Discrepancies in the RIFLE classification are due to the method used to assess the level of derangement of kidney function

Abstract

Purpose: We hypothesized that RIFLE based on creatinine clearance (CrCl) is superior to that based on serum creatinine (sCr) or Cockroft-Gault (C-G) because it is an earlier marker of kidney dysfunction.

Materials and methods: At day 3 of admission, we compared the RIFLE based on sCr, C-G, and CrCl with 28-day mortality and development of RIFLE-F during intensive care unit stay.

Results: Percentages in the RIFLE levels were similar for the 3 estimates, but the patients included in each level were different; with CrCl as the reference, κ statistic was 0.29 (95% confidence interval, 0.15-0.43) for sCr and 0.21 (0.07-0.36) for C-G. Mortality at day 28 was 19.3%, with percentages of mortality increasing with RIFLE based in CrCl but not sCr or C-G (area under the curve, 0.57 [45-72] for C-G; 0.57 [44-72] for sCr; and 0.64 [52-79] for CrCl). Logistic regression only showed an independent relationship with mortality for RIFLE measured with CrCl.

Conclusions: RIFLE classification using sCr or C-G at the third day of admission predicts outcome less accurately than with the use of CrCl. Because of the delay in the rise of sCr after a sudden glomerular filtration rate decrease, RIFLE based in CrCl may represent an advantage in terms of precocity