A critical reanalysis of the DAPA-HF and DELIVER trials reveals that the survival proportional odds model provides greater statistical power and more robust treatment effect estimates than traditional Cox regression in heart failure populations characterized by high risk heterogeneity.