How low can you go? PD-L1 expression as a biomarker in trials of cancer immunotherapy

Accumulating evidence suggests that the effectiveness of immune checkpoint inhibitors depends on tumour expression of programmed death-ligand 1 (PD-L1). Here we review evidence relating to this biomarker and highlight problems in the design and reporting of immunotherapy trials that can lead to suboptimal treatment. Expression of PD-L1 is assessed using immunohistochemistry.1Patel S.P. Kurzrock R. PD-L1 Expression as a predictive biomarker in cancer immunotherapy.Mol Cancer Ther. 2015; 14: 847-856Crossref PubMed Scopus (1276) Google Scholar There are two dominant scoring methods: tumour proportion score (TPS) quantifies the proportion of tumour cells that stain for PD-L1, whereas combined positive score (CPS) provides an aggregate assessment of PD-L1 staining of tumour cells, lymphocytes and macrophages as a proportion of total tumour cells.2de Ruiter E.J. Mulder F.J. Koomen B.M. et al.Comparison of three PD-L1 immunohistochemical assays in head and neck squamous cell carcinoma (HNSCC).Mod Pathol. 2020; https://doi.org/10.1038/s41379-020-0644-7Crossref Scopus (22) Google Scholar Other assays, such as tumour infiltrating immune cells (IC), quantify such cells staining positive for PD-L1 as a proportion of tumour area.3Cimino-Mathews A. Thompson E. Taube J.M. et al.PD-L1 (B7-H1) expression and the immune tumor microenvironment in primary and metastatic breast carcinomas.Hum Pathol. 2016; 47: 52-63Crossref PubMed Scopus (216) Google Scholar, 4Schmid P. Adams S. Rugo H.S. et al.Atezolizumab and nab-paclitaxel in advanced triple-negative breast cancer.N Engl J Med. 2018; 379: 2108-2121Crossref PubMed Scopus (1967) Google Scholar, 5Huang R.S.P. Haberberger J. Severson E. et al.A pan-cancer analysis of PD-L1 immunohistochemistry and gene amplification, tumor mutation burden and microsatellite instability in 48,782 cases.Mod Pathol. 2021; 34: 252-263Crossref PubMed Scopus (29) Google Scholar, 6Eckstein M. Cimadamore A. Hartmann A. et al.PD-L1 assessment in urothelial carcinoma: a practical approach.Ann Transl Med. 2019; 7: 690Crossref PubMed Google Scholar Unfortunately, these scoring methods are not directly comparable.2de Ruiter E.J. Mulder F.J. Koomen B.M. et al.Comparison of three PD-L1 immunohistochemical assays in head and neck squamous cell carcinoma (HNSCC).Mod Pathol. 2020; https://doi.org/10.1038/s41379-020-0644-7Crossref Scopus (22) Google Scholar,6Eckstein M. Cimadamore A. Hartmann A. et al.PD-L1 assessment in urothelial carcinoma: a practical approach.Ann Transl Med. 2019; 7: 690Crossref PubMed Google Scholar, 7De Marchi P. Leal L.F. Duval da Silva V. et al.PD-L1 expression by tumor proportion score (TPS) and combined positive score (CPS) are similar in non-small cell lung cancer (NSCLC).J Clin Pathol. 2021; https://doi.org/10.1136/jclinpath-2020-206832Crossref PubMed Scopus (12) Google Scholar, 8Guo H. Ding Q. Gong Y. et al.Comparison of three scoring methods using the FDA-approved 22C3 immunohistochemistry assay to evaluate PD-L1 expression in breast cancer and their association with clinicopathologic factors.Breast Cancer Res. 2020; 22: 69Crossref PubMed Scopus (23) Google Scholar, 9Kulangara K. Guerrero L. Posch A. et al.Investigation of PD-L1 expression and response to pembrolizumab (pembro) in gastric cancer (GC) and cervical cancer (CC) using combined positive score (CPS) and tumor proportion score (TPS).J Clin Oncol. 