The relationship between the administered dose d of Gd-DTPA and the accuracy of measurements of the glomerular filtration rate G and the cardiac output O was investigated. For a wide range of values the concentration of Gd-DTPA can be uniquely derived from MR signals and precontrast longitudinal relaxation time. Fixed and random errors in these measured variables were analyzed. Depending on noise level and the level of renal function, errors in G reach a minimum for d = 1.4-2.8 mmol. Random errors in G are relatively insensitive to d as long as d > 1.5 mmol. These results establish the feasibility of dynamic MR measurements using doses of Gd-DTPA that are several times lower than current standards.