of Life

The following questionnaires were used to assess HRQoL: the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30), European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire – Breast Module 23 (EORTC QLQ-BR23) and the EQ visual analogue scale (EQ VAS). Overall, there were no differences between groups in terms of HRQoL on any of the questionnaires.

Progression-Free Survival

At the IA2 data cut-off (December 11, 2019), the KEYNOTE-355 trial met the coprimary end point of PFS in patients with PD-L1–positive tumours (CPS ≥ 10). At IA2, the median PFS was 9.7 months in the pembrolizumab plus chemotherapy and 5.6 months in the placebo plus chemotherapy group. The HR was 0.65, (95% CI, 0.49 to 0.86; 1-sided P value = 0.0012). Based on the prespecified success criteria (alpha = 0.00411), pembrolizumab plus chemotherapy statistical significantly improved PFS compared with placebo plus chemotherapy in patients with a CPS of 10 or higher.

The findings on PFS from the final analysis were consistent with the results at IA2. At the final analysis data cut-off (June 15, 2021), a total of 144 PFS events (65.5%) had occurred in the pembrolizumab plus chemotherapy group compared to 81 events (78.6%) in the placebo plus chemotherapy group in patients with PD-L1–positive tumours (CPS ≥ 10). The HR at final analysis was 0.66 (95% CI, 0.50 to 0.88; P = 0.0018) in patients with a PD-L1 CPS of 10 or higher. The median PFS was 9.7 months (95% CI, 7.6 to 11.3) in the pembrolizumab plus chemotherapy group compared to 5.6 months (95% CI, 5.3 to 7.5) in the placebo plus chemotherapy group in patients with PD-L1–positive tumours (CPS ≥ 10). Overall, pembrolizumab plus chemotherapy continued to show a statistically significant improvement in PFS compared with placebo plus chemotherapy in patients with PD-L1–positive tumours (CPS ≥ 10).

Objective Response Rate

At the final analysis data cut-off, 116 patients (52.7%) had an objective response (95% CI, 45.9 to 59.5%) in the pembrolizumab group compared to 42 patients (40.8%) with an objective response (95% CI, 31.2 to 50.9%) in the placebo group among those with a PD-L1 CPS of 10 or higher. The secondary hypotheses pertaining to ORR in all patients and in patients with a PD-L1 CPS of 1 or higher were tested at IA1 and not formally tested at the final analysis. The ORR for patients with a PD-L1 CPS of 10 or higher was not included in the multiplicity strategy. These ORR results from the final analysis are consistent with the IA1 results. Overall, pembrolizumab plus chemotherapy showed greater improvements in ORR (according to RECIST 1.1 by a blinded independent central review) compared with placebo plus chemotherapy in patients with a PD-L1 CPS of 10 or higher.

Duration of Response

In the final analysis, the median DOR was greater in the pembrolizumab plus chemotherapy group at 12.8 months compared to the placebo plus chemotherapy group at 7.3 months in patients with a PD-L1 CPS of 10 or higher. Median time to response was 1.9 months in both groups. DOR results from the final analysis were consistent with those in the previously reported IA2 results and showed a sustained DOR benefit in the pembrolizumab plus chemotherapy group over a longer duration of follow-up. The proportions of patients with an extended response at 6 months or later and 12 or later months by Kaplan-Meier estimation were 87.1% and 55.5% in the pembrolizumab plus chemotherapy group and 55.5% and 37.9% in the placebo plus chemotherapy group, respectively.

Harms

Table 2 presents a summary of key safety results from the KEYNOTE-355 trial. Overall, almost all patients in the pembrolizumab plus chemotherapy group (99.6%) and in the placebo plus chemotherapy and placebo group (98.2%) reported at least 1 AE by the June 15, 2021, data cut-off. Grade 3 or higher AEs were slightly more common in the pembrolizumab plus chemotherapy group (77.9%) compared to placebo plus chemotherapy (73.7%). The most common AEs in both groups were neutropenia, decreased neutrophil count, anemia, thrombocytopenia, decreased white blood cell count, and leukopenia.

Table 2: Summary of Key Results From Pivotal Study

AE = adverse event; CI = confidence interval; CPS = combined positive score; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; EQ VAS = EQ visual analogue scale; LS = least squares; PD-L1 = programmed cell death 1 ligand; 1; PFS = progression-free survival; SAE = serious adverse event; WDAE = withdrawal due to adverse event.

Notes: “+” indicates there is no progressive disease by the time of last disease assessment. Data cut-off: June 15, 2021 (final analysis).

^aFrom product-limit (Kaplan-Meier) method for censored data.

^bBased on Cox regression model with the Efron method of tie handling with treatment as a covariate stratified by chemotherapy on study (taxane vs. gemcitabine and/or carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^cThe multiplicity-adjusted, 1-sided, prespecified P value boundary for statistical significance was P = 0.0113. One-sided P value based on log-rank test stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^dBased on a constrained longitudinal data analysis model with the patient-reported outcome scores as the response variable, and treatment by time point interaction, and stratum (defined by stratification factors of chemotherapy on study [taxane vs. gemcitabine and carboplatin] and prior treatment with same class of chemotherapy in the (neo)adjuvant setting [yes vs. no]) as covariates.

^eNominal P value (i.e., the type I error rate has not been controlled). Two-sided P value.

^fNominal P value (i.e., the type I error rate has not been controlled); 1-sided P value based on log-rank test stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^gBased on Miettinen and Nurminen method stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^hNominal P value. One-sided P value for testing. Null hypothesis: difference in percentage of 0 vs. hypothesis 1: difference in percentage greater than 0. Confirmed responses are included.

Source: Clinical Study Report.¹⁵

AEs of grade 3 to 5 reported in at least 5% of patients were also generally similar in both treatment arms; these included neutropenia (30.1% versus 30.2%), decreased neutrophil count (18% versus 20.3%), anemia (18% versus 16.4%), thrombocytopenia (11.2% versus 11.7%), and decreased white blood cell count (10.5% versus 10.3%) in the pembrolizumab plus chemotherapy and pembrolizumab group and placebo plus chemotherapy and placebo group, respectively. Overall, 3.0% and 1.8% of AEs resulted in death in the pembrolizumab plus chemotherapy and the placebo plus chemotherapy groups, respectively. AEs leading to discontinuation of any study intervention were more common in the pembrolizumab plus chemotherapy group (10.7%) than in the placebo plus chemotherapy group (5.3%).

Notable harms were more common in the pembrolizumab plus chemotherapy group compared to placebo plus chemotherapy, except for infusion reactions. The most common notable harms in the pembrolizumab versus placebo chemotherapy group were hypothyroidism (15.8 versus 3.2%), hyperthyroidism (4.3% versus 1.1%), infusion reactions (3.7% versus 5.0%), and pneumonitis (2.5% versus 0).

Critical Appraisal

In the KEYNOTE-355 trial, the consistency between the interim results and final results helped strengthen the robustness of the findings at the final analysis. For example, PFS assessed at the final analysis was not formally tested based on the statistical hierarchy, yet the results were consistent with the results at IA2, in which a statistically significant result was demonstrated. OS and PFS are validated outcomes recommended by the FDA for trials that include patients with TNBC.¹⁶ The clinical experts considered these outcomes important for patients with mTNBC in Canadian practice.

A potential concern in the KEYNOTE-355 study included the lack of stratification randomization conducted on patients with a PD-L1 CPS of 10 or higher or lower than 10, which was the primary focus of this reimbursement review (subpopulation of PD-L1 CPS ≥ 10). This may have contributed to imbalances between the treatment arms, particularly potential unknown confounding covariates. The Health Canada Reviewer Report raised the same concern that was addressed by the sponsor by noting that an evaluation and sensitivity analysis of the potential for imbalance in the population with a CPS of 10 or higher was conducted. It was concluded that the impact of not having a CPS of 10 or higher as a stratification factor would be minimal, and the imbalances in baseline factors were limited and did not change the conclusion of the primary analysis. The Health Canada Reviewer Report acknowledged that the sponsor’s responses to this concern were detailed and adequate.¹⁷ The HRQoL surveys were conducted at week 15; however, this time point might not be able to capture an accurate picture of patients’ experiences with the combination of pembrolizumab plus chemotherapy for a prolonged period of time. The assessment time point at week 15 was selected to ensure at least 60% completion and 80% compliance with patient-reported outcome assessments to ensure validity of the longitudinal model of change in patient-reported outcome scores over time. Analyses performed on the patient-reported outcomes were noninferential.

The inclusion and exclusion criteria of the KEYNOTE-355 study were considered appropriate and the baseline and demographic characteristics were considered generalizable to Canadian practice by the clinical experts consulted by CADTH for this review. The magnitude of benefit of pembrolizumab in patients who did not meet the inclusion criteria outlined in the KEYNOTE-355 study is uncertain. The choice of chemotherapy and duration of follow-up implemented in the KEYNOTE-355 study were considered appropriate by the experts. The clinical experts noted that the concomitant medications allowed in the KEYNOTE-355 study were also commonly used in Canadian practice and were considered appropriate. The experts did not identify any major concerns and discrepancies between the trial and real-world settings in terms of concomitant medication use and duration of follow-up.

Indirect Treatment Comparison

Description and Methods of Published Network Meta-Analysis

The sponsor-submitted indirect treatment comparison (ITC) involved a systematic review and used a Bayesian NMA to evaluate the relative efficacy and safety of pembrolizumab plus chemotherapy compared to other treatments including nab-paclitaxel or paclitaxel, atezolizumab plus nab-paclitaxel, bevacizumab plus paclitaxel, carboplatin, docetaxel, ixabepilone plus bevacizumab, bevacizumab plus nab-paclitaxel, and pembrolizumab plus nab-paclitaxel or paclitaxel for the first-line treatment of patients, with previously untreated locally recurrent inoperable or metastatic TNBC. The efficacy outcomes of interest were PFS and OS.

Efficacy Results

The sponsor-submitted ITC reported that the results for OS favoured pembrolizumab plus chemotherapy over nab-paclitaxel or paclitaxel (HR = 0.54; 95% credible interval [CrI], 0.36 to 0.82), carboplatin (HR = 0.36; 95% CrI, 0.19 to 0.68) and docetaxel (HR = 0.30; 95% CrI, 0.17 to 0.55).

The sponsor-submitted ITC reported that the results for PFS favoured pembrolizumab plus chemotherapy over nab-paclitaxel or paclitaxel (HR = 0.51; 95% CrI, 0.33 to 0.78) However, pembrolizumab plus chemotherapy was not favoured versus other comparators included in the ITC for PFS.

Critical Appraisal

Due to limited data availability, the sponsor-submitted ITC was not able to obtain and compare baseline patient characteristics between the included trials; it was also not able to estimate between-trial heterogeneity due to the small number of trials in the network. Consequently, there is substantial uncertainty around the ITC results and firm conclusions cannot be drawn on the relative efficacy between pembrolizumab plus chemotherapy and relevant comparators, aside from direct evidence provided by the KEYNOTE-355 study.

Other Relevant Evidence

No long-term extension studies or other relevant studies were included in the sponsor’s submission to CADTH.

Conclusions

One pivotal study (KEYNOTE-355) and 1 sponsor-submitted ITC provided evidence for this review. The OS and PFS benefits observed with pembrolizumab plus chemotherapy versus placebo plus chemotherapy in patients with a PD-L1 CPS of 10 or higher in the KEYNOTE-355 trial were statistically significant, considered clinically meaningful by the clinical experts consulted by CADTH, and aligned with the outcomes important to patient groups. The secondary outcomes, ORR and DOR, were supportive of the observed OS and PFS results. There was no signal suggesting that the addition of pembrolizumab to chemotherapy resulted in a significant decrease in HRQoL from baseline to week 15. The results of the submitted ITC, which compared the efficacy of pembrolizumab plus nab-paclitaxel or paclitaxel to other comparators and results, suggested that OS favoured pembrolizumab plus chemotherapy in a comparison to nab-paclitaxel or paclitaxel, carboplatin, and docetaxel, and PFS favoured pembrolizumab plus chemotherapy in a comparison to nap-paclitaxel and/or paclitaxel, but not to other treatments. However, no firm conclusions could be drawn from the ITC results based on several limitations. No new safety concerns were identified for the use of pembrolizumab plus chemotherapy for the treatment of locally recurrent inoperable or metastatic TNBC. The clinical experts stated that there is experience using pembrolizumab for other indications, and oncologists are familiar with AEs due to pembrolizumab.