2018; 36: 4065Crossref Google Scholar, 10Kulangara K. Zhang N. Corigliano E. et al.Clinical utility of the combined positive score for programmed death ligand-1 expression and the approval of pembrolizumab for treatment of gastric cancer.Arch Pathol Lab Med. 2019; 143: 330-337Crossref PubMed Scopus (259) Google Scholar In general, the probability of response to immunotherapy increases with higher PD-L1 expression, although there are exceptions.1Patel S.P. Kurzrock R. PD-L1 Expression as a predictive biomarker in cancer immunotherapy.Mol Cancer Ther. 2015; 14: 847-856Crossref PubMed Scopus (1276) Google Scholar For example, the phase II KEYNOTE-001 trial of pembrolizumab for patients with non-small-cell lung cancer (NSCLC) reported response rates of 45%, 17% and 11%, corresponding to TPS scores of ≥50%, 1%-49% and <1%, respectively.11Reck M. Rodriguez-Abreu D. Robinson A.G. et al.Pembrolizumab versus chemotherapy for PD-L1-positive non-small-cell lung cancer.N Engl J Med. 2016; 375: 1823-1833Crossref PubMed Scopus (5682) Google Scholar Similar trends were observed in trials for patients with hepatocellular and bladder cancer.12Zhu A.X. Finn R.S. Edeline J. et al.Pembrolizumab in patients with advanced hepatocellular carcinoma previously treated with sorafenib (KEYNOTE-224): a non-randomised, open-label phase 2 trial.Lancet Oncol. 2018; 19: 940-952Abstract Full Text Full Text PDF PubMed Scopus (1127) Google Scholar,13Balar A.V. Castellano D. O’Donnell P.H. et al.First-line pembrolizumab in cisplatin-ineligible patients with locally advanced and unresectable or metastatic urothelial cancer (KEYNOTE-052): a multicentre, single-arm, phase 2 study.Lancet Oncol. 2017; 18: 1483-1492Abstract Full Text Full Text PDF PubMed Scopus (704) Google Scholar In contrast, the CheckMate-067 trial of combined ipilimumab + nivolumab for patients with melanoma reported no trends for differences in survival in groups defined by TPS of <1%, ≥1%, 1%-5%, <5% or ≥5%.14Larkin J. Chiarion-Sileni V. Gonzalez R. et al.Five-year survival with combined nivolumab and ipilimumab in advanced melanoma.N Engl J Med. 2019; 381: 1535-1546Crossref PubMed Scopus (1308) Google Scholar Heterogeneity among tumours and scoring systems makes it difficult to interpret the appropriateness of PD-L1 cut-offs in clinical trials. While receiver operating characteristic curves can be used to define a PD-L1 cut-off,11Reck M. Rodriguez-Abreu D. Robinson A.G. et al.Pembrolizumab versus chemotherapy for PD-L1-positive non-small-cell lung cancer.N Engl J Med. 2016; 375: 1823-1833Crossref PubMed Scopus (5682) Google Scholar,14Larkin J. Chiarion-Sileni V. Gonzalez R. et al.Five-year survival with combined nivolumab and ipilimumab in advanced melanoma.N Engl J Med. 2019; 381: 1535-1546Crossref PubMed Scopus (1308) Google Scholar it is preferable to demonstrate that an immunotherapy drug does or does not improve survival compared with conventional therapy for groups defined by ranges of PD-L1 expression.15Camidge D.R. Doebele R.C. Kerr K.M. Comparing and contrasting predictive biomarkers for immunotherapy and targeted therapy of NSCLC.Nat Rev Clin Oncol. 2019; 16: 341-355Crossref PubMed Scopus (203) Google Scholar Many trials have analysed their primary endpoint for groups defined by PD-L1 expression above a pre-defined level, and the cut point has sometimes been changed during conduct of the trial.16Burtness B. Harrington K.J. Greil R. et al.Pembrolizumab alone or with chemotherapy versus cetuximab with chemotherapy for recurrent or metastatic squamous cell carcinoma of the head and neck (KEYNOTE-048): a randomised, open-label, phase 3 study.Lancet. 2019; 394: 1915-1928Abstract Full Text Full Text PDF PubMed Scopus (874) Google Scholar,17Cortes J. Cescon D.W. Rugo H.S. et al.Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial.Lancet. 