Fig. 5: Association of thymic health with overall survival in patients treated with immunotherapy for various cancer types.

a–f, Kaplan–Meier estimates and Cox proportional hazards models of OS according to thymic health (an imaging-based proxy for thymic functionality). Analyses show the comparison of high or average thymic health to the reference of low thymic health. The forest plots show the same data after multivariate adjustments for sex and continuous age. Patients with melanoma (a), breast (b), renal (c), bladder (d), oesophageal (e) and pooled other cancer types with fewer than 100 cases per entity (f). a–f, Cox proportional hazards regression was used to estimate HRs. In the forest plots, the centre of each box represents the estimated HR, and the whiskers denote the corresponding 95% CI; arrowheads indicate that the 95% CI extends beyond the visualized limits; shaded box size is for visualization only and does not encode statistical weight. The overall contribution of thymic health to uni- or multivariable models was evaluated using likelihood ratio tests (χ² tests) comparing full models with nested models excluding thymic health (type III test, two-sided), with no adjustments for multiple comparisons.