We studied the kinetic behavior of the reaction of alkaline picrate and creatinine and evaluated a nonlinear curve-fitting method for quantifying creatinine in serum. Using a 3 X 3 factorial experimental design, we evaluated interactive effects among temperature and concentrations of creatinine, picrate, and NaOH. We found no evidence of interference by glucose or unconjugated bilirubin; the effects of the acetoacetate reaction, which is fast, are easily compensated by the curve-fitting method. The reaction with human serum albumin is very complex, but its effects are compensated by the curve-fitting method and by preparing standards containing 50 g of albumin per liter. Calibration plots are linear under a wide variety of conditions for both aqueous standards and standard additions of creatinine to pooled serum. Reproducibility studies with standards containing creatinine at 2, 10, and 20 mg/L yielded relative standard deviations (RSD) of 8.2, 2.5, and 1.3%, corresponding to absolute variations of 0.16, 0.25, and 0.26 mg/L. The average SD for 17 sera containing creatinine at 15-50 mg/L was 0.7 mg/L. The averages of ratios (as percent) of determined vs expected concentrations in 17 sera with added creatinine (7.27 mg/L) were 97.8% for aqueous standards, 99.9% for standards with added albumin.