2020; 396: 1817-1828Abstract Full Text Full Text PDF PubMed Scopus (355) Google Scholar Cut-point selection for PD-L1 expression is opaque and at the discretion of the sponsor. Phase III trials that have used this ‘bottom-up’ reporting are summarized in Table 1.Table 1PD-L1 reporting in phase III immunotherapy trials and the cut-offs utilized for analysis of their primary endpoint when bottom-up analysis was undertakenTrialTumour siteComparisonPD-L1 assayPD-L1 cut-off for primary endpointNotesCheckMate-06714Larkin J. Chiarion-Sileni V. Gonzalez R. et al.Five-year survival with combined nivolumab and ipilimumab in advanced melanoma.N Engl J Med. 2019; 381: 1535-1546Crossref PubMed Scopus (1308) Google ScholarMelanomaIpilimumab/nivolumab or nivolumab versus ipilimumab aloneTPSNo PD-L1 cut-offExtensive justification for not using tumour PD-L1 expression as cut-offKEYNOTE-04218Mok T.S.K. Wu Y.L. Kudaba I. et al.Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial.Lancet. 2019; 393: 1819-1830Abstract Full Text Full Text PDF PubMed Scopus (1387) Google ScholarLungPembrolizumab versus chemotherapyTPS≥1%, ≥20% and ≥50%Multiple protocol amendmentsKEYNOTE-04816Burtness B. Harrington K.J. Greil R. et al.Pembrolizumab alone or with chemotherapy versus cetuximab with chemotherapy for recurrent or metastatic squamous cell carcinoma of the head and neck (KEYNOTE-048): a randomised, open-label, phase 3 study.Lancet. 2019; 394: 1915-1928Abstract Full Text Full Text PDF PubMed Scopus (874) Google ScholarHead and neckPembrolizumab versus cetuximab/chemotherapyCPS≥1, ≥20Tested 14 primary hypothesesKEYNOTE-06119Shitara K. Özgüroğlu M. Bang Y.J. et al.Pembrolizumab versus paclitaxel for previously treated, advanced gastric or gastro-oesophageal junction cancer (KEYNOTE-061): a randomised, open-label, controlled, phase 3 trial.Lancet. 2018; 392: 123-133Abstract Full Text Full Text PDF PubMed Scopus (664) Google ScholarGastricPembrolizumab versus paclitaxelCPS≥1Response rate doubled at higher CPS cut-off but primary endpoint set at ≥1KEYNOTE-06220Shitara K. Van Cutsem E. Bang Y.J. et al.Efficacy and safety of pembrolizumab or pembrolizumab plus chemotherapy vs chemotherapy alone for patients with first-line, advanced gastric cancer: the KEYNOTE-062 phase 3 randomized clinical trial.JAMA Oncol. 2020; 6: 1571-1580Crossref PubMed Scopus (229) Google ScholarGastricPembrolizumab +/− chemotherapy versus chemotherapyCPS≥1Concludes non-inferiority for all patients but may be inferior for CPS <10 and superior for CPS ≥10JAVELIN-10021Powles T. Park S.H. Voog E. et al.Avelumab maintenance therapy for advanced or metastatic urothelial carcinoma.N Engl J Med. 2020; 383: 1218-1230Crossref PubMed Scopus (333) Google ScholarBladderAvelumab maintenance after chemotherapy versus no maintenanceHybrid scorePDL-1 positive by any of several criteria, total populationUsed a binary score derived from Massard et al.22Massard C. Gordon M.S. Sharma S. et al.Safety and efficacy of durvalumab (MEDI4736), an anti-programmed cell death ligand-1 immune checkpoint inhibitor, in patients with advanced urothelial bladder cancer.J Clin Oncol. 2016; 34: 3119-3125Crossref PubMed Scopus (622) Google Scholar; utilized elements of TC, IC and other scoring elementsKEYNOTE-04513Balar A.V. Castellano D. O’Donnell P.H. et al.First-line pembrolizumab in cisplatin-ineligible patients with locally advanced and unresectable or metastatic urothelial cancer (KEYNOTE-052): a multicentre, single-arm, phase 2 study.Lancet Oncol. 2017; 18: 1483-1492Abstract Full Text Full Text PDF PubMed Scopus (704) Google ScholarBladderPembrolizumab versus chemotherapyCPS≥10 and total populationSubgroup analysis for <1 raises questions as to efficacy in PD-L1 negative group. Further data not presented in supplementKEYNOTE-35517Cortes J. Cescon D.W. Rugo H.S. et al.Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial.Lancet. 