Fig. 3
A survival scatter plot shows that ∆ξ is a predictor of differential patient sensitivity. For the subC-LP signature, observed PFS times are plotted versus the cross-validated differential log-hazard-ratio ∆ξ for each the n = 209 patients considered in the AFLAME panel. Points for control and aflibercept arm patients are displayed in red and blue, respectively,with a mix of censored (full dots) and uncensored (open dots) data in each category. The binary classification into “sensitive” (S) and “relatively-resistant” (R) groups that obtains with threshold ∆ξc = − 0.815 (dashed vertical line) is indicated by the arrows at the top of the plot, with resulting aflibercept-to-control hazard ratios hR shown alongside. For ∆ξ ≤ − 0.815, it can be seen that patients in the treatment arm (blue) tend to have longer survival than those in the control arm (red), with many blue dots higher than red dots, and that this asymmetry is greater than for patients with ∆ξ > − 0.815