Chapter 3 102 Discussion Since the introduction of PD-1/PD-L1 blockade therapy, the clinical outcomes for advanced stage NSCLC have markedly improved. Nevertheless, a minor subset of patients derive benefit from these treatments, leading to concerns of overtreatment and unnecessary side effects. In addition, healthcare systems deal with increasing costs. Several predictive biomarkers have been identified to support treatment decision-making. Given the complex interactions within the TME influencing the immune response during PD-1/PD-L1 blockade therapy, the likelihood of identifying a single perfect biomarker is minimal. Here, we evaluate the predictive performance of combinations of biomarkers in a cohort of advanced-stage NSCLC patients treated with nivolumab. Our findings show that the selected composite biomarkers did not improve predictive performance compared to PD-1T TILs and TIS alone. At 6 months, none of the selected biomarkers met the prespecified criteria of ≥90% sensitivity and ≥50% specificity in the validation cohort. However, at 12 months, PD-1T TILs and TIS demonstrated a high sensitivity in identifying patients with DC 12m with. Patients without long-term benefit were more accurately identified by PD-1T TILs than TIS. While CD8 or CD3 TILs, in combination with intratumoral localization of CD8 T cells, emerged as the most accurate composite biomarkers for DC 6m in the training cohort, their discriminatory ability was notably low in the validation cohort. The mere presence and localization of TILs may not indicate that all T cells are in a state to recognize and eliminate the tumor37,38. In the present study, this notion is supported by the high predictive accuracy of PD-1T TILs for DC 12m, given that these TILs constitute a distinct subset with an enhanced capacity for tumor recognition24. The consistency of these findings with our previous work can be attributed to the predominant reuse of samples25. Further refinement of this specific T cell population holds promise for the development of new markers or gene signatures, as shown by various studies39–41. Moreover, we recently developed a clinically applicable mRNA expression signature reflecting the presence of PD-1T TILs in the TME35. Since most biomarkers assessed in this study are associated with antitumor immunity and are presumably correlated, PD-1T combinations did not improve specificity. Previous studies have shown the predictive potential of combining CD8+PD-L122,34. However, in the training cohort, the CD8+PD-L1 combination failed to meet the sensitivity and specificity criteria, leading to its exclusion from further evaluation. PD-L1 expression on tumor cells is transient and relies on the production of IFNγ by TILs42. This dynamic nature of PD-L1 expression could contribute to variable tumor PD-L1 expression in the training samples, potentially impacting the predictive
RkJQdWJsaXNoZXIy MTk4NDMw