Introduction

Disease Background

Breast cancer is the most common cancer among females, with approximately 27,700 new cases reported in 2021.⁸ TNBC is an invasive form of breast cancer affecting 10% to 20% of patients.^1,2 TNBC is most common in women under 40 years, Black females, and females with a BRCA-1 mutation.^2,3 TNBC is distinguished by the absence of an ER, PR, and little to no expression of the HER2 gene.^3-5 Breast cancer is clinically categorized into early-stage breast cancer (stage I to II), locally advanced (stage II to III), and metastatic breast cancer (stage IV).¹⁸ Diagnosis of mTNBC (stage IV) is based on imaging tests (i.e., mammogram, breast ultrasound, or MRI), clinical symptoms, and biopsy to confirm breast cancer, with negative results on all 3 tests (ER, PR, and HER2), as well as PD-L1 status.⁹ Approximately 5% of patients with TNBC are initially diagnosed with metastatic disease.¹⁹ The predicted 5-year survival is 12% for mTNBC compared to 77% for all TNBC and 89% for all breast cancer.^7,8

TNBC has the same signs and symptoms as those reported with other common breast cancers, including breast changes such as firm or hard lumps, a lump in the armpit, changes in breast size and shape, changes to the nipple, and discharge from the nipples. Additional symptoms including bone pain, weight loss, nausea, appetite loss, shortness of breath, cough, headache, double vision, and muscle weakness may manifest with tumour size increase or spread to other organs.^3,5 Cases of mTNBC differ from other types of invasive breast cancer in that mTNBC tends to grow and spread faster, has fewer treatment options, and tends to have a worse prognosis.³ Tumours in mTNBC often metastasize to the bloodstream, brain, and lungs rather than the bones and lymph nodes, which is common in breast cancers that are not triple-negative.²⁰ The clinician groups consulted during the review emphasized that patients with TNBC have a higher risk of recurrence and death compared to those with other types of breast cancer. More than 50% of patients with mTNBC (stage IV) are likely to experience recurrence, of whom 37% will die in the first 5 years following surgery.¹

Breast cancers can be diagnosed during screening exams using mammograms. In most Canadian provinces and territories, organized screening via mammography at 2- to 3-year intervals is available for females aged 50 to 74 years with an average risk of breast cancer. Imaging techniques available include breast ultrasounds and radionuclide imaging. Other diagnostic methods include tissue biopsy examinations, hormone receptor status testing, HER2-status testing, complete blood count, blood chemistry tests, tumour marker tests, X-rays, bone scans, and ductography. A repeat biopsy to reassess HER2, ER, and PR, and PD-L1 status is recommended for patients with mTNBC.⁹

Standards of Therapy

The clinical experts and clinician groups consulted during the CADTH review indicated that the standard approach for mTNBC is sequential cytotoxic chemotherapy, including paclitaxel, nab-paclitaxel, docetaxel, gemcitabine, carboplatin, cisplatin, and vinorelbine. Typically, monotherapy is used rather than combination therapy, with the exception of gemcitabine, which is frequently combined with a platinum drug. The median OS for patients with mTNBC treated with conventional chemotherapy is 9 months to 13 months.^10-13 Often patients with more progressive disease are treated with combination chemotherapies. No targeted or immunotherapy has yet become standard of care in this population. Pembrolizumab is currently available through a compassionate access program as noted by the KEYNOTE 355 trial indications in the first-line setting.

Preferred treatment options outlined in the National Comprehensive Cancer Network guidelines for patients with TNBC include anthracyclines (doxorubicin or liposomal doxorubicin), taxanes (paclitaxel), and antimetabolites (capecitabine or gemcitabine). For patients with recurrent unresectable or metastatic TNBC who have received 2 prior therapies, the recommended systemic therapy is sacituzumab govitecan,¹⁹ which is currently the subject of negotiations by the pan-Canadian Pharmaceutical Alliance.

According to the clinical experts consulted for this review, the most important goal in this context is to improve survival, disease symptoms, function, and QoL. Tumour response for a prolonged duration has also been advocated as being important for patients, especially when this translates to improvements in QoL.

Drug

Pembrolizumab is an immunoglobulin G4 monoclonal antibody against PD-1. By inhibiting the PD-1 receptor from binding to its ligands expressed on T cells, pembrolizumab restores the cytotoxic T-cell effector function.¹⁴ Combinations of pembrolizumab and chemotherapy have been studied in an effort to enhance the antitumour activity of single drugs by creating a tumour microenvironment that stimulates responses to immunotherapy.¹⁴

Pembrolizumab underwent a standard review at Health Canada and obtained a Notice of Compliance with conditions on November 19, 2021, for the treatment (in combination with chemotherapy) of adult patients with locally recurrent unresectable or mTNBC who have not received prior chemotherapy for metastatic disease, and whose tumours express PD-L1 as determined by a validated test (CPS ≥ 10).²¹ The sponsor’s reimbursement request is line with the Health Canada indication. On November 13, 2020, the FDA authorized pembrolizumab in combination with chemotherapy for the treatment of patients with locally recurrent unresectable or mTNBC whose tumours express PD-L1 as determined by an FDA-approved test (CPS ≥ 10). Pembrolizumab has been previously reviewed by CADTH for other indications.

The Health Canada–recommended dosage of pembrolizumab in combination with chemotherapy is either 200 mg IV every 3 weeks or 400 mg IV every 6 weeks until disease progression, unacceptable toxicity, or up to 24 months. The alternative dosage is 35 doses for 200 mg or 18 doses for 400 mg, whichever is longer, in patients without disease progression.¹⁴ Additionally, the product monograph states that atypical responses (i.e., an initial transient increase in tumour size or small new lesions within the first few months followed by tumour shrinkage) have been observed. Clinically stable patients with initial evidence of disease progression may remain on treatment until disease progression is confirmed.

Key characteristics of commonly used therapies for mTNBC are shown in Table 3.

Stakeholder Perspectives

Patient Group Input

This section was prepared by CADTH staff based on input provided by patient groups. The full patient group input is included in the Stakeholder section of this review.

Two patient groups, the CBCN and RBC, provided input for this review. The CBCN is a patient-directed charity and a member of the Canadian Cancer Action Network, whose mission is to ensure the best quality of care for all people in Canada affected by breast cancer through the promotion of information, education, and advocacy activities. RBC is a Canadian charity committed to educating, empowering, and advocating for system changes to improve the experience and outcomes of patients with breast cancer, especially for those in underserved groups such as those diagnosed at a younger age, with metastatic breast cancer, and who are systemically marginalized due to race, income, or other factors.

The CBCN collected patient input via 2 online surveys (survey 1: 2017, n = 14 patients in Canada with mTNBC; survey 2: 2012, n = 71 patients in Canada, and n = 16 caregivers of patients in Canada with metastatic breast cancer [mBC]; no respondents had direct experience with pembrolizumab), and a literature review of current studies and grey literature. The input from RBC was based on general observations and insights gathered through various activities (e.g., patient blogs, virtual support groups, working groups, patient advisory boards, peer-support networks, Instagram, and scientific advisory committee meetings); 1 online survey (September 2018 to April 2019, n = 78 patients with mBC); interviews (Zoom videoconferencing, 2022, n = 7 patients with TNBC; one-on-one interviews, 2022, n = 2 patients in Canada with mTNBC, 1 of whom had access to pembrolizumab and 1 of whom was excluded from the pembrolizumab trials due to prior cancer treatment); and meeting and written or online correspondence (2022, n = 1 patient with mTNBC). RBC also shared 2 testimonials from interviews with 2 patients in Canada.

In the CBCN submission, patients emphasized the negative impacts of mBC symptoms such as fatigue (54%), insomnia (39%), and pain (37%). According to the 2017 CBCN survey, of the 14 patients who indicated that they are living with mTNBC, the majority of respondents (10) experienced metastases to their lungs, followed by metastases to other parts of their bodies (6), their bones (5), their liver (3), and their brain (2). These symptoms and metastases impose a heavy physical, emotional, psychosocial, and financial toll, and negatively affect HRQoL. The input provided by RBC indicated that, particularly for patients who are diagnosed at a young age, TNBC may have detrimental effects on patients’ well-being in terms of fertility, childcare, relationships, body image, social activities, employment, and mental health.

According to the 2017 CBCN survey, all 14 patients had been or were currently being treated with chemotherapy, 11 patients previously had surgery, 12 had or were receiving radiation therapy, and 2 had or were currently receiving hormone therapy. The input from patient groups acknowledged that currently available treatments for mBC have only been shown to prolong the progression-free period and emphasized the decreasing response rates in later lines of therapy; while the disease will eventually progress, patients seek to live their remaining months and years with the best possible HRQoL. Patients from both groups identified an unmet need for new treatments for mBC that can prolong survival, control disease, and maintain QoL. The RBC submission suggested that patients value long-term health outcomes over immediate concerns such as reducing symptoms or managing side effects. Patient respondents from CBCN expected new treatments to improve disease control, delay disease progression, improve QoL, have minimal side effects, and be affordable and easy to access.

One patient with mTNBC from RBC who had direct experience with pembrolizumab treatment reported that the drug had helped control the disease, shrink tumour size, and improve HRQoL.

Clinician Input

Input From Clinical Experts Consulted by CADTH

All CADTH review teams include at least 1 clinical specialist with expertise in the diagnosis and management of the condition for which the drug is indicated. Clinical experts are a critical part of the review team and are involved in all phases of the review process (e.g., providing guidance on the development of the review protocol, assisting in the critical appraisal of clinical evidence, interpreting the clinical relevance of the results, and providing guidance on the potential place in therapy). The following input was provided by 2 clinical specialists with expertise in the diagnosis and management of mTNBC.

Unmet Needs

Patients with mTNBC have relatively poor prognoses compared to those with other breast cancer subtypes. Therefore, improvement in survival and/or its quality remain the most important unmet need for many years. Tumour response has also been described as being important for patients, particularly when associated with improvements in QoL. Effective treatments with better tolerance and convenience are sought by patients and oncologists treating mTNBC.

Place in Therapy

According to the clinical experts, pembrolizumab’s unique mechanism of action would create a shift in the treatment paradigm for this population. Pembrolizumab would be used in the first-line setting for the treatment of mTNBC. In the opinion of the clinical experts, pembrolizumab with chemotherapy would be the new standard of care and will replace current treatment (e.g., chemotherapy alone), rather than being reserved for patients that are intolerant to existing treatments. The clinical expert would not recommend patients try other treatments if they present with locally advanced mTNBC that has not previously been treated for that setting.

Patient Population

The clinical experts advised that patients best suited for treatment with pembrolizumab combined with chemotherapy are those who meet the inclusion criteria (e.g., PD-L1 score) of the KEYNOTE-355 study. Patients are identified though PD-L1 antigen testing. Those with a CPS of 10 of higher will be eligible for pembrolizumab. According to the clinical experts, mTNBC is easily diagnosed at most Canadian cancer centres, and no major issues are expected regarding misdiagnosis. Presymptomatic patients can be considered for treatment with pembrolizumab as these represent the majority of the cases. The clinical experts advised that patients least suitable for pembrolizumab included those who are unfit, frail, have a poor ECOG PS, or have active autoimmune disease regardless of tumour size and number nodes. According to the clinical experts, apart from a PD-L1 expression score of greater than 10%, which is a necessary criterion, there are no other reliable predictors of benefit.

Assessing Response to Treatment

The clinical experts indicated that radiological and clinical investigations are used to evaluate response on a regular basis. The experts also noted that PD-L1 status has been predictive of benefit in metastatic cases. The most common assessment modality is the shrinkage of tumour on clinical and/or radiological tests. Other assessments of clinical response include improvement (or no deterioration) of symptoms and patient ability to function and survival. The clinical experts advised that, as long as there is a response, its magnitude is not typically used in treatment decisions. Typical clinical practice is to assess every 3 to 4 cycles of cytotoxic chemotherapy (approximately every 3 months).

