Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Construction and optimization of gene expression signatures for prediction of survival in two-arm clinical trials

Fig. 2

Role of the {treatment-arm × gene expression} interaction term in the Cox models. a Schematic equation for the log-hazard-ratio ξ as a function of treatment arm and gene expression, showing both direct and interaction terms. b ‘no interaction’ case: log-hazard-ratio profiles in the absence of interaction effects (γ = 0), with β0, β1 < 0: the aflibercept treatment arm is always favored, and the signature is only prognostic, with the same dependence on gene expression in both treatment arms. c ‘moderate interaction case’: with γ > 0 and β0, β1 < 0, and the range of x indicated, all patients benefit from aflibercept but some more than others. d ‘strong interaction case’: with γ < 0, β0 < 0 and | γ | > β1 > 0, the optimal treatment may ‘flip’ depending on x

Back to article page