The PD-L1 Immunohistochemistry Biomarker: Two Steps Forward, One Step Back?
2018; Elsevier BV; Volume: 13; Issue: 3 Linguagem: Inglês
10.1016/j.jtho.2018.01.020
ISSN1556-1380
Autores Tópico(s)Pancreatic and Hepatic Oncology Research
ResumoDrugs targeting programmed death 1 (PD-1) or its ligand, programmed death ligand 1 (PD-L1), have proved successful in treating advanced-stage NSCLC and revolutionized the way patients with this disease are now treated.1Brahmer J. Reckamp K.L. Baas P. et al.Nivolumab versus docetaxel in advanced squamous-cell non-small-cell lung cancer.N Engl J Med. 2015; 373: 123-135Crossref PubMed Scopus (6236) Google Scholar, 2Borghaei H. Paz-Ares L. Horn L. et al.Nivolumab versus docetaxel in advanced nonsquamous non-small-cell lung cancer.N Engl J Med. 2015; 373: 1627-1639Crossref PubMed Scopus (6773) Google Scholar, 3Garon E.B. Rizvi N.A. Hui R. et al.Pembrolizumab for the treatment of non-small-cell lung cancer.N Engl J Med. 2015; 372: 2018-2028Crossref PubMed Scopus (4358) Google Scholar, 4Fehrenbacher L. Spira A. Ballinger M. et al.Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial.Lancet. 2016; 387: 1837-1846Abstract Full Text Full Text PDF PubMed Scopus (1990) Google Scholar, 5Herbst R.S. Baas P. Kim D.W. et al.Pembrolizumab versus docetaxel for previously treated, PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised controlled trial.Lancet. 2016; 387: 1540-1550Abstract Full Text Full Text PDF PubMed Scopus (4549) Google Scholar, 6Reck M. Rodríguez-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 (6215) Google Scholar By interrupting the interaction of PD-1 and PD-L1, the negative regulatory effects of this immune checkpoint on specific T-cell–driven immune responses are reversed. These treatments, however, work in only around 10% to 20% of unselected patients with advanced-stage NSCLC. Although there may be other generic immune inhibitory mechanisms, such as inhibitory tumor metabolism, or inhibitory cells or soluble factors in the tumor microenvironment that are active in the tumor,7Blank C.U. Haanen J.B. Ribas A. et al.CANCER IMMUNOLOGY. The "cancer immunogram.".Science. 2016; 352: 658-660Crossref PubMed Scopus (504) Google Scholar we presume that patients deriving benefit from such therapy are among those in whom PD1–PD-L1 interaction is inhibiting an available, specific immune response against the patient's disease. A characteristic of those tumors should be expression of PD-L1 on the surface of tumor cells (TCs) and/or tumor-infiltrating immune cells (ICs). And so it seems to be. In almost all trials of anti–PD-1 or anti–PD-L1 monotherapy reported so far, there is a consistent finding of greater enrichment of response and survival benefit in treatment groups defined by higher levels of PD-L1 expression as assessed by using PD-L1 immunohistochemistry (IHC).8Abdel-Rahmann O. Correlation between PD-L1 expression and outcome of NSCLC patients treated with anti-PD1/PD-L1 agents. A meta-analysis.Crit Rev Haematol Oncol. 2016; 101: 75-85Abstract Full Text Full Text PDF PubMed Scopus (98) Google Scholar PD-L1 IHC is not a perfect biomarker, and it has probably been unfairly compared with addictive oncogene biomarkers that dichotomized patients with NSCLC more clearly, and targeting addictive oncogenic drivers is a very effective therapeutic strategy. In immunotherapy the drugs and biomarkers perform very differently. Although PD-L1 is effectively the target of the drugs in question, PD1–PD-L1 interaction is only one of many factors that may determine the eventual outcome of a potential tumor-directed immune response.7Blank C.U. Haanen J.B. Ribas A. et al.CANCER IMMUNOLOGY. The "cancer immunogram.".Science. 2016; 352: 658-660Crossref PubMed Scopus (504) Google Scholar Furthermore, both assessment of the biomarker and definition of the treatment group on the basis of biomarker data have issues. PD-L1 expression on TCs or ICs is heterogeneous, meaning that sampling error during tissue acquisition may be a significant factor that is more likely to lead to underestimation rather than overestimation of PD-L1 expression. This may be a major reason why treatment responses are occasionally seen in cohorts of patients deemed biomarker negative (i.e. expressing PD-L1 at a level below a treatment-determining threshold). Also, PD-L1 expression is not an all or nothing phenomenon. It is a biological continuum that is more easily understood in the context of TC expression, embracing cases ranging from no PD-L1 expression to those in which all TCs