An inverse relationship between estimates of renal function, with formulas such as the Modification of diet in renal disease (MDRD) study equation or the Cockcroft-Gault formula, and mortality has been suggested. These formulas both contain the variables sex, serum creatinine and age and the latter also contains body weight. We investigated whether these formulas predict mortality better than the variables they contain together in patients with Type 2 diabetes.
In 1998, 1143 primary care patients with Type 2 diabetes participated in the Zwolle Outpatient Diabetes project Integrating Available Care (ZODIAC) Study, in the Netherlands. Clinical and laboratory data were collected at baseline. Life status was assessed after 6 years. We used Cox proportional hazard modelling to investigate the association between estimates of renal function (continuous data) and the variables they contain and mortality, adjusting for confounders. Both formulas were compared with models consisting of the variables present in the formulas. Predictability was assessed using Bayesian information criterion (BIC) and Harrell's C statistics.
At follow-up, 335 patients had died. All variables, except sex, influenced mortality. Predictive capability, indicated by lower BIC values and higher Harrell's C values, was up to 10% better for models containing the separate variables as compared with Cockcroft-Gault or MDRD.
Using estimates of renal function to assess mortality risk decreases predictability as compared with the combination of the risk factors they contain. These formulas, therefore, could be used to estimate renal function; however, they should not be used as a tool to predict mortality risk.