It has not yet been fully clarified whether the plasma or renal clearance approach is the most reliable to investigate GFR in humans. The study presented here aimed to compare plasma decay with renal clearance of 51Cr-EDTA in 27 diabetic patients with patterns of renal function broadly dispersed in a wide range of values. Moreover, the comparison was also performed with renal clearance of nonlabeled iothalamate in a subgroup of 17 patients. A biexponential function was found to fulfill statistical and heuristic criteria for the modeling analysis of plasma 51Cr-EDTA decay with 19 samples after bolus intravenous 51Cr-EDTA injection. Individual GFR values from 51Cr-EDTA plasma clearance highly correlated with those from renal clearance (r2 = 0.977, P < 0.0001), but resulted on average about 2.5 mL.min-1.1.73 m-2 higher (66.8 +/- 6.5 mL.min-1.1.73 m-2 (mean +/- SE) versus 64.3 +/- 6.4, P < 0.02). This difference remained relatively constant from patients with normal renal function to those with impaired renal function, suggesting that the plasma clearance is slightly less accurate than renal clearance approach because of a constant extrarenal clearance rate. In the subgroup studied, a similar difference was found between GFR values from 51Cr-EDTA plasma clearance (84.7 +/- 7.3) and renal clearance of iothalamate (82.8 +/- 7.3), although not statistically significant (P = 0.4). Individual GFR values well correlated (r2 = 0.913, P < 0.0001). The precision and reproducibility of the experimental approaches were assessed by comparing three coefficients of variation: (1) CVb of the bolus injection, because of measurement errors; (2) CVc of the continuous infusion, which additionally includes errors of urine volume measurement and physiological variability in the same day; and (3) CVr of repeated measurements by using bolus injection, which also accounts for physiological variability in different days. CVc of iothalamate and 51Cr-EDTA infusions were 7.5 +/- 1.9% and 7.4 +/- 1.2% respectively. CVb and CVr of bolus injection of 51Cr-EDTA were 2.6 +/- 0.3% and 3.5 +/- 0.8% respectively. CVb and CVr of bolus injection of 51Cr-EDTA, but not CVc of iothalamate and 51Cr-EDTA infusions were twofold to tenfold lower than the percent yearly change reported in IDDM and NIDDM patients. More particularly, CVr was significantly less than CVc. In order to make the test less cumbersome, a reduced sampling schedule with seven samples was designed and validated. GFR measured with seven samples was 66.1 +/- 6.4 (P = 0.1 when compared with the full 19-sample schedule) with a CVb of 3.5 +/- 0.5%. This seven-sample protocol was not different from that obtained with the previously described simplified method of Brøchner-Mortensen (63.9 +/- 6.8, P = 0.16), yet yielding a statistically more accurate estimate (coefficient of variation for Brøchner-Mortensen method = 12.1 +/- 2.9, P = 0.004). Moreover, only bolus injection, along with modeling analysis of plasma clearance rate, allows the accurate measurement of the extracellular fluid volume, an important parameter in diabetic patients. It was concluded that the reduced seven-plasma sample protocol is able to detect as small as 4 to 5% changes per year in a single patient. Moreover, it provides precise and accurate estimate of GFR in diabetic patients with hyperfiltration, who are postulated to be at higher risk to develop renal damage.