Background: In the phase 3 ARAMIS study (NCT02200614), darolutamide significantly improved metastasis-free survival in patients with non-metastatic castration-resistant prostate cancer (nmCRPC). Following the primary analysis, the study was unblinded, and placebo recipients were permitted to cross over to open-label darolutamide. Despite crossover, darolutamide significantly improved overall survival (OS). We conducted sensitivity analyses to estimate the effect of placebo–darolutamide crossover on OS. Methods: Patients with nmCRPC were randomised to oral darolutamide 600 mg twice daily (n = 955) or placebo (n = 554). Prespecified (rank-preserving structural failure time [RPSFT] and iterative parameter estimation [IPE]) and post hoc (OS-adjusted censoring and inverse probability of censoring weighting [IPCW], with weightings for baseline testosterone and prostate-specific antigen) sensitivity analyses were conducted. Results: After unblinding, 170 of 554 placebo recipients (30.7%) crossed over to darolutamide. At the final OS intention-to-treat analysis (median 11.2 months after unblinding), darolutamide significantly improved OS by 31% versus placebo (hazard ratio [HR] 0.69, 95% confidence interval [CI] 0.53–0.88; P = 0.003). The benefit increased in the analyses adjusting for crossover is as follows: RPSFT HR 0.68, 95% CI 0.51–0.90; P = 0.007; IPE HR 0.66, 95% CI 0.51–0.84; P < 0.001; OS-adjusted censoring HR 0.59, 95% CI 0.45–0.76; IPCW HR 0.63, 95% CI 0.48–0.81. The favourable safety profile of darolutamide was maintained, including in crossover patients. Conclusions: After adjusting for crossover, darolutamide reduced the risk of death by up to 41% in patients with nmCRPC. The effect of darolutamide on OS may have been underestimated in the original intention-to-treat analysis.