Several whole-body variables are available for indexing glomerular filtration rate (GFR), including extracellular fluid volume (ECF volume, ECV), which has the advantage that GFR based only on the terminal exponential, alpha 2, of the chromium-51 ethylene diamine tetraacetic acid (51Cr-EDTA) plasma clearance curve is a measure already indexed for ECV. This slope-only method, however, is open to the theoretical criticism of a lack of robustness not present in the more conventional slope/intercept method, which generates GFR in absolute units before indexation to body surface area (BSA). To further evaluate the slope-only method, a large database of routine three-sample GFR measurements, based on 51Cr-EDTA, was analysed with the aim of comparing GFR/BSA with GFR/ECV, identifying the main sources of error in their respective measurements and explaining why they might be discrepant. The database was subdivided into 304 patients (group A) in whom the correlation coefficient of the fit to the three data points was greater than 0.99 and 31 patients in whom it was less than 0.99 (noisy data; group B). There was modest agreement between GFR/BSA and GFR/ECV in group A (r = 0.77; mean ratio 1.01 +/- 0.21). The difference between them correlated significantly with BSA (r = 0.52; P < 0.001), as would be expected given that small individuals have a relatively high BSA. Subdividing group A into four subgroups stratified according to BSA led to a slight improvement in the correlation between GFR/BSA and GFR/ECV. Because an error in alpha 2 is balanced by an opposing error in the intercept (which leads to a change in distribution volume in the opposite direction), and therefore impacts on both GFR/BSA and GFR/ECV, whereas an error in administered dose affects only GFR/BSA, it is possible to evaluate such errors by examining the respective relationships of GFR/BSA and GFR/ECV with indexed ECV. In group A, GFR/BSA correlated positively with ECV/1.73 m2, not surprisingly as ECV/BSA is effectively the ratio GFR/BSA to GFR/ECV, while GFR/ECV correlated negatively with ECV/1.73 m2. This implies that errors other than alpha 2, probably principally in administered dose, were as important as errors in alpha 2. This conclusion was supported by disappearance of the positive correlation between GFR/BSA and ECV/BSA in group B, brought about by the greater errors in alpha 2 in this group. This study suggests that, because of the effects of errors in the slope/intercept method that do not affect alpha 2, GFR based on slope only is at least as robust as that based on slope/intercept.