express PD-L1 (tumor proportion score [TPS] 100%). The treatment group, however, is defined as all patients whose PD-L1 TPS is above a certain threshold. Outcome data are presented as an average for the treatment cohort; yet especially when the threshold is set very low (e.g., 1%), it is likely that expectations for benefit should be very different between those patients who just qualify for treatment (e.g., TPS 1% and 50%AtezolizumabGenentech/RocheSP142 (Ventana)Ventana BenchMark ULTRATC > 1%, 10%, and 50%IC > 1%, 5%, and 10%DurvalumabAstraZenecaSP263 (Ventana)VentanaBenchmarkTC ≥ 25%AvelumabPfizer/Merck Serono73-10 (Dako)Dako Link 48TC ≥ 1%, 50%, and 80%PD-L1, programmed death ligand 1; Ab, antibody; TC, tumor cells by percentage staining for PD-L1; IC, percentage of tumor area infiltrated by PD-L1–positive immune cells.a Variable according to trials and line of therapy. Open table in a new tab PD-L1, programmed death ligand 1; Ab, antibody; TC, tumor cells by percentage staining for PD-L1; IC, percentage of tumor area infiltrated by PD-L1–positive immune cells. In this month's issue of the Journal of Thoracic Oncology, Hendry et al. present the latest comparative study of trial-validated PD-L1 assays in cases of NSCLC, the second largest of the series.13Hendry S. Byrne D.J. Wright G.M. et al.Comparison of four PD-L1 immunohistochemistry assays in lung cancer.J Thorac Oncol. 2018; 13: 367-376Abstract Full Text Full Text PDF Scopus (113) Google Scholar Previous studies have more or less agreed that at a technical level, the trial-validated assays based on the 28-8, 22C3, and SP263 clones were comparable whereas the SP142 assay was different.9Tsao M.S. Kerr K.M. Dacic S. Yatabe Y. Hirsch F.R. IASLC Atlas of PD-L1 Immunohistochemistry Testing in Lung Cancer. Editorial Rx Press, North Fort Myers, FL2017Google Scholar, 14Scheel A.H. Dietel M. Heukamp L.C. et al.Harmonized PD-L1 immunohistochemistry for pulmonary squamous-cell and adenocarcinomas.Mod Pathol. 2016; 29: 1165-1172Crossref PubMed Scopus (292) Google Scholar, 15Hirsch F.R. McElhinny A. Stanforth D. et al.PD-L1 immunohistochemistry assays for lung cancer: results from phase 1 of the Blueprint PD-L1 IHC Assay Comparison Project.J Thorac Oncol. 2017; 12: 208-222Abstract Full Text Full Text PDF PubMed Scopus (936) Google Scholar, 16Ratcliffe M.J. Sharpe A. Midha A. et al.Agreement between programmed cell death ligand-1 diagnostic assays across multiple protein expression cut-offs in non-small cell lung cancer.Clin Cancer Res. 2017; 23: 3585-3591Crossref PubMed Scopus (242) Google Scholar, 17Rimm D.L. Han G. Taube J.M. et al.A prospective multi-institutional, pathologist-based assessment of four immunohistochemistry assays for PD-L1 expression in non-small cell lung cancer.JAMA Oncol. 2017; 3: 1051-1058Crossref PubMed Scopus (565) Google Scholar, 18Adam J. Rouquette I. Damotte D. et al.Multicentric French harmonization study for PD-L1 IHC testing in non-small cell lung cancer. IASLC World Conference on Lung Cancer, Vienna, 2016.J Thorac Oncol. 2016; 11 (PL04.04a)Google Scholar None of these studies showed identical outcomes in terms of case allocation around various cutpoints, but there is no consensus on what degree of statistically expressed agreement is clinically acceptable. In Blueprint phase 1, the percentage match in terms of case allocation around a 1% cutoff with use of alternate assays was, at best, around 85% to 90%.15Hirsch F.R. McElhinny A. Stanforth D. et al.PD-L1 immunohistochemistry assays for lung cancer: results from phase 1 of the Blueprint PD-L1 IHC Assay Comparison Project.J Thorac Oncol. 2017; 12: 208-222Abstract Full Text Full Text PDF PubMed Scopus (936) Google Scholar The decision to award the SP263 assay European Conformity–marked status for use as an alternative to both the 28-8 assay with nivolumab and the 22C3 assay with pembrolizumab was presumably based on these technical data because there were no clinical outcome data based on alternative assays in the public domain. This decision was welcome because laboratory access to Ventana platforms (Ventana Medical Systems, Tucson, AZ) is generally greater than for Dako platforms (Dako, Glostrup, Denmark). The implication here was that we could use either of these three trial validated assays interchangeably and expect "comparable" technical performance; so far so good. The conclusion from Hendry et al.13Hendry S. Byrne D.J. Wright G.M. et al.Comparison of four PD-L1 immunohistochemistry assays in lung cancer.J Thorac Oncol. 