Background: Cystatin C is increasingly used to estimate renal function, but its large intraindividual variability limits its practical value. This study aimed at verifying whether the clinical practice of associating cystatin C determination with serum creatinine (Cr) improved the ability of the Cockcroft and Gault formula to estimate creatinine clearance (CrCl).
Methods: It was an observational cross-sectional study of 134 in-patients with mildly impaired renal function. Using the Hoek et al formula (glomerular filtration rate (GFR)/1.73m2 = - 4.32 + 80.35/cystatin C mg/L), multivariate linear regressions (LREG) and artificial neural networks (ANN), we integrated cystatin C in the Cockcroft and Gault formula and analyzed the potential superiority of this procedure by comparing its performance with that of the two algorithms taken separately.
Results: The inclusion of cystatin C in the Cockcroft and Gault formula using the data of an LREG (CrCl = 0.371 x (Hoek et al) + 0.589 x Cockcroft and Gault), a simple mean between the two algorithms or ANN ameliorated the CrCl estimation precision allowing an absolute error reduction of approximately 4, 4 and 6%, respectively (relative values 12, 12 and 17%).
Conclusions: Although the combination of the Hoek et al and Cockcroft and Gault formulae using both linear and non-linear mathematical methods allowed a statistically significant reduction in the estimation error generated by Cockcroft and Gault, considering the small impact on the estimation precision and the large intraindividual variation of both cystatin C and Cr, this procedure probably has no clinical relevance.