2020; 396: 1817-1828Abstract Full Text Full Text PDF PubMed Scopus (355) Google ScholarTNBCPembrolizumab + chemotherapy versus chemotherapyCPS≥1 and ≥10Reported PFS above and below three separate cut points in exploratory analysisIMpassion-1304Schmid P. Adams S. Rugo H.S. et al.Atezolizumab and nab-paclitaxel in advanced triple-negative breast cancer.N Engl J Med. 2018; 379: 2108-2121Crossref PubMed Scopus (1967) Google ScholarTNBCAtezolizumab + nab-paclitaxel versus nab-paclitaxelIC≥1% and total populationBottom-up PFS reportingKEYNOTE-42623Rini B.I. Plimack E.R. Stus V. et al.Pembrolizumab plus axitinib versus sunitinib for advanced renal-cell carcinoma.N Engl J Med. 2019; 380: 1116-1127Crossref PubMed Scopus (1430) Google ScholarRCCPembrolizumab + axitinib versus sunitinibCPSNo PD-L1 cut-offExploratory analysisMajority of patients have PD-L1 ≥1%, a group which appears to have better outcomes with immunotherapy as per forest plotCheckMate-21424Motzer R.J. Tannir N.M. McDermott D.F. et al.Nivolumab plus ipilimumab versus sunitinib in advanced renal-cell carcinoma.N Engl J Med. 2018; 378: 1277-1290Crossref PubMed Scopus (2189) Google ScholarRCCIpilimumab/nivolumab versus sunitinibTPSNo PD-L1 cut-offKaplan Meir curves shows benefit in non-expressors. Adequate reportingCPS, combined positive score; IC, tumour infiltrating immune cells; PD-L1, programmed death-ligand 1; RCC, renal cell carcinoma; TPS, tumour proportion score; TNBC, triple-negative breast cancer. Open table in a new tab CPS, combined positive score; IC, tumour infiltrating immune cells; PD-L1, programmed death-ligand 1; RCC, renal cell carcinoma; TPS, tumour proportion score; TNBC, triple-negative breast cancer. The KEYNOTE-042 trial, comparing pembrolizumab with chemotherapy for patients with advanced NSCLC, demonstrates why bottom-up reporting is problematic.18Mok T.S.K. Wu Y.L. Kudaba I. et al.Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial.Lancet. 2019; 393: 1819-1830Abstract Full Text Full Text PDF PubMed Scopus (1387) Google Scholar,25Tannock I.F. Templeton A.J. Flawed trials for cancer.Ann Oncol. 2020; 31: 331-333Abstract Full Text Full Text PDF PubMed Scopus (4) Google Scholar Pembrolizumab was already known to improve overall survival (OS) compared with chemotherapy for NSCLC patients with PD-L1 expression by TPS ≥50%.11Reck M. Rodriguez-Abreu D. Robinson A.G. et al.Pembrolizumab versus chemotherapy for PD-L1-positive non-small-cell lung cancer.N Engl J Med. 2016; 375: 1823-1833Crossref PubMed Scopus (5682) Google Scholar The original primary endpoint in KEYNOTE-042 of OS in patients with TPS ≥50% was changed to sequential testing of OS in patients with TPS ≥50%, ≥20% and ≥1%. Of the total trial population, 47% had tumours with TPS ≥50%, and the benefit of pembrolizumab was confirmed for this group.18Mok T.S.K. Wu Y.L. Kudaba I. et al.Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial.Lancet. 2019; 393: 1819-1830Abstract Full Text Full Text PDF PubMed Scopus (1387) Google Scholar For patients with TPS ≥50%, ≥20% and ≥1%, median survival with pembrolizumab was 20, 18, and 17 months, respectively, demonstrating dilution of treatment effect. The trial showed better survival for pembrolizumab than for chemotherapy in patients with TPS ≥1%, and pembrolizumab was approved for such patients by the American Food and Drug Administration (FDA). However, exploratory analysis of patients whose tumours had PD-L1 expression between 1% and 49% showed no benefit of immunotherapy compared with chemotherapy.18Mok T.S.K. Wu Y.L. Kudaba I. et al.Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial.Lancet. 2019; 393: 1819-1830Abstract Full Text Full Text PDF PubMed Scopus (1387) Google Scholar Lowering of the PD-L1 cut-off for pembrolizumab from 50% to 1% in NSCLC is disingenuous and driven in part by a desire to increase market share. Proponents might argue that pembrolizumab gives similar survival to chemotherapy in those with PD-L1 expression of 1%-49% and might have less toxicity, but financial toxicity of pembrolizumab is substantial and the trial was designed to demonstrate superiority rather than non-inferiority. Some trials in progress have been modified to evaluate new cut-point levels of PD-L1. For