Discontinuing Treatment

The clinical experts stated that disease progression (tumour enlargement unless pseudoprogression is suspected) and the occurrence of intolerable AEs would be considered when deciding whether to discontinue treatment.

Prescribing Conditions

The clinical experts advised that prescribing depends on the comfort level of the centre with administration of the drug and handling side effects. Given that pembrolizumab has already been used for a few years in other diseases, most Canadian centres should have a relatively sophisticated understanding of how to deliver this drug safely.

Clinician Group Input

This section was prepared by CADTH staff based on the input provided by clinician groups. The full clinician group input is included in the Stakeholder section of this review.

Input from the following 3 clinician groups was provided: the OH-CCO Breast Cancer Drug Advisory Committee (providing input from 1 clinician), breast cancer medical oncologists in Canada, led by TOHCC (providing input from 6 medical oncologists), and Provincial Breast Tumour Group (Alberta) (providing input from 2 medical oncologists). The OH-CCO’s Drug Advisory Committee provides timely evidence-based clinical and health-system guidance on drug-related issues in support of Cancer Care Ontario’s mandate, including the Provincial Drug Reimbursement Programs and the Systemic Treatment Program. The TOHCC group consists of medical oncologists throughout Canada with an expertise in breast oncology and systemic therapies in advanced TNBC, in particular. The Provincial Breast Tumour Group (Alberta) is responsible for treating patients with TNBC in Alberta.

The clinician groups indicated that the current first-line treatment for mTNBC, cytotoxic chemotherapy (taxane, platinum, and anthracycline), showed limited benefit in delaying disease progression and improving both duration and QoL. The clinician groups identified an unmet need for more effective and tolerable treatment options in the metastatic setting. The OH-CCO noted that the lack of CPS testing within the province may delay the delivery of test results and the diagnosis of TNBC. The clinician groups mentioned that pembrolizumab would fit in the first-line metastatic setting in combination with chemotherapy in treating patients with TNBC. The clinician groups emphasized that pembrolizumab with chemotherapy would be expected to shift the current treatment paradigm by replacing chemotherapy alone. The patients suitable for pembrolizumab with chemotherapy identified by the clinician groups were those who expressed PD-L1 (CPS ≥ 10), with an ECOG PS of 0 to 2 and a time from completion of adjuvant treatment of greater than 6 months, which were aligned with inclusion criteria for the KEYNOTE-355 trial. Patients least suitable were those who do not have a CPS of 10 or higher, those in other lines of therapy, or those who had received adjuvant and/or neoadjuvant pembrolizumab for early-stage disease within 12 months and with a high risk of adverse toxicity related to immunotherapy.

Clinician groups agreed that treatment response should be measured using assessments based on symptoms, laboratory markers, radiographic scans, and tumour measurements. The appropriate interval for assessments is 3 months. Improvement of organ function (bone, liver, and lung) and severity of symptoms, maintenance or improvement of performance status, tumour radiographic response with either stabilization of disease or response by RECIST 1.1 were considered clinically meaningful responses. When deciding whether to discontinue pembrolizumab with chemotherapy treatment, disease progression, intolerable or dangerous toxicity (grade 3 or higher immune-mediated toxicity) and patient preference or refusal should be considered. Clinician groups agreed that outpatient settings, such as hospitals or specialty clinics that have oncology specialists (medical oncologists, chemotherapy nurses, or oncology pharmacists) to administer systemic cancer therapies and monitor and manage treatment-related toxicities, are appropriate. In addition, the clinician groups pointed out that the combination of pembrolizumab and chemotherapy has been considered a new standard of care by internationally accepted guidelines, has been shown to be well tolerated with a manageable toxicity profile, and is highly accepted and valued by patients in Canadian clinical practice; it is therefore imperative that Canadians have access to this treatment.

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may affect their ability to implement a recommendation. The implementation questions and corresponding responses from the clinical experts consulted by CADTH are summarized in Table 4.

Table 4: Summary of Drug Plan Input and Clinical Expert Response

CPS = combined positive score; ECOG PS = Eastern Cooperative Oncology Group Performance Status; ER = estrogen receptor; mTNBC = metastatic triple-negative breast cancer; PAG = Provincial Advisory Group; PD-L1 = programmed cell death 1 ligand 1; pERC = CADTH pan-Canadian Oncology Drug Review Expert Review Committee; PR = progesterone receptor; TNBC = triple-negative breast cancer.

Clinical Evidence

The clinical evidence included in the review of pembrolizumab (Keytruda) is presented in 2 sections. The first section, the systematic review, includes pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those studies that were selected according to an a priori protocol. The second section includes a submitted ITC. No indirect evidence was identified in the literature that met the selection criteria specified for this CADTH review.

Systematic Review (Pivotal and Protocol-Selected Studies)

Objectives

To perform a systematic review of the beneficial and harmful effects of pembrolizumab (200 mg IV administered every 3 weeks or 400 mg every 6 weeks) for the treatment of adult patients with locally recurrent unresectable or mTNBC who have not received prior chemotherapy for metastatic disease and whose tumours express PD-L1 as determined by a validated test (CPS ≥ 10).

Methods

Studies selected for inclusion in the systematic review included pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those meeting the selection criteria presented in Table 5. Outcomes included in the CADTH review protocol reflect outcomes considered to be important to patients, clinicians, and drug plans.

The literature search for clinical studies was performed by an information specialist using a peer-reviewed search strategy according to the PRESS Peer Review of Electronic Search Strategies checklist.²³

Published literature was identified by searching the following bibliographic databases: MEDLINE All (1946–) via Ovid and Embase (1974–) via Ovid. All Ovid searches were run simultaneously as a multifile search. Duplicates were removed using Ovid deduplication for multifile searches, followed by manual deduplication in EndNote. The search strategy comprised both controlled vocabulary, such as the National Library of Medicine’s MeSH (Medical Patient Headings), and keywords. The main search concepts were Keytruda (pembrolizumab) and TNBC. Clinical trials registries searched included the US National Institutes of Health’s clinicaltrials.gov, WHO’s International Clinical Trials Registry Platform search portal, Health Canada’s Clinical Trials Database, and the European Union Clinical Trials Register.

Table 5: Inclusion Criteria for the Systematic Review

AE = adverse event; CPS = combined positive score; DOR = duration of response; ECOG PS = European Cooperative Oncology Group Performance Status; HRQoL = health-related quality of life; mTNBC = metastatic triple-negative breast cancer; ORR = overall response rate; OS = overall survival; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival; SAE = serious adverse event; TEAE = treatment-emergent adverse event; WDAE = withdrawal due to adverse event.

^aThese outcomes were identified as being of particular importance to patients in the input received by CADTH from patient groups.

No filters were applied to limit the retrieval by study type. Retrieval was not limited by publication date or by language. Conference abstracts were excluded from the search results. Appendix 1 provides detailed search strategies.

The initial search was completed on June 27, 2022. Regular alerts updated the search until the meeting of the CADTH pan-Canadian Oncology Drug Review Expert Committee on November 9, 2022.

Grey literature (literature that is not commercially published) was identified by searching relevant websites from the Grey Matters: A Practical Tool For Searching Health-Related Grey Literature checklist.²⁴ Included in this search were the websites of regulatory agencies (FDA and European Medicines Agency). Google was used to search for additional internet-based materials. Appendix 1 provides more information on the grey literature search strategy.

These searches were supplemented through contacts with appropriate experts. In addition, the sponsor of the drug was contacted for information regarding unpublished studies.

Findings From the Literature

One report of a single study was identified from the literature for inclusion in the systematic review (Figure 1). The included study is summarized in Table 6.

Figure 1: Flow Diagram for Inclusion and Exclusion of Studies

Table 6: Details of Included Study

AE = adverse event; ASCO = American Society of Clinical Oncology; AUC = area under the curve; CAP = College of American Pathologists; CHF = congestive heart failure; CIV = central imaging vendor; CNS = central nervous system; CPS = combined positive score; DCR = disease control rate; DOR = duration of response; ECOG = Eastern Cooperative Oncology Group Performance Status; EORTC QLQ-BR23 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire – Breast Module 23; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; HRQoL = health-related quality of life; IA1 = first interim analysis; IA2 = second interim analysis; ORR = objective response rate; OS = overall survival; PD-1 = programmed cell death 1 protein; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival; q.3.w. = every 3 weeks; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1; TNBC = triple-negative breast cancer.

Note: Data cut-off: June 15, 2021.

Source: Clinical Study Report¹⁵ and sponsor’s general submission materials.²⁶

Description of Study

The KEYNOTE-355 trial is an ongoing, phase III, randomized, multicentre, double-blind, 2-part, placebo-controlled trial. The primary objective of the trial is to evaluate the efficacy and safety of pembrolizumab plus chemotherapy versus placebo plus chemotherapy for patients with locally recurrent inoperable or metastatic TNBC not been previously treated with chemotherapy for metastatic disease. The study was conducted in 2 parts (Figure 2). Part 1 was the safety run-in (N = 30) and part 2 was the efficacy evaluation (N = 847). For the purpose of this CADTH review, only part 2 was evaluated. The 2 coprimary efficacy outcomes of OS and PFS were assessed in all patients, and patients with PD-L1–positive tumours. Secondary outcomes included ORR, DOR, disease control rate, and HRQoL and were assessed in all patients and patients with PD-L1–positive tumours. The study was considered to have met its primary end point if the combination of pembrolizumab and chemotherapy was superior to placebo and chemotherapy in either PFS or OS in either all patients or in patients with PD-L1–positive tumours (CPS ≥ 1 or CPS ≥ 10) at either an interim analysis or the final analysis (OS only). In total, 75.1% and 38.1% had a tumour tissue PD-L1 expression score of CPS of 1 or higher and a CPS of 10 or higher, respectively. Given that the indication for the sponsor-submitted reimbursement request is for patients with a PD-L1 expression (CPS) of 10 or higher, this review focuses on patients with a PD-L1 CPS of 10 or higher and all patients. Efficacy results for patients with a PD-L1 CPS of 1 or higher can be found in Appendix 2. In total, 1,372 patients had been screened, of whom 847 were randomized via an interactive voice recognition system in the ITT population in a 2:1 ratio between the pembrolizumab plus chemotherapy and the placebo plus chemotherapy groups. In total, 566 patients were randomized to receive pembrolizumab plus chemotherapy and 281 to receive placebo plus chemotherapy. Randomization was stratified based on 3 factors:

• type of chemotherapy on study (paclitaxel or nab-paclitaxel or gemcitabine and carboplatin)

• PD-L1 expression at baseline (CPS ≥ 1 or < 1)

• prior treatment with the same class of chemotherapy in the neoadjuvant or adjuvant setting (yes or no).

Patients already in the screening phase may have been enrolled even after the maximum sample size had been reached. A total of 843 patients received at least 1 dose of the study intervention and were included in the population of all patients as treated (APaT) (562 in the pembrolizumab plus chemotherapy group and 281 in the placebo plus chemotherapy group).

The KEYNOTE-355 trial was initiated in August 2016 and has 251 participating centres across 29 countries in North America (5 sites in Canada), South America, Europe, Asia, and Australia. Both patients and investigators were blinded to study treatments administered during the trial. All patients enrolled were female. Majority of patients were younger than 65 years of age, white, postmenopausal, and had an ECOG PS of 0.

Populations

Inclusion and Exclusion Criteria

Eligible patients included those at least 18 years of age with locally recurrent inoperable TNBC not previously treated with chemotherapy and that could not be treated with curative intent, or mTNBC not previously treated with chemotherapy. Patients were to have completed treatment for stage I to III breast cancer, if indicated, with at least 6 months between the completion of treatment with curative intent and first documented local or distant disease recurrence. Patients who received taxane, gemcitabine, or platinum drugs in the (neo)adjuvant setting could have been treated with the same class of chemotherapy and could participate if at least 12 months had elapsed between the completion of treatment with curative intent and first documented local or distant disease recurrence. Patients were required to have centrally confirmed TNBC as defined by American Society of Clinical Oncology or College of American Pathologists guidelines, measurable disease according to RECIST 1.1 by local radiology review, an ECOG PS of 0 or 1, and a predicted life expectancy of 12 weeks or longer from randomization. Patients were enrolled regardless of PD-L1 biomarker status but were required to provide a tumour tissue sample for central biomarker analysis.