2018; 13: 367-376Abstract Full Text Full Text PDF Scopus (113) Google Scholar that the four assays used are not comparable may therefore initially cause alarm. However, the largest significant difference in this article was the lack of TC staining using the SP142 assay, which is a consistent finding in all comparative studies and one that does not affect the key debate around the comparability of the three other assays. Nonetheless, the data regarding the SP263 assay should make us pause for thought. In this study, the SP263 assay scored significantly more cases (34.6%) at or over the 1% threshold when compared with the 22C3 or 28-8 assays (23.9 and 25.9%, respectively), which translates to about 10% more patients at or above the 1% threshold. For the 50% cutpoint, however, there was no significant difference between the three assays. Furthermore, the differences in κ scores in pairwise assessments are not great; they are best between 22C3 and 28-8, but the scores comparing SP263 with these assays are barely worse, especially for the 25% and 50% cutpoints. These findings are, in broad terms, not greatly different from those in the Blueprint phase 1 study.15Hirsch F.R. McElhinny A. Stanforth D. et al.PD-L1 immunohistochemistry assays for lung cancer: results from phase 1 of the Blueprint PD-L1 IHC Assay Comparison Project.J Thorac Oncol. 2017; 12: 208-222Abstract Full Text Full Text PDF PubMed Scopus (936) Google Scholar These data imply that the SP263 assay could put more patients over the 1% threshold but not over the 50% threshold. So although there may be no implication for use of this alternative assay for pembrolizumab in the first-line setting, more patients would qualify for this drug in the second-line companion diagnostic setting or receive a more optimistic prognosis when considering a complementary diagnostic test with nivolumab. Given that we do see limited responses in biomarker-negative patients with these drugs in this setting, this may be an insignificant issue. Alternatively, there must be some question about the likely benefit to be gained for these patients who "qualify" for treatment with TPS, which is marginally over the 1% threshold. We have seen very few data for response or outcome for these particular patients (TPS 1–10 or 25%, for example), but data from the Keynote 001 trial suggest that the benefit is no better than for PD-L1–negative patients. The SP263 assay is acknowledged to be a more sensitive assay with stronger staining characteristics.9Tsao M.S. Kerr K.M. Dacic S. Yatabe Y. Hirsch F.R. IASLC Atlas of PD-L1 Immunohistochemistry Testing in Lung Cancer. Editorial Rx Press, North Fort Myers, FL2017Google Scholar, 16Ratcliffe M.J. Sharpe A. Midha A. et al.Agreement between programmed cell death ligand-1 diagnostic assays across multiple protein expression cut-offs in non-small cell lung cancer.Clin Cancer Res. 2017; 23: 3585-3591Crossref PubMed Scopus (242) Google Scholar Scheel at al.14Scheel A.H. Dietel M. Heukamp L.C. et al.Harmonized PD-L1 immunohistochemistry for pulmonary squamous-cell and adenocarcinomas.Mod Pathol. 2016; 29: 1165-1172Crossref PubMed Scopus (292) Google Scholar also reported similarly, but the difference in added "positive" cases was not significant. Why Hendry et al. have produced this slightly uncomfortable outcome with the SP263 assay is not immediately obvious, but there are some potential issues. The number of positive cases in this study was quite low (less than 10% of cases with TPS >50%), whereas in most trials this cohort is around one-third, but it is hard to judge the possible impact of this difference. The lower section thickness of 3 μm used by Hendry et al.13Hendry S. Byrne D.J. Wright G.M. et al.Comparison of four PD-L1 immunohistochemistry assays in lung cancer.J Thorac Oncol. 2018; 13: 367-376Abstract Full Text Full Text PDF Scopus (113) Google Scholar might lead to underrecognition of weaker staining. The scoring was carried out by only one pathologist over a 10-month period. The pathologist was trained in 22C3 reading only, which is possibly an issue given that SP263 staining characteristics are acknowledged as different from those of 22C3 and 28-8, and although on the learning curve for implementing new techniques, all pathologists will change the way they do things until a level is reached with experience. We have no way of knowing whether this could have affected the relative scores determined for 22C3 and 28-8 assays versus for SP263 over this long period. Because the SP263 assay stained TCs more strongly (and the authors comment that this assay was easier to read for that reason), it is conceivable that, in the context of a subjective assessment such as PD-L1 IHC scoring, stronger staining means that positive cells are more likely to be recognized. The particular issues with making this call around the 1% level, as demonstrated in the recently reported Blueprint phase 2,19Tsao M.S. Kerr K.M. Yatabe Y. et al.PL 03.03 Blueprint 2: PD-L1 immunohistochemistry comparability study in real-life, clinical samples.J Thorac Oncol. 