In elderly individuals, serum creatinine may remain normal as glomerular filtration rate (gfr) declines. Therefore, the estimation of glomerular filtration utilizing mathematical models incorporates age as an important variable. In order to adjust drug dosages and diagnose renal disease earlier in the elderly, a variety of such simplified estimates of gfr have been applied. Unfortunately, no estimator is as accurate as the cumbersome gold standards (e.g. inulin or iothalamate clearance) and the reliability of each may vary with the particular clinical setting. The purpose of this study was to critically evaluate three commonly used estimators of gfr-i.e., creatinine clearance (CC), Cockroft-Gault (CG), and 100 over serum creatinine (100/SC)-comparing them to iothalamate clearance (IC) in a group of healthy ambulatory geriatric subjects (n = 41; ages 65-85). IC declined 1 ml/min per year of age in our sample. CC demonstrated a similar decline, a correlation of 0.83 with IC, and moderate error relative to IC of 17% at the mean (standard error [SE] = 12.3). In contrast, 100/SC correlated only 0.56 with IC, demonstrated a large positive bias (41 ml/min), and showed no age-related decline. An age correction to 100/SC similar to that utilized in the CG formula was clearly necessary. Despite the age and weight correction used in the CG formula, we found the estimates from it to be inaccurate (correlation = 0.5; SE = 23.8). A simpler age-corrected formula (Est. IC = 1/2 [100/SC] + 88-age) was derived and proved significantly superior to CG in our ambulatory geriatric sample, but still exhibited enough error (SE = 16.4) to question its clinical utility. It appears that serum creatinine based estimates of gfr in the elderly may not provide accurate results.