Figure 2: KEYNOTE-355 Study Design

CNS = central nervous system; CPS = combined positive score; PD-L1 = programmed cell death 1 ligand 1; TNBC = triple-negative breast cancer.

Notes: Treatment was to continue until confirmation of progressive disease or death. Patients from part 1 (safety run-in) were not included in part 2 (efficacy evaluation). Data cut-off date: June 15, 2021.

Source: Clinical Study Report.^15,27

Baseline Characteristics

All patients were female and most were younger than 65 years of age; white, non-Hispanic or non-Latino; postmenopausal; and had an ECOG PS of 0. Most patients entered the study with mTNBC (recurrent [66.9%] or de novo [29.8%] metastatic disease). More patients in the pembrolizumab group (21%) versus the placebo plus chemotherapy group (16.5%) had a disease-free interval of less than 12 months. Most patients (78.3%) who received prior neoadjuvant and/or adjuvant treatment were not treated with the same class of chemotherapy in the study. Most patients (75.1%) had a tumour-tissue PD-L1 CPS of 1 or higher, and 38.1% of patients had a tumour-tissue PD-L1 CPS of 10 or higher. In the ITT population for patients with a PD-L1 CPS of 10 or higher, demographics and other baseline characteristics data were generally well balanced between the 2 intervention groups and consistent with those of the APaT population. A majority of patients (approximately 54%) received gemcitabine and carboplatin as their chemotherapy in addition to pembrolizumab and placebo. A summary of baseline and demographic characteristics of patients enrolled in the KEYNOTE-355 study at final analysis is presented in Table 7.

Table 7: Summary of Baseline Characteristics at Final Analysis — ITT Population

CPS = combined positive score; ECOG PS = Eastern Cooperative Oncology Group Performance Status; ITT = intention-to-treat; IVRS = interactive voice recognition system; PD-L1 = programmed cell death 1 ligand 1.

Notes: Breast, chest wall, and lymph node also include locally recurrent lesions. Data cut-off: June 15, 2021.

Source: Clinical Study Report.¹⁵

Interventions

Patients were randomized in a 2:1 ratio to the 2 study arms:

• pembrolizumab (200 mg) every 3 weeks plus 1 of the following chemotherapies:

  • nab-paclitaxel (100 mg/m² on days 1 and 8, and 15 every 28 days)

  • paclitaxel (90 mg/m² on days 1 and 8, and 15 every 28 days)

  • gemcitabine and carboplatin (gemcitabine 1,000 mg/m² with carboplatin at the area under the curve 2 on days 1 and 8 every 21 days)

• placebo (normal saline) every 3 weeks plus 1 of the following chemotherapies as described in group 1 (e.g., nab-paclitaxel, paclitaxel, or gemcitabine and carboplatin), using the same dosing and schedule.

The chemotherapy for which each patient was eligible was determined by the previous chemotherapy use and the time elapsed since that use, as described in the inclusion criteria. If chemotherapy options included more than 1 regimen, the choice of study chemotherapy was at the physician’s discretion.

Treatment Administration

Chemotherapies were administered by an unblinded pharmacist, but patients, investigators, other study site staff, and the sponsor were blinded to the administration of pembrolizumab or placebo.

Pembrolizumab was administered at a 200 mg fixed dose via IV every 3 weeks until confirmed disease progression or cessation of study treatment. Patients were followed closely for unacceptable toxicities for 21 days in patients receiving gemcitabine and carboplatin chemotherapy or 28 days in patients receiving either taxane chemotherapy (i.e., paclitaxel or nab-paclitaxel) after the first administrations. According to the protocol of the KEYNOTE-355 trial, patients had to discontinue from pembrolizumab or placebo once they completed 35 administrations of pembrolizumab or placebo. However, patients were allowed to continue chemotherapy treatment at the investigator’s discretion.

Dose Modification and Interruptions

Dose reductions were not permitted for pembrolizumab. Dose modifications were conducted in a stepwise pattern and were based on maximum toxicity experienced during the previous treatment. Toxicity (except for alopecia) needed to resolve to no worse than grade 1 or baseline before resuming treatment with the same drug(s). A maximum of 2 dose modifications per chemotherapy (if applicable) for toxicities throughout the course of the study was permitted. Dose-modification strategies for other chemotherapy drugs were allowed in the study protocol. Local guidelines and practices were recommended if dosing was different from the sponsor’s recommendations. Dose reductions were not permitted for pembrolizumab.

In the event of drug-related toxicity, including severe or life-threatening AEs, pembrolizumab was withheld. Dosing interruptions were permitted in the case of medical or surgical events or logistical reasons not related to study therapy (e.g., elective surgery, unrelated medical events, patient vacation, and/or holidays). Chemotherapy administration was allowed to be interrupted due to AEs for a maximum of 4 weeks; pembrolizumab was allowed to be interrupted due to AEs for a maximum of 12 weeks. Patients were to resume study therapy within 3 weeks of the scheduled interruption, unless otherwise discussed with the sponsor.

Treatment Discontinuation Criteria

Patients were allowed to discontinue 1 or more components of the study treatment at the discretion of the investigator in case of undesirable effects. Study treatment was continued if 1 of the following occurred: central imaging vendor (CIV)-verified progressive disease as defined by RECIST 1.1, unacceptable toxicity, intercurrent illness that necessitated discontinuation of study treatment, investigator’s decision to withdraw the patient, pregnancy, patient noncompliance with study treatment or procedure requirements, withdrawal of consent to treatment, death, the end of the study, or other administrative reasons requiring cessation of treatment.

Concomitant Medications

In the trial, all treatments that the investigator considered necessary for a patient’s welfare may be administered at the discretion of the investigator in keeping with the community standards of medical care. All concomitant medications received within 30 days before randomization while on study treatment, and up to 30 days after the last dose of study treatment, were recorded. Rescue medications included the use of oral or IV treatment with corticosteroids, as well as additional anti-inflammatory drugs, if symptoms did not improve with administration of corticosteroids.

Patients were prohibited from receiving the following therapies during the screening, treatment, and re-treatment phase of this study:

• antineoplastic systemic chemotherapy or biological therapy

  • immunotherapy not specified in this protocol

  • chemotherapy not specified in this protocol

  • investigational drugs other than pembrolizumab

  • radiation therapy within 2 weeks before randomization; however, radiation therapy for a symptomatic solitary lesion or the brain may be allowed after consultation with sponsor (except during screening).

• herbal supplements

• live vaccines within 30 days before randomization and while participating in the study; examples of live vaccines include, but are not limited to: measles, mumps, rubella, varicella, herpes zoster, yellow fever, rabies, Bacillus Calmette–Guérin, and typhoid (oral) vaccines; seasonal influenza vaccines for injection are generally killed virus vaccines and are allowed; however, intranasal influenza vaccines (e.g., Flu-Mist) are live attenuated vaccines, and are not allowed

• glucocorticoids for any purpose other than the following:

  • to modulate symptoms from an AE of suspected immunologic etiology

  • inhaled steroids for management of asthma

  • physiologic doses of prednisone 10 mg (or equivalent) per day

  • use of prophylactic corticosteroids to avoid allergic reactions (e.g., to paclitaxel and/or IV contrast dye) is permitted.

• any medication prohibited in combination with chemotherapy as described in the respective product labels for nab-paclitaxel, paclitaxel, gemcitabine, and carboplatin.

Outcomes

A list of efficacy end points identified in the CADTH review protocol that were assessed in the clinical trials included in this review is provided in Table 8 and summarized in the following section. A detailed discussion and critical appraisal of the outcome measures is provided in Appendix 3.

Table 8: Summary of Outcomes of Interest Identified in the CADTH Review Protocol

AE = adverse event; EORTC QLQ-BR23 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire – Breast Model 23; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; EQ VAS = EQ visual analogue scale; SAE = serious adverse event.

Source: Clinical Study Report.¹⁵

Primary Outcomes

Overall Survival

Overall survival was a coprimary outcome of the KEYNOTE-355 trial. OS was defined as the time from randomization to death due to any cause. Patients without documented death at the time of the analysis were censored at the date of the last follow-up.

Health-Related Quality of Life

HRQoL measures were a secondary outcome of the KEYNOTE-355 trial. HRQoL measures were assessed using the EORTC QLQ-C30, EORTC QLQ-BR23, and EQ-5D questionnaires. A detailed discussion and critical appraisal of the EORTC QLQ-C30 and EORTC QLQ-BR23 is available in Appendix 3; the EQ VAS is a commonly used and validated measure in oncology trials. The primary patient-reported outcome analysis time point is defined as the latest time point at which the completion and compliance rates are still high enough based on blinded data review (approximately 60% completion and 80% compliance rates). The resulting time frame was 15 weeks in this case; the minimum completion and compliance rates were required to minimize the impact of missing data assumptions on patient-reported outcome analysis outcomes. The key patient-reported outcome end point was the mean score changes from baseline to the primary analysis time point in EORTC QLQ-C30 global health status and QoL score.

Supportive patient-reported outcome end points included the following: the mean score changes from baseline to the primary analysis time point in EQ VAS as measured by the EQ-5D questionnaire; the mean score changes from baseline to the primary analysis time point for the QLQ-C30 physical and emotional functioning scales; time to deterioration defined as the time from start of treatment to first onset of 10 points or more worsening from baseline the EORTC QLQ-C30 global health status and QoL score and physical and emotional functioning scales; the mean score changes from baseline to the primary analysis time point for other items or scales of the EORTC QLQ-C30 and EORTC QLQ-BR23 as appropriate.

The patient-reported outcomes were assessed every 3 weeks (e.g., every cycle of treatment). After the third cycle and until the end of year 1, patient-reported outcomes were collected every third cycle (every 9 weeks) until progressive disease, while the patient was receiving study treatment. During year 2, they will occur every fourth cycle (every 12 weeks) until progressive disease, while the patient is receiving study treatment. For the analysis, patient-reported outcomes assessed at visits of “end of treatment” and “safety follow-up” will be mapped into different time points according to the actual visit time. If there are multiple patient-reported outcome collections within the time window of a specific visit, the collection closest to the target day will be used in the analysis.

The EORTC QLQ-C30 is a multidimensional, cancer-specific, self-administered measure of HRQoL. The EORTC QLQ-C30 is composed of both multi-item scales and single-item measures. These include 5 functional scales (physical, role, cognitive, emotional, and social), 3 symptom scales (fatigue, pain, and nausea and vomiting), a global health status or QoL scale, and 6 single items assessing additional symptoms commonly reported by cancer patients (dyspnea, loss of appetite, insomnia, constipation and diarrhea) as well as perceived financial impact of the disease.²⁸ The EORTC QLQ-C30 uses a 1-week recall period to assess functional status and symptoms. All scales and single-item measures are scored from 0 to 100. Scale sum scores are transformed such that a high score on the functional scales represents a high or healthy level of functioning, a high score on the symptom scales represents a high level of symptomatology and a high score on the global health status or QoL scale represents a high HRQoL.²⁹

The EORTC QLQ-BR23 includes an additional 23 items that are specific to breast cancer, with 5 multi-item scales assessing systemic therapy side effects, group symptoms, breast symptoms, body image, and sexual functioning, as well as single items assessing sexual enjoyment, hair loss, and future perspective.³⁰ All scales range in score from 0 to 100, and higher scores represent higher responses. A high score for functioning represents high functioning while a high symptom score represents a high symptom burden. The EORTC QLQ-BR23 is not a freestanding instrument as it does not measure a multidimensional construct of body image; instead, it is administered in conjunction with the EORTC QLQ-C30.³⁰

The 5-Level EQ-5D questionnaire consists of 2 separate elements: the utility score (or descriptive system) that captures health states across 5 dimensions: mobility; self-care; usual activities, pain and/or discomfort; and anxiety and/or depression. Each unique health state is defined by combining response levels from each of the 5 dimensions. The second component, a vertical visual analogue scale, records the respondent’s self-rated health on a scale ranging from 0 to 100, for which the end points are “best imaginable health state” (100) and “worst imaginable health state” (0).²⁶

Progression-Free Survival

PFS was a coprimary outcome of the KEYNOTE-355 trial. PFS was defined as the time from randomization to the first documented disease progression as defined by RECIST 1.1 based on assessments by a CIV or death due to any cause, whichever occurs first. Table 9 provides PFS censoring rules.