2017; 12: S1606Abstract Full Text Full Text PDF PubMed Google Scholar raise the possibility that at least some pathologists might make a higher score with more strongly stained cells. However, the aforementioned argument regarding the true clinical significance of a low PD-L1 TPS score just over 1% still prevails; we do not know which assay gives the "true" result. A more sensitive assay perhaps means more patients in the treatment group, but it might also mean fewer patients benefiting. On the other hand, a less sensitive assay may select fewer patients with a higher probability of benefiting, but with the risk of leaving behind in a "biomarker-negative" group some who might benefit (see later in this editorial). Of course, in the second-line NSCLC setting, immunotherapy is preferred, even if the trial data match only the limited performance of docetaxel because of better toxicity profiles. Overall, this study by Hendry et al.13Hendry S. Byrne D.J. Wright G.M. et al.Comparison of four PD-L1 immunohistochemistry assays in lung cancer.J Thorac Oncol. 2018; 13: 367-376Abstract Full Text Full Text PDF Scopus (113) Google Scholar is a valuable addition to the literature, and although the article is thought-provoking, it should not undermine recent developments in PD-L1 IHC testing. There are other observations in this article regarding laboratory-developed tests and storage of cut section that inform the wider discussions around PD-L1 IHC, but they are beyond the remit of this editorial. The aforementioned studies make technical comparison of these assays and gauge the consistency of likely patient allocation to treatment groups defined by PD-L1 IHC scores allied to particular drugs and indications. None of these studies contained any clinical outcome data showing that the alternative way of selecting patients for treatment actually resulted in the same outcome for patients. In this edition of the Journal of Thoracic Oncology, Fujimoto et al. provide some such evidence.20Fujimoto D. Sato Y. Uehara K. et al.Predictive performance of four programmed cell death ligand 1 assay systems on nivolumab response in previously treated patients with non-small cell lung cancer.J Thorac Oncol. 2018; 13: 377-386Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar At the outset, we need to note that this is a small study of only 40 patients, all treated in the second-line setting with nivolumab, a scenario that in practice, is allied only with a complementary diagnostic test. Nonetheless, we may regard these data as proof of principle and a first step. This study also presented yet more comparative data on staining performance and reached the same basic conclusions as described earlier in this editorial, namely, that the 22C3, 28-8, and SP263 based assays used here showed similar technical performance and SP142 demonstrated the expected lower staining of TC. It is implied, though not explicit, that the assays used were those validated in the clinical trials; no comment is made regarding any training in PD-L1 assay evaluation received by any of the pathologists who scored the cases. Further analysis of this aspect of the article is not warranted in this review. It is, of course, the clinical outcome data that are of more interest. The receiver operating characteristic analysis data infer that the 28-8, 22C3, and SP263 assays used all performed equally well, and their performance was equivalent to that of the 22C3 assay reported in the Keynote 001 trial3Garon E.B. Rizvi N.A. Hui R. et al.Pembrolizumab for the treatment of non-small-cell lung cancer.N Engl J Med. 2015; 372: 2018-2028Crossref PubMed Scopus (4358) Google Scholar whereas the SP142 assay did less well. The data presented by Fujimoto et al.20Fujimoto D. Sato Y. Uehara K. et al.Predictive performance of four programmed cell death ligand 1 assay systems on nivolumab response in previously treated patients with non-small cell lung cancer.J Thorac Oncol. 