Chapter 3 98 Table 3. Predictive accuracy of selected individual and composite biomarkers, summary of training and validation results Training Clinical outcome Bio-marker Predictor Cut-off Samples (n) AUC (95% CI) DC 6 months PD-1T TILs 90 42 0.82 (0.69-0.95) TIS 6.65 28 0.81 (0.65-0.98) CD8+ IT-CD-8 = 1 1 −C De8x p+( −03. .054734 9∗ +I T 0−. 0C0D381) ∗ 0.167 55 0.83 (0.73-0.94) CD3+ IT-CD-8 = 1 1C D−3e x+p ( 0−. 02 1. 37852 1∗ +I T 0−. 0C8D086 +∗ 0.0069 ∗ CD3 ∗ IT−CD8) 0.161 53 0.78 (0.65-0.91) DC 12 months PD-1T TILs 90 42 0.82 (0.70-0.94) TIS 6.65 28 0.77 (0.58-0.96) CD8+ IT-CD-8 = 1 1 − exp (−4.0644 + 0.003 ∗ 8 + 0.0436 ∗ − 8) 0.122 55 0.85 (0.73-0.96) CD8+ TIS = 1 1 −CeDx8p (+− 50..72935426 +∗ T0I. 0S 2+2 4 ∗ − 0.0021 ∗ CD8 ∗ TIS) 0.124 28 0.91 (0.79-1.00) Accuracy of individual and composite biomarkers to predict DC at 12 months Approximately 70-80% of patients undergoing second-line treatment with PD-1/ PD-L1 blockade experience disease progression within 12 months2–4. Our previous work demonstrated the superior efficacy of PD-1T TILs in identifying patients with DC at 12 months (DC 12m) compared to those with DC 6m, as well as identifying a subgroup without long-term benefit25. In light of this, we extended our analysis to evaluate the predictive performance of all biomarkers for DC 12m. Similar to the analysis performed for DC 6m, we constructed ROC curves to determine optimal cut-off values corresponding to a sensitivity of ≥90% and specificity of ≥50% for each composite and individual biomarker. Four patients in the training and nine patients in the validation cohort experienced disease progression between 6 and 12 months, allocating them into the PD group. In the training cohort, 16 composite biomarkers, along with individual biomarkers PD-1T TILs and TIS, met the prespecified sensitivity and specificity ciritera (Table S3). PD-1T TIL and TIS combinations did not significantly improve predictive accuracy (Fig. S3A,B and Table S3). Noteworthy, the combination of CD8 with TIS (CD8+TIS) showed an 18% increase in specificity compared to TIS
RkJQdWJsaXNoZXIy MTk4NDMw