Patients were evaluated with radiographic imaging at baseline, after randomization at weeks 8, 16, and 24, then every 9 weeks for the remainder of year 1, and every 12 weeks thereafter to assess response to treatment.

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

PD = progressive disease; PFS = progression-free survival.

Source: Clinical Study Report.¹⁵

Objective Response Rate

The ORR was a secondary outcome of the KEYNOTE-355 trial. The ORR was defined as the proportion of the patients in the analysis population with a complete or partial response. Responses are based on assessments by a CIV as defined by RECIST 1.1.

Duration of Response

The DOR was a secondary outcome of the KEYNOTE-355 trial. For patients who demonstrate a complete or partial response, DOR was defined as the time from first documented evidence of a complete or partial response until disease progression or death due to any cause, whichever occurs first, based on CIV assessments according to RECIST 1.1. Table 10 provides DOR censoring rules.

Sensitivity analyses were performed for DOR based on site investigator or local radiology review.

Table 10: Censoring Rules for the Analysis of Duration of Response

PD = progressive disease.

Note: A missed disease assessment includes any assessment that is not obtained or is considered inadequate for evaluation of response.

Source: Clinical Study Report.¹⁵

Harms Outcomes

Safety parameters such as incidence of AEs, SAEs including fatal SAEs, immune-related AEs, laboratory abnormalities, rates of dose interruption and discontinuation due to AEs, and events of clinical interest were monitored throughout the trial.

Schedule of Events

The end-of-treatment visit occurred at the time all study treatments are discontinued for any reason. The safety follow-up visit occurred 30 days after the last dose of all study treatments or before the initiation of a new anticancer treatment, whichever came first. Patients who are eligible for re-treatment with pembrolizumab may have up to 2 safety follow-up visits: 1 after the treatment period and 1 after the second course phase. Patients who discontinued all study treatments for a reason other than disease progression were moved into the disease status follow-up phase and were assessed according to the already-followed tumour-imaging schedule.

Statistical Analysis

Sample Size and Power Calculation

A total sample size of approximately 828 patients was initially planned for the study. It was expected that approximately 664 OS events among all patients and approximately 240 OS events among patients with a CPS of 10 or higher would be observed at the final analysis.

• PFS in all patients: At IA2, the trial had approximately 89% power at a 1-sided 0.111% alpha level if the true HR is 0.70. At IA2, with approximately 634 events, the HR at boundary for success was approximately 0.77. At IA2, PFS in all patients could only be tested if both hypotheses of PFS in patients with a CPS of 10 or higher and PFS in patients with a CPS of 1 or higher were supported.

• PFS in patients with a CPS of 10 or higher: at IA2 the trial had approximately 86% power at a 1-sided 0.411% alpha level if the true HR was 0.60. At IA2, with approximately 235 events, the HR at boundary for success was approximately 0.69 (approximately 2.4 months improvement over control median PFS of 5.5 months).

• OS in all patients: the trial had approximately 60% power at a 1-sided 0.75% alpha level, if the true HR was 0.80. With approximately 664 events, the HR at boundary for success at final analysis was approximately 0.81 (approximately 4.0 months improvement over control median OS of 17.5 months). After IA1, OS in all patients could be tested if the hypothesis of OS in patients with a CPS of 1 or higher was supported.

• OS in patients with a CPS of 10 or higher: the trial had approximately 79% power at a 1-sided 1.011% alpha level if the true HR was 0.65. With approximately 240 events, the HR at boundary for success at final analysis was approximately 0.72 (approximately 6.8 months improvement over control median OS of 17.5 months).

Primary Outcome Analyses

Overall Survival

The nonparametric Kaplan-Meier method was used to estimate the survival curves. The treatment difference in survival was assessed by the stratified log-rank test. A stratified Cox proportional hazards model with the Efron method of tie handling was used to assess the magnitude of the treatment difference (i.e., the HR). The HR and its 95% CI from the stratified Cox model with a single treatment covariate was reported. The stratification factors used for randomization were applied, as stratification factors for analysis, to both the stratified log-rank test and the stratified Cox model. Patients in the placebo plus chemotherapy group were expected to discontinue treatment earlier than would patients in the pembrolizumab plus chemotherapy group and were not allowed to cross over to the pembrolizumab plus chemotherapy group; however, they were allowed to be treated with another anti–PD-1 drug following the verification of progressive disease by blinded CIV. Exploratory analyses (adjustments) were performed to determine the impact on patients in the placebo group who were treated with an anti–PD-1 drug following progressive disease. Sensitivity analyses for OS were conducted to check the robustness of the data using the ITT principle. Table 12 provides an overview of the statistical analyses of efficacy end points.

Progression-Free Survival

The nonparametric Kaplan-Meier method was used to estimate the PFS curve in each treatment group. The treatment difference in PFS will be assessed by the stratified log-rank test. A stratified Cox proportional hazard model with the Efron method of tie handling will be used to assess the magnitude of the treatment difference (i.e., HR) between the treatment arms. The HR and its 95% CI, from the stratified Cox model with the Efron method of tie handling and with a single treatment covariate will be reported. The stratification factors used for randomization were applied, as stratification factors used for analysis, to both the stratified log-rank test and the stratified Cox model.

For patients who have progressive disease, the true date of disease progression will be approximated by the date of the first assessment at which progressive disease is objectively documented based on RECIST 1.1 as assessed by a CIV. Death is always considered a confirmed progressive-disease event. Patients who do not experience a PFS event will be censored at the last disease assessment.

In addition, 1 primary and 2 sensitivity analyses with a different set of censoring rules will be performed to assess the robustness of the PFS results. The first sensitivity analysis followed the ITT principle, whereby progressive diseases and/or deaths were counted as events regardless of missed study visits or initiation of new anticancer therapy. The second sensitivity analysis considered initiation of new anticancer treatment or discontinuation of treatment due to reasons other than complete response to be a progressive-disease event for patients without documented progressive disease or death. The censoring rules for primary and sensitivity analyses are summarized in Table 9. If a patient meets multiple criteria for censoring, the censoring criterion that occurs earliest was applied.

Multiplicity Testing

The multiplicity strategy was applied to the 6 primary hypotheses and the 2 secondary hypotheses of part 2: primary hypotheses of superiority of pembrolizumab plus chemotherapy compared to placebo plus chemotherapy in PFS and OS in all patients and in patients with PD-L1–positive tumours (for CPS ≥ 1 and CPS ≥ 10 separately), and secondary hypotheses of superiority of pembrolizumab plus chemotherapy compared to placebo plus chemotherapy in ORR in all patients and in patients with PD-L1–positive tumours (CPS ≥ 1).

The familywise type I error rate for this study was controlled at 2.5% (1-sided) across all 6 primary hypotheses on PFS and OS as well as 2 secondary hypotheses on ORR. A summary of the revised initial alpha allocation before any alpha reallocation is presented in Table 11. Based on emerging biomarker data external to this study, the initial alpha allocation among the 6 primary hypotheses and 2 secondary hypotheses was revised after IA1. The revision was based on emerging biomarker data external to the study (outlined in Amendment 5).

Table 11: Initial Alpha Allocation

IA1 = interim analysis 1; IA2 = interim analysis 2; ORR = objective response rate; OS = overall survival; PFS = progression-free survival.

Source: Clinical Study Report.¹⁵

Figure 3 displays the revised multiplicity strategy diagram of the study. The initial 1-sided alpha allocation for each hypothesis is shown in the rectangle representing the hypothesis. The weights for reallocation from each hypothesis to the others are represented in the numbers along the lines connecting hypotheses. Overall, 0.5% alpha was allocated to PFS end points, a 1.8% alpha was allocated to OS end points, and a 0.2% alpha was allocated to ORR end points.

Subgroup Analyses

To determine whether the treatment effect was consistent across various subgroups, the between-group treatment effect (with a nominal 95% CI) for the primary end points was estimated and plotted within each category of the following classification variables in all patients and in patients with PD-L1–positive tumours (CPS ≥ 1 and CPS ≥ 10):

• chemotherapy on study (nab-paclitaxel versus paclitaxel versus gemcitabine and carboplatin; taxane versus gemcitabine and carboplatin)

• tumour PD-L1 status (positive versus negative). Note: this subgroup analysis will only be conducted in the “all patients” population

• prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes versus no).

• prior (neo)adjuvant chemotherapy (yes versus no)

• prior (neo)adjuvant taxane treatment (yes versus no)

• prior (neo)adjuvant platinum treatment (yes versus no)

• menopausal status (for females only; premenopausal versus postmenopausal)

• age (< 65 years versus ≥ 65 years)

• geographic region (Europe, Israel, North America, or Australia versus Asia versus rest of world)

• ethnic origin (Hispanic versus non-Hispanic)

• ECOG PS (0 versus 1)

• HER2 status (2+ by immunohistochemistry versus 0 to 1+ by immunohistochemistry)

• disease-free interval (de novo metastasis versus < 1 2 months versus ≥ 12 months)

• number of metastatic sites (< 3 versus ≥ 3)

• visceral disease (yes versus no)

• lactate dehydrogenase (≥ 2.0 × upper limit of normal versus < 2.0 × upper limit of normal).

The subgroups that aligned with the CADTH review protocol included the following: ECOG PS, prior treatment with the same class of chemotherapy in the adjuvant or neoadjuvant setting, and chemotherapy on study. Only those subgroups identified in the CADTH review protocol are reported in the efficacy section.

Figure 3: Multiplicity Strategy for the KEYNOTE-355 Trial

CPS = combined positive score; final analysis = final analysis; FA = final analysis; H = hypothesis; IA1 = interim analysis 1; IA2 = interim analysis 2; IA3 = interim analysis 3; ORR = objective response rate; OS = overall survival; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival.

Note: The shaded boxes represent the alpha that has already been spent at IA1 and will be considered lost for future analyses. These alphas will no longer be reallocated to other hypotheses under the graphical approach, nor can they be used to account for correlation among group sequential tests within each end point across different time points.

^a Nominal alpha for testing will be calculated based on the Spiessens and Debois method accounting for correlation between ORR in all patients and ORR in patients with CPS ≥ 1. While the nominal alpha will be calculated and used for testing, the allocated alpha (0.001) will be passed to H7 when applicable.

^b H5 alpha = 0.0075, which includes 0.00036 already spent at IA1.

Source: Clinical Study Report.¹⁵

Secondary Outcome Analyses

Health-Related Quality of Life

The patient-reported outcomes were considered secondary and exploratory objectives in this study, and no formal hypotheses were formulated. Nominal P values without multiplicity adjustment were provided and should be interpreted with caution. The treatment effect on patient-reported outcome score change from baseline was evaluated at the primary analysis time point. Between-group comparisons were performed and the differences in the least squares mean change from baseline at the primary analysis time point were reported, together with 95% CIs and nominal P values. In addition, model-based least squares mean scores with corresponding 95% CIs were provided by treatment group at the primary analysis time point.

Patients with confirmed progressive disease or those feeling worse due to a drug-related AE may have missing patient-reported outcome assessments. The missing data were handled using the constrained longitudinal data analysis model to obtain valid statistical analysis results. This method implicitly treats missing data as missing at random.

Descriptive statistics (e.g., mean and standard error) of change from baseline with no imputation for missing data of the following score/scales were also plotted: EORTC QLQ-C30 global health status and QoL, EORTC QLQ-C30 physical and emotional functioning, and EQ VAS.

Objective Response Rate

The stratified Miettinen and Nurminen method was used for the comparison of ORR between 2 treatment arms. The difference in ORR and its 95% CI from the stratified Miettinen and Nurminen method with strata weighting by sample size was reported. The stratification factors used for randomization were applied to the analysis.

The ORR hypotheses were tested according to the hypotheses testing plan. Sensitivity analyses were performed for ORR based on site investigator or local radiology review.

The ORR was included in the multiplicity testing strategy and is described in Table 11.

Duration of Response

The DOR was summarized descriptively using the nonparametric Kaplan-Meier method. Only the subset of patients who achieved complete or partial response was included in this analysis.

