Comparison of 2 methods for calculating adjusted survival curves from proportional hazards models

Context Adjusted survival curves are often presented in medical research ar ticles. The most commonly used method for calculating such curves is the me an of covariates method, in which average values of covariates are entered into a proportional hazards regression equation. Use of this method is wide spread despite published concerns regarding the validity of resulting curve s. Objective To compare the mean of covariates method to the less widely used corrected group prognosis method in an analysis evaluating survival in pati ents with and without diabetes. In the latter method, a survival curve is c alculated for each level of covariates, after which an average survival cur ve is calculated as a weighted average of the survival curves for each leve l of covariates. Design, Setting, and Patients Analysis of cohort study data from 11468 Albe rta residents undergoing cardiac catheterization between January 1, 1995, a nd December 31, 1996. Main Outcome Measures Crude and risk-adjusted survival for up to 3 years af ter cardiac catheterization in patients with vs without diabetes, analyzed by the mean of covariates method vs the corrected group prognosis method. Results According to the mean of covariates method, adjusted survival at 10 44 days was 94.1% and 94.9% for patients with and without diabetes, respect ively, with misleading adjusted survival curves that fell above the unadjus ted curves. With the corrected group prognosis method, the corresponding su rvival values were 91.3% and 92.4%, with curves that fell more appropriatel y between the unadjusted curves. Conclusions Misleading adjusted survival curves resulted from using the mea n of covariates method of analysis for our data. We recommend using the cor rected group prognosis method for calculating risk-adjusted curves.