example, the three-arm KEYNOTE-048 trial for head and neck cancer initially planned to test 3 primary hypotheses but ultimately tested 14: superiority of pembrolizumab alone and of pembrolizumab with chemotherapy versus cetuximab with chemotherapy for OS and progression-free survival (PFS) for the total population and for groups with PD-L1 CPS ≥20 and ≥1 and non-inferiority of pembrolizumab alone and pembrolizumab with chemotherapy for OS in the total population.16Burtness B. Harrington K.J. Greil R. et al.Pembrolizumab alone or with chemotherapy versus cetuximab with chemotherapy for recurrent or metastatic squamous cell carcinoma of the head and neck (KEYNOTE-048): a randomised, open-label, phase 3 study.Lancet. 2019; 394: 1915-1928Abstract Full Text Full Text PDF PubMed Scopus (874) Google Scholar Survival was improved significantly for pembrolizumab alone in CPS ≥20 and CPS ≥1 groups compared with controls and was non-inferior for the total population. However, 63% of responders to pembrolizumab and 71% of complete responders in the CPS ≥1 group had tumours with CPS ≥20. OS and PFS were not reported for the CPS 1-19 group, but only 14.5% of these patients responded to pembrolizumab compared with 31.3% of controls. Despite this, the FDA approved a new indication for pembrolizumab for all patients with tumours having CPS ≥1. Problems arise when endpoint analysis is carried out on the total trial population despite knowledge that PD-L1 expression influences outcome. The IMpassion-130 trial of atezolizumab + nab-paclitaxel versus nab-paclitaxel alone for women with triple-negative breast cancer (TNBC) illustrates this concept. Despite prior studies showing no responses to atezolizumab for women with TNBC and PD-L1 expression <1% by IC, the primary endpoints of OS and PFS were evaluated in the entire population and in patients with IC ≥1%.4Schmid P. Adams S. Rugo H.S. et al.Atezolizumab and nab-paclitaxel in advanced triple-negative breast cancer.N Engl J Med. 2018; 379: 2108-2121Crossref PubMed Scopus (1967) Google Scholar,26Emens L.A. Cruz C. Eder J.P. et al.Long-term clinical outcomes and biomarker analyses of atezolizumab therapy for patients with metastatic triple-negative breast cancer: a phase 1 Study.JAMA Oncol. 2019; 5: 74-82Crossref PubMed Scopus (354) Google Scholar The trial met its primary PFS endpoint in the total population, but this was driven by the PD-L1 positive group. Analysis restricted to the PD-L1 positive subgroup (41% of patients) could not be tested formally because of the statistical design, but exploratory analyses suggest a larger benefit favouring atezolizumab for those with PD-L1 ≥1% [median PFS: 25.0 versus 15.5 months; hazard ratio (HR) = 0.62]. The authors concluded that atezolizumab improved PFS in the whole population without acknowledging that efficacy may be limited to the PD-L1 positive subgroup. Another risk of bottom-up reporting is masking the true benefit in a subgroup of patients. The KEYNOTE-061 trial compared pembrolizumab to paclitaxel for advanced gastric cancer.19Shitara K. Özgüroğlu M. Bang Y.J. et al.Pembrolizumab versus paclitaxel for previously treated, advanced gastric or gastro-oesophageal junction cancer (KEYNOTE-061): a randomised, open-label, controlled, phase 3 trial.Lancet. 2018; 392: 123-133Abstract Full Text Full Text PDF PubMed Scopus (664) Google Scholar With the pre-specified cut point of CPS ≥1, there was no significant difference in the primary endpoint of OS (medians 9.1 versus 8.3 months: HR = 0.82). Post-hoc analysis revealed that for patients with CPS ≥10, there was a strong trend for improved survival with immunotherapy (medians 10.4 versus 8.2 months, HR = 0.64). The follow-up KEYNOTE-062 trial tested non-inferiority of pembrolizumab compared to chemotherapy for treatment of advanced gastric cancer in both CPS ≥1 and ≥10 groups.20Shitara K. Van Cutsem E. Bang Y.J. et al.Efficacy and safety of pembrolizumab or pembrolizumab plus chemotherapy vs chemotherapy alone for patients with first-line, advanced gastric cancer: the KEYNOTE-062 phase 3 randomized clinical trial.JAMA Oncol. 