Table 12: Statistical Analysis of Efficacy End Points

CIV = central imaging vendor; DOR = duration of response; ORR = objective response rate; OS = overall survival; PD-1 = programmed cell death 1 protein; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1.

Note: For stratified analyses, the stratification factors used for randomization were used as stratification factors for analysis.

Source: Clinical Study Report.¹⁵

Harms

Safety parameters were assessed via point estimates with 95% CIs provided for between-group comparisons using the Miettinen and Nurminen method. In addition, point estimates were provided by treatment group. Some adverse experiences (specific terms as well as system organ class terms) required that at least 4 patients in any treatment group exhibit the event. Given that many 95% CIs may be provided without adjustment for multiplicity, the CIs were regarded as a helpful descriptive measure to be used in review, not a formal method for assessing the statistical significance of the between-group differences in adverse experiences and predefined limits of change.

Planned Analyses

Three efficacy interim analyses were planned in addition to the final analysis for part 2 of the KEYNOTE-355 trial. Patients from part 1 (safety run-in) were not included in part 2 (efficacy evaluation).

Interim Analysis 1 (Final ORR Analysis)

The primary purpose of IA1 was to perform the final ORR, interim PFS, and interim OS analyses. The ORR analysis at IA1 is considered the final ORR analysis of the study. IA1 was to be performed after enrolment is completed, and approximately 9 months after the first 640 patients were randomized to part 2.

At IA1, ORR analyses were based on data from the first approximately 640 patients randomized to part 2 and were to be tested in all patients and patients with PD-L1–positive tumours. All patients randomized on or before the date the 640th patient in part 2 were randomized were to be included in the ORR analysis.

The success boundary to demonstrate ORR superiority at IA1 corresponds to an observed ORR difference of approximately 12.6% at 1-sided an alpha of 0.1% for all patients, and an observed ORR difference of approximately 14.1% at a nominal 1-sided alpha of approximately 0.145% for patients with PD-L1–positive tumours, if there were 640 patients in all patients and 480 patients in patients with PD-L1–positive tumours available for analysis (assuming a PD-L1 positivity prevalence of 75%).

Interim Analysis 2 (Final PFS Analysis)

The primary purpose of efficacy IA2 (data cut-off date: December 11, 2019) was to evaluate the superiority of pembrolizumab plus chemotherapy versus placebo plus chemotherapy in OS, and to perform the final PFS analysis. The analysis was performed approximately 12 months after the last patient was randomized.

Interim Analysis 3

The primary purpose of efficacy interim analysis 3 was to evaluate the superiority of pembrolizumab plus chemotherapy versus placebo plus chemotherapy in OS. The analysis was to be performed approximately 17 months after the last patient randomized.

Final Analysis (Final Overall Survival Analysis)

The final analysis of the study was driven by events and follow-up times, and was to be conducted after approximately 500 OS events have been observed among all patients, or after approximately 23 months since the last patient was randomized, whichever occurred later. The final analysis may be delayed for up to 4 months if fewer than 360 OS events have been observed among patients with PD-L1–positive tumours. The success boundary to demonstrate OS superiority at final analysis corresponds to an observed HR of approximately 0.80 (approximately 4.5 months improvement over control median OS of 17.5 months) at a 1-sided alpha of approximately 0.84% for all patients, and an observed HR of approximately 0.77 (approximately 5.4 months of improvement over a control median OS of 17.5 months) at a 1-sided alpha of approximately 0.84% for patients with PD-L1–positive tumours, if the planned numbers of OS events were analyzed, respectively.

A summary of the timing, sample size, and decision guidance of the 3 efficacy interim analyses and final analysis, assuming no alpha reallocation between hypotheses, is presented in Table 13. ORR boundaries assumed 640 randomized patients and a 75% PD-L1 prevalence and may be updated at the time of the analyses using the actual observed numbers. PFS and OS boundaries were based on the estimated number of events and may be updated at the time of the analyses according to the actual observed number of events, spending functions, and the spending time approach as noted above.

Table 13: Summary of Timing, Sample Size, and Decision Guidance of Efficacy Interim Analyses and Final Analysis

HR = hazard ratio; IA1 = interim analysis 1; IA2 = interim analysis 2; IA3 = interim analysis 3; NA = not applicable; ORR = overall response rate; OS = overall survival; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival.

Source: Clinical Study Report.¹⁵

Analysis Populations

Efficacy Analysis Populations

The ITT population was analyzed for primary efficacy analyses. All randomized patients were included in this population in the treatment group to which they were randomized. The ITT population was further analyzed by subgroups of a CPS of 1 or higher and a CPS of 10 or higher.

Safety Analysis Populations

The APaT population was used for the analysis of safety data in the KEYNOTE-355 trial. The APaT population consisted of all randomized patients who received at least 1 study treatment. Patients were included in the treatment group corresponding to the study treatment they actually received for the analysis of safety data using the APaT population. At least 1 laboratory or vital sign measurement obtained subsequent to at least 1 dose of the study treatment is required as well as a baseline measurement for inclusion in the analysis of each specific parameter.

Patient-Reported Outcomes Analysis Populations

Patient-reported outcome analyses (for HRQoL) were based on the patient-reported outcome final analysis population, defined as all randomized patients who received at least 1 dose of study intervention and had completed at least 1 patient-reported outcome assessment analyses and who received at least 1 study treatment.

Protocol Amendments

The original protocol was approved on April 21, 2016. The final protocol was approved October 4, 2019. Five protocol amendments were made before the final analysis data-cut-off. No changes were made after the unblinding of data.

Amendment 1 (December 6, 2016) harmonized the pneumonitis exclusion criterion throughout the pembrolizumab program.

Amendment 2 (February 5, 2018) aligned the protocol with the most current label and safety information for pembrolizumab and added guidelines for dose modification in the event of myocarditis and updated guidelines for several other conditions.

Amendment 3 (August 31, 2018) adjusted the allocation of alpha over the primary end points and key secondary end points to allocate the initial alpha to ORR hypotheses and allow testing of ORR hypothesis at IA1 independent of the outcome of the other hypotheses.

Amendment 4 (March 20, 2019) changed the timing of the final analysis from OS-driven to both OS-event– and follow-up-time–driven to ensure adequate follow-up duration at time of the final analysis.

Amendment 5 (October 4, 2019) changed the objectives, hypotheses, and statistical analysis plan to include patients with PD-L1–positive tumours with a higher CPS cut-off (≥ 10).

Results

Patient Disposition

By the final analysis data cut-off (June 15, 2021), 1,372 patients had been screened, 525 patients had failed screening, and 847 patients had been successfully randomized in a 2:1 ratio to receive either pembrolizumab plus chemotherapy (N = 566) or placebo plus chemotherapy (N = 281). All but 4 randomized patients received at least 1 dose of the study intervention. A summary of the patient disposition at the final analysis data cut-off is presented in Table 14.

As of the final analysis data cut-off, a total of 707 patients had discontinued the study. The major reason for discontinuation in the subset of patients with a PD-L1 CPS of 10 or higher was death (68.6% versus 79.6% in the pembrolizumab and placebo group, respectively). Similarly, a total of 778 (92.3%) patients had discontinued treatment. The discontinuation rate was similar between the 2 treatment groups in all patients and in patients with a PD-L1 CPS of 10 or higher. The most common reason for discontinuation of treatment in the subset of patients with a PD-L1 CPS of 10 or higher was progressive disease (58% pembrolizumab group versus 68.9% placebo group) followed by clinical progression (58% pembrolizumab group versus 68.9% placebo group).

Table 14: Patient Disposition at Final Analysis — Intention-to-Treat Population

APaT = all patients as treated; CPS = combined positive score; PD-L1 = programmed cell death 1 ligand 1.

Notes: Clinical progression and progressive disease were based on investigator assessments and may be different from the data used in the primary analysis. Progressive disease refers to disease progression based on Response Evaluation Criteria in Solid Tumours Version 1.1 and does not include clinical progression. Study medication discontinuation refers to discontinuation of all study medications. Patients without a completed study medication discontinuation form or without a completed study disposition form were labelled “status not recorded.” Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Premature Unblinding

In total, 150 premature unblinding events occurred due to the following reasons:

• sponsor-approved nonemergency unblinding request for discontinued patients to determine subsequent therapy in patients failing study intervention; N = 119

• inadvertent unblinding; N = 15

• emergency unblinding through the call centre; N = 16.

Protocol Deviations

Important protocol deviations are highlighted in Table 15. Overall, the number of clinically important protocol deviations was low in both groups (3.5% in the pembrolizumab group and 2.5% in the placebo group).

Table 15: Summary of Protocol Deviations Considered to Be Clinically Important by Trial Investigators at Final Analysis — ITT Population

ASCO = American Society of Clinical Oncology; CAP = College of American Pathologists; ITT = intention-to-treat; TNBC = triple-negative breast cancer.

Notes: Every patient is counted a single time for each applicable row and column. Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Exposure to Study Treatments

A summary of drug exposure is presented in Table 16 and a summary of the duration of exposure is presented in Table 17. The median duration of exposure to study intervention for all drugs in all patients was generally similar between the pembrolizumab plus chemotherapy group (26.4 weeks; range = 0.1 to 212.1 weeks) and the placebo plus chemotherapy group (23.1 weeks; range = 0.1 to 224.1weeks). The median exposure time of all drugs in patients with a PD-L1 CPS of 10 or higher was ||||||||||||||||||||||||||| in patients in the pembrolizumab plus chemotherapy group versus the placebo plus chemotherapy group, respectively.

In the subpopulation with a PD-L1 CPS of 10 or higher, the median number of weeks on therapy was ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| for placebo in pembrolizumab versus placebo groups, respectively. The companion chemotherapies such as nab-paclitaxel ||||||||||||||||||||||||||||||||||||) and paclitaxel (|||||||||||||||||||||||||||||||||||| were all consistently higher in the median number of weeks on treatment in the pembrolizumab group than in the placebo group. Gemcitabine and carboplatin were roughly balanced with or a little higher than the placebo group and these patterns of differences in treatment exposure between the 2 groups were consistently observed in all patients.

As of the final analysis data cut-off, median duration of follow-up in the ITT population was 16.7 months (range, 0.2 to 53.1 months) and was similar for both treatment groups. Median duration of follow-up was |||||||||||||||||||||||||||||||||||||||||||||) in patients with a PD-L1 CPS of 10 or higher.

Table 16: Summary of Drug Exposure in Patients With PD-L1 CPS ≥ 10 and All Patients at Final Analysis — APaT Population

APaT = all patients as treated; CPS = combined positive score; NA = not applicable; PD-L1 = programmed cell death 1 ligand 1.

Notes: Patients who did not have neoadjuvant treatments but had surgery were included in the APaT population in neoadjuvant phase. Patients who had postsurgery radiation therapy but did not have adjuvant treatment were included in the APaT population in the adjuvant phase. Data cut-off date: June 15, 2021.

Sources: Clinical Study Report¹⁵ and sponsor’s additional information package.³¹

Table 17: Exposure by Duration in All Patients — APaT Population

APaT = all patients as treated.

Notes: Each patient is counted once on each applicable duration category row. Duration of exposure is the time from the first dose date to the last dose date.

Source: Clinical Study Report.¹⁵

A summary of concomitant medication use (incidence > 30% in 1 or more treatment groups) is presented in Table 18.

The reported concomitant medications were generally balanced between the 2 treatment groups in part 2, with the exception that the following were higher (≥ 5 percentage point difference) in the pembrolizumab plus chemotherapy group:

• antidiarrheals, intestinal anti-inflammatory or anti-infective medications (17.1% versus 9.3%)

• medications for constipation (30.6% versus 24.6%)

• systemic antibacterial medications (56.2% versus 47.3%)

• topical corticosteroids (18.3% versus 12.5%)

• thyroid therapy (25.1% versus 13.5%).

Table 18: Patients With Specific Concomitant Medication at Final Analysis (Incidence ≥ 30% in 1 or More Treatment Groups) in Patients With PD-L1 CPS ≥ 10 and All Patients — APaT Population

APaT = all patients as treated; CPS = combined positive score; PD-L1 = programmed cell death 1 ligand 1.