2018; 13: 377-386Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar in their table 3 prompt the following observations, although little more should be inferred because the numbers in the subgroups are very small. The responses to second-line nivolumab for patients with a TPS of 50% or higher are impressive; equally notable are the poor responses in each of the TPS score groupings 1% to 49% and 1% or lower. This issue has already been discussed in this editorial. The exception might be with the SP142 assay; here one could argue that the slightly better outcomes in the lower-expressing cohort reflects the possibility of true higher expressers being consigned to this group by a less sensitive test. Although yet to be published, a retrospective analysis of alternative biomarker assessment of samples in the OAK trial demonstrated that the 22C3 assay was also effective in selecting patients for second-line atezolizumab at a 50% cutoff when compared with the TC3/IC3 assay algorithm using the SP142 assay.21Oncology Pro. Clinical efficacy of atezolizumab (Atezo) in PD-L1 subgroups defined by SP142 and 22C3 IHC assays in 2L+ NSCLC: results from the Randomized OAK Study. http://oncologypro.esmo.org/Meeting-Resources/ESMO-2017-Congress/Clinical-Efficacy-of-Atezolizumab-Atezo-in-PD-L1-Subgroups-Defined-by-SP142-and-22C3-IHC-Assays-in-2L-NSCLC-Results-From-the-Randomized-OAK-Study. Accessed January 21, 2018.Google Scholar More such data are desperately required to settle this argument about the clinical applicability of alternative PD-L1 IHC assays. Both these studies represent small forward steps in the journey toward harmonization of PD-L1 IHC biomarker testing. Testing is clearly an effective enrichment tool, and although alternative biomarker approaches are enthusiastically sought, the fundamental importance of PD-L1 as a biomarker for PD1/PD-L1 inhibitory therapy cannot be ignored. The preliminary results of the Blueprint 2 study underpin these latest findings from Hendry et al.13Hendry S. Byrne D.J. Wright G.M. et al.Comparison of four PD-L1 immunohistochemistry assays in lung cancer.J Thorac Oncol. 2018; 13: 367-376Abstract Full Text Full Text PDF Scopus (113) Google Scholar and also introduce the latest trial-validated assay for avelumab based on the 73-10 clone into the fray. These PD-L1 biomarkers will be with us for the foreseeable future, but effort is still required to make testing easier and more accessible without compromising the clinical efficacy of the testing data. Predictive Performance of Four Programmed Cell Death Ligand 1 Assay Systems on Nivolumab Response in Previously Treated Patients with Non–Small Cell Lung CancerJournal of Thoracic OncologyVol. 13Issue 3PreviewNivolumab has demonstrated efficacy against metastatic NSCLC. Four programmed cell death ligand 1 (PD-L1) immunohistochemistry (IHC) assay systems are available for identification of responders among patients with NSCLC, and these assays show some differing characteristics. Accordingly, in this study, we evaluated the ability of these assays to identify responders to nivolumab therapy. Full-Text PDF Open ArchiveComparison of Four PD-L1 Immunohistochemical Assays in Lung CancerJournal of Thoracic OncologyVol. 13Issue 3PreviewFour different programmed death ligand 1 immunohistochemical assays are approved or in development as companion or complementary diagnostics to different immunotherapeutic agents in lung carcinoma. We sought to determine whether these assays are technically equivalent and whether one antibody can be used on an alternate staining platform. Full-Text PDF Open ArchiveErratum to: Kerr KM. The PD-L1 immunohistochemistry biomarker: two steps forward, one step back? J Thorac Oncol. 2018;13(3):291-294.Journal of Thoracic OncologyVol. 13Issue 8PreviewIn the above-referenced article, there is an error in Table 1, which has been corrected below (correction in bold): Full-Text PDF Open ArchiveResponse to Editorial: The PD-L1 Immunohistochemistry Biomarker: Two Steps Forward, One Step Back?Journal of Thoracic OncologyVol. 13Issue 9PreviewWe are grateful to the author for the thoughtful and thought-provoking editorial regarding our study comparing four programmed death ligand 1 (PD-L1) immunohistochemical assays in lung cancer.1,2 As with several other studies, the staining characteristics of SP142 (Ventana, Tucson, Arizona) diverged from the other antibodies; however, we still question if the variation of the SP263, 22C3, and 28-8 assays in our study and others is clinically acceptable, and whether pathologist training influenced our findings. Full-Text PDF Open Archive
Referência(s)