Skip to main content
. 2025 Feb 25;13(2):e010541. doi: 10.1136/jitc-2024-010541

Figure 2. (A) Soluble plasma IO proteins (Olink) associated with treatments. Heatmap illustrating the temporal expression of soluble plasma proteins in patients randomized to Arm A or Arm B. Values represent Z-scores of the estimated marginal means (EMMs) derived from linear mixed-effects model for the interaction of treatment and time point, accounting for crossover (time period), tumor genotype and random patient-level effects. C1, C2, C3, P are blood collection time points at C1D1, C2D2, C4D1 and progression, respectively. Time points are grouped per treatment arms: Arm B nivo (blue) or Arm A nivo+cabo (red). Boxplots highlight selected biomarkers upregulated by nivo+cabo combination treatment (red frame: HO-1, FASLG), immunotherapy in general (green frame: PDCD1, CXCL10), or downregulated by nivo+cabo combination (blue frame: IL-12, VEGFR2). Trend lines represent adjusted marginal means per time point per arm with estimated 95% CIs. Statistical significance was inferred from post hoc marginal mean difference analysis for all possible pairwise comparisons: between time points per treatment or between treatments per time point. (B) Exploratory biomarkers associated with prior immunotherapy (prior IO) status in nivo+cabo treated patients from Arm A, Arm B after crossover, and Arm C. Values are z-scores of EMM of NPX for prior-IO status and time point interaction, controlling for genomics, histology, (carcinosarcoma yes or no) and patient-level random effect. Heatmap of selected biomarkers elevated at C1D1 in patients being exposed to prior IO. Time points are grouped per prior IO status: IO-naive (N, gray) or pretreated (Y, green). The middle boxplot shows temporal dynamics of PD-1 soluble levels. The bottom scatterplot shows the regression of PD-1 soluble levels (NPX values, log2) as a function of time to previous anti-PD-1 treatment in days. (C.) Association of tumor genomics subtypes, copy number high (CN-H) and copy number low (CN-L), with longitudinal profiles of soluble analytes in nivo+cabo treated patients from Arm A, Arm B after crossover, and Arm C. From this model, EMM of NPX for genomics and time point are computed, controlling prior-IO, carcinosarcoma, and patient-level variability. Left—boxplots showing GZMB and TNFRSF4/OX40 levels in CN-H versus CN-L during combination treatment. Right—clustermap showing signature of soluble analytes associated with differential patterns in CN-H versus CN-L. IL, interleukin; IO, immuno-oncology; NPX, normalized protein expression; PD-1, programmed cell death protein 1; VEGFR, vascular endothelial growth factor receptor.

Figure 2