Notes: Every patient is counted a single time for each applicable specific concomitant medication. A patient with multiple concomitant medications within a medication category is counted a single time for that category. Database cut-off date: June 15, 2021

Sources: Clinical Study Report¹⁵ and sponsor’s additional information package.³¹

A summary of subsequent therapy use is presented in Table 19. A lower percentage of patients in the pembrolizumab plus chemotherapy group received subsequent anticancer treatment compared with the placebo plus chemotherapy group in all patients (60.5% versus 69.0%, respectively) and in patients with PD-L1–positive tumours (||||||||||||||||||||||||||||||||||||). The most common type of subsequent therapy administered in both treatment groups was antineoplastic drugs. More patients in the placebo plus chemotherapy group received subsequent immunotherapy than did patients in the pembrolizumab plus chemotherapy group in the subset of patients with PD-L1–positive tumours (CPS ≥ 10) (||||||||||||||||||||||||||||||||||||||||||||||||||||||).

Table 19: Patients With Subsequent Antineoplastic Therapy or Immunotherapy Use (Incidence > 10% in 1 or More Treatment Groups) in Patients With PD-L1 CPS ≥ 10 and All Patients — APaT Population

APaT = all patients as treated; CPS = combined positive score; PD-L1 = programmed cell death 1 ligand 1.

Notes: Every patient is counted a single time for each applicable specific medication. A patient with multiple medications within a medication category is counted a single time for that category. A medication class or specific medication appears on this report only if its incidence in 1 or more of the columns meets the incidence criterion in the report title, after rounding. Data cut-off date: June 15, 2021.

Sources: Clinical Study Report¹⁵ and sponsor’s additional information package.³¹

Efficacy

Only those efficacy outcomes and analyses of subgroups identified in the review protocol are reported. Additional efficacy results for patients with a PD-L1 CPS > 1 are provided in Appendix 2.

Overall Survival

A summary of OS results at the final analysis data cut-off are presented in Table 20 and Figure 4. By the final analysis data cut-off, in patients with a PD-L1 CPS of 10 or higher, 155 of 220 predefined events had occurred in the pembrolizumab group, representing approximately 70.5% of OS events compared to 81.6% in the placebo group. The HR between groups was 0.73 (95% CI, 0.55 to 0.95) and the observed 1-sided P value of P = 0.0093 successfully crossed the prespecified efficacy boundary at the final analysis (the multiplicity-adjusted, 1-sided, nominal alpha level was 0.0113). The primary end point of OS was therefore met in patients with PD-L1–positive tumours (CPS ≥ 10). However, there was no significant difference in OS events in the all patients group (HR = 0.89, 95% CI, 0.76 to 1.05; P = 0.0797). The OS results for all patients were not formally tested given the OS results for patients with a PD-L1 CPS of 1 or higher did not reach statistical significance as defined by the prespecified analysis plan. The median OS was longer for patients with PD-L1–positive tumours (CPS ≥ 10) in the pembrolizumab plus chemotherapy group (23 months, 95% CI, 19 to 26.3) compared with the placebo plus chemotherapy group (16.1 months, 95% CI, 12.6 to 18.8).

Table 20: Overall Survival at Final Analysis in Patients With PD-L1 CPS ≥ 10 and All Patients — ITT Population

CI = confidence interval; ITT = intention-to-treat; NR = not reached; OS = overall survival.

Note: Data cut-off date: June 15, 2021

^aFrom product-limit (Kaplan-Meier) method for censored data.

^bBased on Cox regression model with the Efron method of tie handling with treatment as a covariate stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^cThe multiplicity-adjusted, 1-sided, prespecified P value boundary for statistical significance was P = 0.0113. One-sided P value based on log-rank test stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^dNominal P value (i.e., not multiplicity-adjusted).

Source: Clinical Study Report.¹⁵

Subgroup Analyses

A summary of OS results by subgroup in patients with a PD-L1 CPS of 10 or higher is presented in Table 21. The HR of OS across the different subgroups analyzed was generally consistent with the primary finding in the main analysis. Overall results within subgroups were similar, except for chemotherapy on study. Patients who received taxanes as chemotherapy (e.g., nab-paclitaxel and paclitaxel) had lower OS HRs (0.63 and 0.34, respectively) compared with those given gemcitabine and carboplatin (0.88). However, given the exploratory nature, small sample sizes, and lack of statistical testing, subgroup results should be interpreted with caution.

Figure 4: Kaplan-Meier Estimates of OS at Final Analysis in Patients With PD-L1 CPS ≥ 10 — ITT Population

CPS = combined positive score; ITT = intention-to-treat; OS = overall survival; PD-L1 = programmed cell death 1 ligand 1.

Table 21: Subgroup Analysis of OS in Patients With PD-L1 CPS ≥ 10 — ITT Population

CI = confidence interval; CPS = combined positive score; ECOG PS = European Cooperative Oncology Group Performance Status; ITT = intention-to-treat; IVRS = integrated voice recognition system; OS = overall survival; PD-L1 = programmed cell death 1 ligand 1.

Note: Data cut-off date: June 15, 2021.

^aAnalysis (hazard ratio and 95% CI) in the overall population is based on Cox regression model with the Efron method of tie handling with treatment as a covariate stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin), tumour PD-L1 status (CPS ≥ 1 vs. CPS < 1) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no); analysis in the subgroups is based on the unstratified Cox model. If any level of a subgroup variable has fewer than 30 patients, subgroup analysis is not performed in that level of the subgroup variable.

Source: Clinical Study Report.¹⁵

Health-Related Quality of Life

Findings presented are for the final analysis data cut-off using the final analysis population.

EORTC QLQ-C30

Compliance rates were similar at baseline for the EORTC QLQ-C30 in both the pembrolizumab plus chemotherapy and placebo plus chemotherapy groups (≥ 93%). After 15 weeks, compliance rates were 87.7% and 81.1% in the pembrolizumab and placebo groups, respectively. HRQoL estimates measured up to week 15 may not provide an accurate picture of patients’ experiences with the pembrolizumab plus chemotherapy combination for a prolonged period of time.

Change from baseline for the EORTC QLQ-C30 global health status and QoL emotional and physical functioning scales, and symptom or item scales at week 15 in patients with a PD-L1 CPS of 10 or higher are presented in Figure 5 and Figure 6, respectively. Over 15 weeks of follow-up, there was either no decrease or a similar decrease in the prespecified global health status and QoL physical and emotional functioning scores observed for both treatment groups. Among patients with PD-L1–positive tumours (CPS ≥ 10), the least squares mean change from baseline to week 15 in scores for the EORTC QLQ-C30 global health status and QoL and symptom scales were similar between the 2 treatment groups. Similar results were observed among all patients.

Analysis of time to deterioration for the EORTC QLQ-C30 across the emotional functioning, physical functioning, and global health status scales are presented in Table 23. Overall, there was no difference in time to deterioration in the EORTC QLQ-C30 global health status and QoL emotional and physical functioning scores between the pembrolizumab plus chemotherapy and placebo plus chemotherapy groups.

Table 22: Analysis of Change From Baseline in EORTC QLQ-C30 Global Health Status and QoL at Week 15 in Patients With PD-L1 CPS ≥ 10 — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; LS = least squares; PD-L1 = programmed cell death 1 ligand 1; QoL = quality of life; SD = standard deviation.

Notes: For baseline and week 15, N is the number of patients in each treatment group with nonmissing assessments at the specific time point; for change from baseline, N is the number of patients in the analysis population in each treatment group. Data cut-off date: June 15, 2021.

^aBased on a constrained longitudinal data analysis model with the patient-reported outcome scores as the response variable, and treatment by time point interaction, and stratum (defined by stratification factors of chemotherapy on study [taxane vs. gemcitabine and carboplatin] and prior treatment with same class of chemotherapy in the [neo]adjuvant setting [yes vs. no]) as covariates.

^bNominal P value (i.e., the type I error rate has not been controlled); 2-sided P value.

Source: Clinical Study Report.¹⁵

Figure 5: Change From Baseline for EORTC QLQ-C30 Global Health Status and QoL and Functional Scales at Week 15 (LS Mean Change and 95% CI) in Patients With PD-L1 CPS ≥ 10 — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; QoL = quality of life; PD-L1 = programmed cell death 1 ligand 1.

Notes: For global health status and QoL scores and all functional scales, a higher score denotes better health-related QoL or function. For symptoms scales, a higher score denotes worse symptoms. Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Figure 6: Change From Baseline for EORTC QLQ-C30 Symptom Scales and Items at Week 15 (LS Mean Change and 95% CI) in Patients With PD-L1 CPS ≥ 10 — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; LS = least squares; PD-L1 = programmed cell death 1 ligand 1; QoL = quality of life.

Notes: For global health status and QoL and all functional scales, a higher score denotes better health-related QoL or function. For symptoms scales, a higher score denotes worse symptoms. Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Table 23: Analysis of Time to Deterioration for EORTC QLQ-C30 in Patients With PD-L1 CPS ≥ 10 Across Different Scales — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; NR = not reached; PD-L1 = programmed cell death 1 ligand 1; QoL = quality of life; SD = standard deviation.

Notes: A high score in a functional scale represents a high or healthy level of functioning; a high score in the global health status and QoL represents a high QoL, and a high score for a symptom scale or item represents a high level of symptomatology or problems. Time to deterioration (event) is defined as time to first onset of 10 points or more worsening from baseline. Data cut-off date: June 15, 2021.

^aFrom product-limit (Kaplan-Meier) method for censored data.

^bBased on Cox regression model with the Efron method of tie handling with treatment as a covariate stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^cP value has not been adjusted for multiple testing (i.e., the type I error rate has not been controlled). One-sided P value based on log-rank test stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

Source: Clinical Study Report.¹⁵

EORTC QLQ-BR23

Compliance rates at baseline were similar for the EORTC QLQ-BR23 in both the pembrolizumab and placebo arms (93.5% versus 95.0%, respectively). The change from baseline in EORTC QLQ-BR23 functional and symptom scales and items at week 15 are presented in Figure 7 and Figure 8.

After 15 weeks of follow-up, the compliance rate was 87% versus 81.4% in the pembrolizumab and placebo arms, respectively. In patients with PD-L1–positive tumours (CPS ≥ 10), the least squares mean change from baseline to week 15 in scores for both the EORTC QLQ-BR23 functional scales and items (body image, sexual functioning, sexual enjoyment, and future perspective) and the EORTC QLQ-BR23 symptom scales (systemic therapy side effects, breast symptoms, group symptoms, and upset by loss of hair) were similar between the 2 treatment groups. The analyses of domains of the EORTC QLQ-BR23 demonstrate that the addition of pembrolizumab to chemotherapy did not result in a significant decrease in HRQoL. Similar results were observed among all patients.

Figure 7: Change From Baseline for EORTC QLQ-BR23 Functional Scales and Items at Week 15 (LS Mean Change and 95% CI) in Patients With PD-L1 CPS ≥ 10 — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EORTC QLQ-BR23 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire – Breast Module 23; LS = least squares; PD-L1 = programmed cell death 1 ligand 1.

Notes: For functional scales and items, a higher score denotes better function. For symptom scales and items, a higher score denotes worse symptom. N is the number of patients in the analysis population in each treatment group. Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Figure 8: Change From Baseline for EORTC QLQ-BR23 Symptom Scales and Items at Week 15 (LS Mean Change and 95% CI) in Patients With PD-L1 CPS ≥ 10 — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EORTC QLQ-BR23 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire – Breast Module 23; LS = least squares; PD-L1 = programmed cell death 1 ligand 1; QoL = quality of life.

Notes: For global health status and QoL score and all functional scales, a higher score denotes better health-related QoL or function. For symptoms scales, a higher score denotes worse symptoms. If a patient did not lose any hair, then upset by hair loss was imputed as lowest score (0). Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

EQ Visual Analogue Scale

At the final analysis data cut-off, compliance rates were similar at baseline for EQ VAS in both the pembrolizumab and placebo plus arms (92.8% versus 96.3% respectively). After 15 weeks of follow-up, the compliance rate was 87% versus 81.4% in the pembrolizumab and placebo arms, respectively. Over 15 weeks of follow-up, patients receiving pembrolizumab plus chemotherapy and placebo plus chemotherapy had similar small decreases (worsening) in prespecified EQ VAS scores. Similar results were observed among all patients. A summary of change from baseline in the EQ VAS at week 15 is presented in Table 24 and Figure 9.

