We are gratified that our article has generated a healthy discussion. It was not our intent to initiate a debate about "whose GFR equation is best." All of the available equations, including ours, have limitations. Instead, we believe we should concentrate our efforts on whether a generalizable equation based on serum creatinine can be developed (this may not be possible). We also believe that it is important to be cautious when interpreting the prevalence of a reduced estimated GFR in various populations.

The discrepancy between our results and those of Froissart and colleagues may be related to differences in methods. In our study, we compared patients with chronic kidney disease who had an estimated (not measured) GFR greater than 60 mL/min per 1.73 m² with healthy persons. Since the objective was to compare the accuracy of equations in predicting measured GFR, we did not identify or stratify the chronic kidney disease and healthy samples by measured GFR. The quadratic GFR equation (derived by using two-thirds healthy persons and one-third patients with chronic kidney disease) should be tested in populations where the diagnosis of chronic kidney disease is unknown, such as the general population. Even with careful attention to creatinine assay calibration, other investigators showed that at a GFR of 60 mL/min per 1.73 m², the MDRD equation underestimated measured GFR by 30 mL/min per 1.73 m² among healthy persons (1).

We agree with Delanaye and colleagues that the relationship between serum creatinine concentration and GFR differs among many populations and clinical presentations (for example, good health vs. chronic kidney disease). In our equation, the cutoff of 0.8 mg/dL (71 µmol/L) for serum creatinine concentration was determined as the peak of the parabola; otherwise, a further decrease in creatinine concentration would lead to a decrease in estimated GFR. The version of the Cockcroft–Gault equation that we had referenced predicted GFR and was derived from patients with chronic kidney disease (2). Despite its limitations, if GFR was not indexed to body size in some way, smaller persons would be more likely to have a reduced GFR and thus chronic kidney disease. A similar argument exists between cardiac output and cardiac index.

The letter by Maaravi and colleagues nicely illustrates a point. An equation derived by using patients with chronic kidney disease (the MDRD equation) increased the prevalence of reduced GFR and weakened epidemiologic associations between GFR and risk factors. However, an equation derived by using healthy persons and patients with chronic kidney disease (the quadratic equation) decreased the prevalence of a reduced GFR and strengthened epidemiologic associations. To better understand the epidemiology of early chronic kidney disease, we need studies that measure rather than estimate GFR in populations where the diagnosis of chronic kidney disease is unknown.