Integration of longitudinal deep-radiomics and clinical data improves the prediction of durable benefits to anti-PD-1/PD-L1 immunotherapy in advanced NSCLC patients

Table 6 Hazard ratios and C-indexes of longitudinal and ensemble models trained for endpoint PFS9 to predict PFS and OS in the independent test set

Model	Features	PFS			OS
Model	Features	HR [95% CI]	p-value	C-index	HR [95% CI]	p-value	C-index
RF-longitudinal	Clinical data	0.52 [-1.16,2.20]	0.542	0.575	1.73 [-0.67,4.13]	0.157	0.613
RF-longitudinal	DF-imm	1.35 [-0.23,2.92]	0.093	0.642	1.72 [-0.18,3.62]	0.076	0.641
Ensemble RF-longitudinal	DF-imm Clinical data	2.38 [0.23,4.54]	0.030	0.685	2.94 [0.40,5.48]	0.023	0.736

ISSN: 1479-5876