We have investigated the influence on reproducibility of total [51Cr]EDTA plasma clearance (E) of various times and number of blood samples in patients with normal (13 patients) and low (14 patients) renal function. The study aims at fixing a clinically useful procedure suitable for all levels of renal function. Six different types of E were evaluated with time periods for blood sampling between 3 and 5h after tracer injection, and the variation from counting radioactivity, sc, was determined as part of total variation, st. Optimum mean time t(E), for blood sampling was calculated as a function of E, as the mean time giving the least change in E for a given change in the 'final slope' of the plasma curve. For patients with normal E, sc did not contribute significantly to ST, and t(E) was about 2h. For patients with low renal function sc contributed significantly to st, and t(E) increased steeply with decreasing E. The relative error of sc from fixed E types was calculated for all levels of renal function. The results indicate that blood sampling individualized according to predicted E values is not necessary. A sufficient precision of E can be achievef for all function levels from three blood samples drawn 180, 240, and 300 min after injection.