Estimated Glomerular Filtration Rate: Fit for What Purpose?

Abstract

There is important nosologic utility in staging chronic kidney disease (CKD) based on estimates of glomerular filtration rate (GFR). These equations have been optimized for estimating GFR at a single point in time. Risk assessment models used for prognosis of specified outcome events have commonly incorporated estimated GFR (eGFR), but the validity of this approach has not been evaluated. The current objective is to evaluate the risk of all-cause mortality over a 10-year follow-up period with multivariable-adjusted Cox regression analysis, comparing CKD stages based on eGFR to Cockcroft-Gault estimated creatinine clearance (eCrCl). There were significant differences between Stage 3A and Stage 3B-5 hazard ratios for all-cause mortality (p = 0.003) using eCrCl categories, but not for the same eGFR categories (p = 0.241). Discrimination analysis showed that a clinically significant difference (relative integrated discrimination improvement 778.6%; p = 0.001) was observed between the 2 models for the age strata ≤64. While eGFR is more precise and accurate than the Cockcroft Gault equation for estimating measured GFR at a single point in time, eGFR does not perform as well as eCrCl for assessing risk of all-cause mortality over 10-year follow-up intervals.