Table 24: Summary of Change From Baseline in EQ VAS at Week 15 in Patients With PD-L1 CPS ≥ 10 — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EQ VAS = EQ visual analogue scale; LS = least squares; PD-L1 = programmed cell death 1 ligand 1; SD = standard deviation.

Notes: For baseline and week 15, N is the number of patients in each treatment group with nonmissing assessments at the specific time point; for change from baseline, N is the number of patients in the analysis population in each treatment group. Data cut-off date: June 15, 2021.

^aBased on constrained longitudinal data analysis model with the patient-reported outcomes scores as the response variable, and treatment by time point interaction, and stratum (defined by stratification factors of chemotherapy on study [taxane vs. gemcitabine and carboplatin] and prior treatment with same class of chemotherapy in the (neo)adjuvant setting [yes vs. no]) as covariates.

^bP value has not been adjusted for multiple testing (i.e., the type I error rate has not been controlled); 2-sided P value.

Source: Clinical Study Report.¹⁵

Progression-Free Survival

A summary of PFS results at the final analysis data cut-off are presented in Table 25. At IA2 (December 11, 2019), the KEYNOTE-355 trial met the success criterion for the primary hypothesis of PFS in patients with PD-L1–positive tumours (CPS ≥ 10). The analyses performed at IA2 were the final prespecified analyses for PFS and updated data were provided at the final analysis, with nominal P values. At IA2, the median PFS was 9.7 months in the pembrolizumab plus chemotherapy and 5.6 months in the placebo plus chemotherapy group. The HR for progression or death, was 0.65, (95% CI, 0.49 to 0.86; 1-sided P = 0.0012). Based on the prespecified success criteria (alpha = 0.00411), pembrolizumab plus chemotherapy significantly improved PFS compared with placebo plus chemotherapy in patients with CPS ≥ 10.

The findings from the final analysis were consistent with the results at IA2. At the final analysis data cut-off, a total of 144 (65.5%) events had occurred in the pembrolizumab plus chemotherapy group and 81 events (78.6%) in the placebo plus chemotherapy group in patients with PD-L1–positive tumours (CPS ≥ 10). The PFS stratified HR at final analysis was 0.66 (95% CI, 0.50 to 0.88; P = 0.0018) in patients with a PD-L1 CPS of 10 or higher. Median PFS was 9.7 months (95% CI, 7.6 to 11.3) in the pembrolizumab plus chemotherapy group compared to 5.6 months (95% CI, 5.3 to 7.5) in the placebo plus chemotherapy group in patients with PD-L1–positive tumours (CPS ≥ 10).

Pembrolizumab plus chemotherapy continued to show a statistically significant improvement in PFS compared with placebo plus chemotherapy in patients with PD-L1–positive tumours (CPS ≥ 10).

Figure 9: Empirical Mean Change From Baseline in EQ VAS Across Time in Patients With PD-L1 CPS ≥ 10 — Final Analysis Population

CI = confidence interval; CPS = combined positive score; EQ VAS = EQ visual analogue scale; PD-L1 = programmed cell death 1 ligand 1.

Note: Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Table 25: PFS Based on BICR Assessment According to RECIST 1.1 in Patients With PD-L1 CPS ≥ 10 and All Patients — ITT Population

BICR = blinded independent central review; CI = confidence interval; CPS = combined positive score; ITT = intention-to-treat; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1.

Notes: According to the statistical analysis plan, multiplicity-adjusted analysis of PFS was conducted at interim analysis 2. Results from the final analysis were consistent with results from interim analysis 2. Data cut-off date: June 15, 2021.

^aFrom product-limit (Kaplan-Meier) method for censored data.

^bBased on Cox regression model with the Efron method of tie handling with treatment as a covariate stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^cNominal P value (i.e., the type I error rate has not been controlled); 1-sided P value based on log-rank test stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

Source: Clinical Study Report.¹⁵

Subgroup Analyses for PFS

A summary of PFS results by subgroup in the population with a PD-L1 CPS of 10 or higher is presented in Table 26. The treatment benefit in PFS for pembrolizumab plus chemotherapy compared with placebo plus chemotherapy in patients with PD-L1–positive tumours (CPS ≥ 10) was also generally consistent across the different subgroups analyzed and when applying 2 prespecified sensitivity censoring rules. Patients who received taxanes as their chemotherapy (e.g., nab-paclitaxel and paclitaxel) had lower PFS HRs (0.63 and 0.34, respectively) than did those receiving gemcitabine and carboplatin (0.88). However, given the exploratory nature, small sample sizes, and lack of statistical testing, subgroup results should be interpreted with caution.

Objective Response Rate

A summary of ORR results at the final analysis data cut-off is presented in Table 27. At the final analysis data cut-off, there were 116 objective responses (52.7%; 95% CI, 45.9 to 59.5) in the pembrolizumab group compared to 42 objective responses (40.8%; 95% CI, 31.2 to 50.9) in the placebo group in patients with a PD-L1 CPS of 10 or higher. The secondary hypotheses pertaining to ORR were tested at IA1 and not formally tested at the final analysis. The ORR was formally tested in all patients and patients with a PD-L1 CPS of 1 or higher and not in patients with a CPS of 10 or higher. The success criterion for ORR hypotheses were not met at IA1. These ORR results from the final analysis were consistent with IA1 and did not show a significant difference between the 2 treatment groups (nominal P value = 0.0213). Overall, pembrolizumab plus chemotherapy showed greater improvements in ORR (as defined by RECIST 1.1 and determined by blinded independent central review) compared with placebo plus chemotherapy in patients with PD-L1–positive tumours (CPS ≥ 10).

Subgroup Analyses of ORR

A summary of ORR results by subgroup in the PD-L1 CPS ≥ 10 population is presented in Table 28. The treatment benefit in ORR for pembrolizumab plus chemotherapy compared with placebo plus chemotherapy in patients with PD-L1–positive tumours (CPS ≥ 10) was also generally consistent across the different subgroups analyzed. Subgroup results were similar in the all patients population. Given the exploratory nature, small sample sizes, and lack of statistical testing, subgroup results should be interpreted with caution.

Figure 10: Kaplan-Meier Estimates of PFS Based on BICR Assessment According to RECIST 1.1 in Patients With PD-L1 CPS ≥ 10 — ITT Population

BICR = blinded independent central review; CPS = combined positive score; ITT = intention-to-treat; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1.

Note: Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Table 26: PFS by Subgroups at Final Analysis in Patients With PD-L1 CPS ≥ 10 — ITT Population

CPS = combined positive score; ECOG PS = European Cooperative Oncology Group Performance Status; IVRS = interactive voice recognition system; PD-L1 = programmed cell death 1 ligand 1.

Note: Data cut-off date: June 15, 2021.

^aAnalysis (hazard ratio and 95% CI) in the overall population is based on Cox regression model with the Efron method of tie handling with treatment as a covariate stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin), tumour PD-L1 status (CPS ≥ 1 vs. CPS < 1) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no); analysis in the subgroups is based on the unstratified Cox model. If any level of a subgroup variable has fewer than 30 patients, subgroup analysis is not performed in that level of the subgroup variable.

Source: Clinical Study Report.¹⁵

Objective Response Rate

A summary of ORR results at the final analysis data cut-off is presented in Table 27. At the final analysis data cut-off, there were 116 objective responses (52.7%; 95% CI, 45.9 to 59.5) in the pembrolizumab group compared to 42 objective responses (40.8%; 95% CI, 31.2 to 50.9) in the placebo group in patients with a PD-L1 CPS of 10 or higher. The secondary hypotheses pertaining to ORR were tested at IA1 and not formally tested at the final analysis. The ORR was formally tested in all patients and patients with a PD-L1 CPS of 1 or higher and not in patients with a CPS of 10 or higher. The success criterion for ORR hypotheses were not met at IA1. These ORR results from the final analysis were consistent with IA1 and did not show a significant difference between the 2 treatment groups (nominal P value = 0.0213). Overall, pembrolizumab plus chemotherapy showed greater improvements in ORR (as defined by RECIST 1.1 and determined by blinded independent central review) compared with placebo plus chemotherapy in patients with PD-L1–positive tumours (CPS ≥ 10).

Subgroup Analyses of ORR

A summary of ORR results by subgroup in the PD-L1 CPS ≥ 10 population is presented in Table 28. The treatment benefit in ORR for pembrolizumab plus chemotherapy compared with placebo plus chemotherapy in patients with PD-L1–positive tumours (CPS ≥ 10) was also generally consistent across the different subgroups analyzed. Subgroup results were similar in the all patients population. Given the exploratory nature, small sample sizes, and lack of statistical testing, subgroup results should be interpreted with caution.

Duration of Response

A summary of DOR results at the final analysis data cut-off is presented in Table 29. The median DOR was greater in the pembrolizumab plus chemotherapy group at 12.8 months compared to the placebo plus chemotherapy group at 7.3 months in patients with PD-L1–positive tumours (CPS ≥ 10). The median time to response was 1.9 months in both groups. DOR results from the final analysis were consistent with the previously reported IA2 results and show a sustained DOR benefit in the pembrolizumab plus chemotherapy group over a longer follow-up (Figure 11). The proportions of patients with an extended response at 6 months or later by Kaplan-Meier estimation were 87.1% in the pembrolizumab plus chemotherapy group and 55.5% in the placebo plus chemotherapy group. At 12 months or later the ongoing response at the data cut-off was again higher for the pembrolizumab plus chemotherapy group (55.5%) compared with the placebo plus chemotherapy group (37.9%) by Kaplan-Meier estimation.

Table 27: Analysis of ORR Based on BICR Assessment According to RECIST 1.1 in Patients With PD-L1 CPS ≥ 10 and All Patients — ITT Population

BICR = blinded independent central review; CI = confidence interval; CPS = combined positive score; ORR = overall response rate; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1.

Note: Data cut-off date: June 15, 2021.

^aBased on Miettinen and Nurminen method stratified by chemotherapy on study (taxane vs. gemcitabine and carboplatin) and prior treatment with same class of chemotherapy in the (neo)adjuvant setting (yes vs. no).

^bNominal P value. One-sided P value for testing. Null hypothesis: difference in percentage of 0 vs. hypothesis 1: difference in percentage greater than 0. Confirmed responses are included.

Source: Clinical Study Report.¹⁵

Figure 11: Kaplan-Meier Estimates of DOR in Patients With Confirmed Response based on BICR Assessment According to RECIST 1.1 in Patients With PD-L1 CPS ≥ 10 — ITT Population

BICR = blinded independent central review; CPS = combined positive score; DOR = duration of response; ITT = intention-to-treat; PD-L1 = programmed cell death 1 ligand 1; PFS = progression-free survival; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1.

Note: Data cut-off date: June 15, 2021.

Source: Clinical Study Report.¹⁵

Harms

Only those harms identified in the CADTH review protocol are reported in this section and in Table 30.

Adverse Events

Overall, 98.6% of patients enrolled in the pembrolizumab plus chemotherapy group and 98.2% in the placebo plus chemotherapy reported at least 1 AE by the final analysis data cut-off. Table 30 presents an overview of AEs in the pembrolizumab plus chemotherapy group and the placebo plus chemotherapy and placebo group. Overall, incidences of common AEs were similar between the 2 treatment groups. The most frequently reported AEs (occurring in > 30% of patients) receiving treatment in the pembrolizumab plus chemotherapy group and the placebo plus chemotherapy were anemia, nausea, neutropenia, alopecia, and fatigue. The incidence of hypothyroidism was greater in the pembrolizumab plus chemotherapy group (15.8%) compared to the placebo plus chemotherapy group (3.2%).

Serious Adverse Events

By the final analysis data cut-off, the incidence of SAEs was 30.1% of patients in the pembrolizumab plus chemotherapy group and 23.8% in the placebo plus chemotherapy group. The most common SAEs (approximately 2.0% of patients) in the pembrolizumab and placebo group were anemia (2.0%), pneumonia (2.0%), thrombocytopenia (2.0%), and vomiting (1.8%).

Table 30: Summary of Adverse Events — All Patients as Treated Population