Evaluation of GFR estimating equations in the general community: implications for screening

The Kidney Disease Outcomes Quality Initiative has recommended the use of GFR estimating equations to detect silent chronic kidney disease (CKD) in the community. The benefit of general reporting of CKD must be balanced with the harm of mislabeling people who do not have CKD. The popular Cockcroft-Gault (CG) and Modification of Diet in Renal Disease (MDRD) GFR estimating equations were compared with the recently devised Rule equation in a representative community population sample (2166) divided into subsamples with (385) and without (1781) previous renal impairment. The prevalence of CKD was CG > MDRD >> Rule estimates. The magnitude of difference in prevalence of CKD as detected by the MDRD and CG versus the Rule equation increases markedly when the subsamples with (30.8 and 29.7 versus 17.5%) and without (12 and 11.3 versus 3.0%) previous kidney impairment are compared. General demographic and potential or known risk factors were used in a logistic regression model to assess the association with CKD. The MDRD estimates note female gender (odds ratio 2.19; 95% confidence interval 1.63 to 2.95) and both MDRD and the Rule equations identify hypertension and diabetes as significant CKD risk factors. All estimating equations identify age to be associated with CKD. The annualized serial decline in GFR was CG > MDRD > Rule estimates. Only the Rule GFR estimates detected a greater decline in renal impaired versus unimpaired populations. The calibrated Rule equation seems to perform better than CG and MDRD (CKD 3 versus 11.3 to 12%) but lacks validation against gold standards for community-based screening.