Composite versus individual biomarkers for predicting clinical benefit to PD-1 blockade in NSCLC 93 3 (95% CI: 0.65-0.92) for CD3+IT-CD8 (Fig. 2C,D). Corresponding cut-off values of 0.167 and 0.161, respectively, were associated with a sensitivity of 94% for both, specificity of 62% and 54%, NPV of 96% and 95% and PPV of 50% and 47% (Table 3). Furthermore, PD-1T TIL numbers and TIS scores were significantly higher in the DC 6m group compared to the PD group (PD-1T TILs, P<0.001 and TIS, P<0.01) (Fig. S2C,D). PD-1T TILs showed an AUC of 0.82 (95% CI: 0.69-0.95), and TIS demonstrated an AUC of 0.81 (95% CI: 0.65-0.98) (Fig. S2E,F). For PD-1T TILs, a cut-off value of 90 per mm2 was selected based on its predictive value in a previous study25. This threshold yielded a sensitivity of 92%, specificity of 67%, NPV of 95% and PPV of 52% (Table 3). A score of 6.65 was identified as the optimal cut-off for TIS, resulting in a sensitivity of 100%, specificity of 55%, NPV of 100% and PPV of 47% (Table 3). Next, we evaluated the predictive performance of the four selected biomarkers (CD8+ITCD8, CD3+IT-CD8, PD-1T TILs and TIS) in the validation cohort. The number of samples with successfully obtained biomarker results in the validation cohort ranged from 40 to 79 (Table 1, Fig. S1). A decrease in the predictive accuracy of CD8+IT-CD8 and CD3+ITCD8 biomarkers was observed when compared to the training cohort, as reflected by the AUC of the ROC curve measuring 0.62 (95% CI: 0.50-0.75) and 0.68 (95% CI: 0.55-0.80), respectively (Fig. 2C,D). Moreover, the probability scores within the DC 6m group did not significantly differ from those in the PD group for CD8+IT-CD8 (P=0.08) (Fig. 2E). This comparison for CD3+IT-CD8 was borderline significant (P=0.01) (Fig. 2F). The selected cut-off value of CD8+IT-CD8 reached a sensitivity of 64%, specificity of 56%, NPV of 76% and PPV of 41%. The predictive accuracy of CD3+IT-CD8, while higher than that of CD8+IT-CD8, remained lower than in the training cohort (sensitivity: 83%, specificity: 46%, NPV: 85% and PPV: 43%) (Table 3). The individual biomarker analysis in the validation cohort showed significantly higher PD-1T TIL numbers in the DC 6m group versus the PD group (P<0.01), whereas no significant difference was observed for TIS scores (P=0.52) (Fig. S2G,H). Although the discriminatory ability of PD-1T TILs was lower compared to the training, it still reached an AUC of 0.72 (95% CI: 0.57-0.87) (Fig. S2E). A cut-off value of 90 PD-1T TILs per mm2 resulted in a sensitivity of 72%, specificity of 74%, NPV of 86% and PPV of 54%. Conversely, TIS obtained a low AUC of 0.57 (95% CI: 0.36-0.77) (Fig. S2F). A TIS score of 6.65 showed a comparable sensitivity (83%), NPV (84%) and PPV (37%) but lower specificity (39%) (Table 3). In summary, these results demonstrate that the combination of CD8+IT-CD8 and CD3+IT-CD8 did not improve predictive accuracy compared to the standalone use of PD-1T TILs and TIS. Furthermore, none of the selected biomarkers met the prespecified performance criteria.
RkJQdWJsaXNoZXIy MTk4NDMw