Chapter 5 190 The results obtained in this study do not stand alone. Weber and colleagues identified a protein classifier from sera of melanoma patients treated with PD-1 inhibitors, employing the same technology that was used in our study. This was validated in multiple patient cohorts treated with PD-1 inhibitors and CTLA4 antagonists14. As here, they were able to identify, prior to initiation of treatment, patients who had a favorable outcome following treatment. Biological processes associated with that classifier included complement, wound healing and acute phase pathways, all upregulated in the poor prognosis group, corroborating our results. Further evidence that the pretreatment circulating proteome provides important information on checkpoint efficacy was provided in the context of a phase II study where atezolizumab was compared with docetaxel as second-line treatment in 272 advanced NSCLC patients33. Similar to our results, a serum protein classifier was established that identified patients with poor (median OS 7.3 months, n=60 (45%)) and good (median OS not reached, n= 73 (55%)) outcomes. This classifier was shown in blinded validation to be predictive for atezolizumab vs. docetaxel for OS and PFS (interaction P<0.01). In that study, as in our own, there was no association between test classification and tumor PD-L1 expression; there was also no association with TMB. Also, among the biological processes that were most significantly associated with classification by PSEA, acute inflammation and complement activation ranked in the top three. There are some limitations to our results. Obviously, the number of patients is low and all three immunotherapy-treated cohorts come from one geographic area and were investigated retrospectively. Also, for historical reasons, validation blinded to all clinical data was only possible for validation set 2. Although we made strong efforts to obtain sufficient tumor tissue samples, we were not able to obtain PD-L1 expression data on all patients. Several factors contributed to this: many patients are diagnosed on the basis of cytology alone and so have no tissue available for PD-L1 analysis; at the time of treatment initiation for these patients, use of PD-L1 expression was still somewhat investigational; and positive PD-L1 expression status is not mandatory for administration of nivolumab in the second and higher line setting. Unfortunately, TMB data was not collected. Investigation of larger cohorts with more complete information on TMB and PD-L1 expression would be useful to examine with more precision the level of association of these markers and how much complementary information each can provide to predict outcome. The nonimmunotherapy-treated control set is small and restricted to one therapy. It would be of interest to study the performance of the test in larger control cohorts in other standard-of-care non-immunotherapy regimens to be able to explore the test’s predictive potential.
RkJQdWJsaXNoZXIy MTk4NDMw