Extended Data Fig. 5: Reductions in ctDNA levels occurred at a higher rate in the atezolizumab arm than in the observation arm and were associated with improvements in clinical outcome in the atezolizumab arm.

a–e, Reduction is assessed in C1D1 ctDNA(+) patients in the C1/C3 BEP and is defined as a decrease in sample MTM/ml from C1D1 to C3D1. The proportion of patients who are ctDNA(+) at C1D1 who have reduced ctDNA by C3D1 for the atezolizumab arm (blue) and the observation arm (red) (a). Kaplan–Meier analyses comparing patients who have reduced ctDNA (‘reduction’, dark blue or dark red) compared with those who have ctDNA levels that increase (‘non-reduction’, light blue or light red) for DFS in the atezolizumab arm (b), DFS in the observation arm (c), OS in the atezolizumab arm (d) and OS in the observation arm (e). f–i, Patients shown are from the atezolizumab arm C1/C3 BEP and are ctDNA(+) at baseline and have different types of ctDNA dynamics based on the percent change in sample MTM/ml from C1 to C3. Note that the scale for the per cent change goes from −100% (clearance) to infinity, where negative values indicate reductions, and positive values indicate increases. Kaplan–Meier analysis of DFS (f) and OS (h) in which ctDNA reduction is split into patients who clear ctDNA (‘reduction with clearance’, dark blue, solid lines) and those who have decreased ctDNA without clearance (‘reduction without clearance’, dark blue, dashed lines). Patients with an increase in ctDNA are also shown (‘increase’, light blue, solid lines). Forest plots for DFS (g) and comparing patients with reduction (anywhere from clearance (−100% change) to minor decreases in ctDNA (<0% change)) (i) using different thresholds for percent change in sample MTM/ml, including −100% (reduction with clearance versus reduction without clearance), −50%, −25% and −10% changes. Forest plots show HRs for recurrence or death estimated using a univariable Cox proportional-hazards model, and 95% confidence intervals of HRs are represented by horizontal bars.