Fig. 4: Association between peripheral blood immune populations as determined by mass cytometry (CyTOF) or peripheral blood or urine protein analytes and cCR or metastasis-free survival.

a, Timing of sample collection. b, Volcano plot for metastasis-free survival (MFS) based on C1D1 peripheral blood CyTOF data (n = 74; patients without samples or outcome data were excluded) showing log-rank test (y axis) and Cox regression hazard ratio (x axis). c, Kaplan–Meier curves for CD4⁺ naive T cells on C1D1 with Gehan–Breslow P values. d, Volcano plot for MFS based on C3D1 peripheral blood CyTOF data (n = 65; patients without samples or outcome data were excluded) showing log-rank test (y axis) and Cox regression hazard ratio (x axis). e, Kaplan–Meier curves with Gehan–Breslow P values for CD8⁺ naive T cells at C3D1. f, Volcano plot showing the differential expression of peripheral blood proteins on C1D1 (n = 71; patients without samples or outcome data were excluded) according to cCR status (−log₁₀(P value) in x axis and log₂ fold change calculated using moderated t-statistic). g, Volcano plot showing the differential expression of peripheral blood proteins on C3D1 (n = 65; patients without samples or outcome data were excluded) according to cCR status (−log₁₀(P value) in x axis and log₂ fold change calculated using moderated t-statistic). h, Heat map showing the normalized expression (z-score) for peripheral blood proteins increased on C3D1 according to cCR status (hierarchically clustered using Ward’s algorithm). i, Volcano plot for MFS based on C1D1 expression of proteins showing log-rank test (y axis) and Cox regression hazard ratio (x axis). j, Kaplan–Meier curves for IL6 expression at C1D1 with Gehan–Breslow P values. k, Volcano plot for MFS based on C3D1 protein expression showing log-rank test (y axis) and Cox regression hazard ratio (x axis). l, Kaplan–Meier curves for ANGPT2 expression at C3D1 with Gehan–Breslow P values. m, Volcano plot for MFS based on urine proteins at time of clinical restaging (n = 59, patients without samples or outcome data were excluded) showing log-rank test (y axis) and Cox regression hazard ratio (x axis). For associations with cCR, adjustment for multiple comparisons was performed using the Benjamini–Hochberg method. a was created with BioRender. FDR, false discovery rate; HR, hazard ratio; Nivo, nivolumab; NA, not applicable.