T cell exhaustion markers in hepatitis C virus-related hepatocellular carcinoma: Expression patterns and prognostic significance in an Egyptian cohort

Table 8 Prognostic modeling and receiver operating characteristic curve performance for hepatocellular carcinoma mortality

Variable	Random forest AUC (95%CI)	Logistic regression OR (95%CI)	P value	ROC AUC (95%CI)	Cut-off	Sensitivity (%)	Specificity (%)
CD4/PD-1 (%)	0.92 (0.85-0.98)	2.3 (1.4-3.8)	< 0.001a	0.92 (0.85-0.98)	> 35%	88	85
CD8/TIM3 (%)	0.90 (0.83-0.97)	2.5 (1.5-4.2)	< 0.001a	0.90 (0.83-0.97)	> 17%	85	82
AFP (ng/mL)		3.2 (1.8-5.7)	< 0.001a	0.89 (0.81-0.96)	> 13000	90	80
AFP + CD8/TIM3	0.95 (0.90-0.99)		< 0.001a	0.95 (0.90-0.99)	> 13000, > 17%	92	90
BCLC stage (D vs A-C)		3.8 (2.0-7.2)	< 0.001a
Clustering: High-expression		50% mortality (n = 30)	< 0.001a		Elevated (PD-1, TIGIT, TIM-3)
Clustering: Low-expression		10% mortality (n = 20)	< 0.001a		Low (PD-1, TIGIT, TIM-3)

^aP < 0.001.

Random forest and logistic regression analyses, with receiver operating characteristic cut-offs determined by Youden’s index. Clustering based on K-means stratification using programmed death 1, T cell immunoglobulin and ITIM domain, and T cell immunoglobulin and mucin domain-containing protein 3 expression levels (standardized z-scores). PD-1: Programmed death 1; TIGIT: T-cell immunoglobulin and ITIM domain; TIM-3: T cell immunoglobulin and mucin domain-containing protein 3; AFP: Alpha-fetoprotein; BCLC: Barcelona Clinic Liver Cancer; AUC: Area under the curve; OR: Odds ratio; CI: Confidence interval; ROC: Receiver operating characteristic.