2020; 6: 1571-1580Crossref PubMed Scopus (229) Google Scholar The trial met its non-inferiority endpoint at the cut-off of ≥1 with median OS of 10.6 months for pembrolizumab and 11.1 months for chemotherapy. Within the pembrolizumab arm, 36% had CPS ≥10 and median OS in this group was 17.4 months versus 10.8 months with chemotherapy, but the statistical design did not permit this analysis. The authors concluded that pembrolizumab is non-inferior to chemotherapy for all patients with CPS ≥1. An alternative conclusion, based on consistent results of these two trials, is that pembrolizumab is inferior to chemotherapy for patients with gastric cancer with CPS <10 but beneficial in those with CPS ≥10. The analysis of some immunotherapy trials demonstrates how reporting of immunotherapy trials might be improved. The CheckMate-067 early melanoma trial of nivolumab or ipilimumab alone or nivolumab + ipilumumab employed extensive exploratory analyses to justify exclusion of a PD-L1 cut-off. Kaplan–Meier curves were derived for groups with tumours above and below the PD-L1 cut-offs of 1%, 5% and 10% and for patients with 1%-5% expression.14Larkin J. Chiarion-Sileni V. Gonzalez R. et al.Five-year survival with combined nivolumab and ipilimumab in advanced melanoma.N Engl J Med. 2019; 381: 1535-1546Crossref PubMed Scopus (1308) Google Scholar For each pair of curves, dual immunotherapy showed superior survival to ipilimumab monotherapy negating the need for a PD-L1 cut point. The KEYNOTE-355 trial, which compared pembrolizumab + chemotherapy to chemotherapy alone for women with TNBC, reported data in a similar manner.17Cortes J. Cescon D.W. Rugo H.S. et al.Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial.Lancet. 2020; 396: 1817-1828Abstract Full Text Full Text PDF PubMed Scopus (355) Google Scholar The primary endpoints were PFS and OS for groups with CPS ≥1 and CPS ≥10. The CPS ≥10 cut point was added after trial enrolment, and exploratory results for median PFS were reported for patients above and below CPS cut points of 1, 10 and 20. Unfortunately, most immunotherapy trials do not report data for variable PD-L1 cut-offs. The fundamental problem with bottom-up reporting is that clinicians cannot determine which patients are most likely to benefit from immunotherapy. We recognize that trials must plan a limited number of primary analyses to avoid statistical problems of multiplicity, but they also need flexibility if the biomarker cut-off that indicates probable benefit is unknown. The ultimate aim of any treatment is to improve survival or quality of life, and the number of comparisons should be reduced by designating OS as the sole primary outcome. A Bayesian approach can determine an initial cut-off in tumour PD-L1 expression based on prior information from other trials, but patients should be stratified across pre-defined ranges of tumour PD-L1 expression. This would allow the primary endpoint to be analysed for different cut points, with the caveat of reduced statistical power for exploratory subgroup analyses. Adaptive analysis could be built into the statistical design to allow modification of the cut-off in PD-L1 expression. Such reporting would address the clinical question as to which patients benefit from immunotherapy, rather than the commercial question as to how to identify the largest possible patient population with a statistically significant improvement in outcome. More rational use of the biomarker would maximize treatment benefit to patients while minimizing overutilization of drugs with toxicity and high financial costs. None declared.