The most common analysis of time to event data uses the proportional hazards (PH) model. The proportional hazard means that the ratio of the hazard rates for treatment versus control is constant over time. If the PH assumption does not hold, the interpretation of the hazard ratio (HR) can be very difficult. The basic concern is that a clinically important treatment effect may be more complex than that implied by a single HR between treatment arms. It may not be appropriate to interpret the hazard ratio as a relative risk. The HR becomes time-dependent and the estimated HR is some type of complex average over the event time. For example, survival curves that diverge then converge or even cross may suggest the mode of action of a research treatment. A simple HR would not be sufficient to interpreting such data. Moreover, the regression coefficients of the covariates including treatment may be biased as a result. In this presentation, we will compare the HR with the restricted mean survival time analysis for several studies of cardiorenal drugs and discuss the differences between the two methods.