The accuracy in the dosimetry for radionuclide therapy shows a great contrast to that obtained in external beam radiotherapy. The dosimetry for [(90)Y-DOTA(0), Tyr(3)] octreotide is evaluated in patients to see whether the accuracy of the dosimetry is high enough to distinguish the probability for radiation nephropathy. The 5% threshold for late end-point nephropathy at 23 Gy with external beam radiotherapy becomes with (90)Y therapy at least 30-35 Gy, when it is given in three or more fractions. More accurate linear-quadratic (LQ) model parameters are, however, needed to predict a more precise threshold for renal damage in this dose rate region. The average MIRD-based dose to the kidneys was 27 +/- 4 Gy (N = 52) with no evidence for renal damage. The variance in the dose is only caused by the high variability in renal uptake kinetics of the compound. Using the actual kidney volumes instead of the phantom values lowered the kidney dose considerably, but the variance in the dose greatly increased. As the peptide specifically localizes in the kidney cortex, the dose to the cortex increased up to a factor 1.5 compared to the MIRD whole kidney dose. Both the sum of uncertainties of ? 40% in the actual dose to the kidneys and the unknown maximum tolerable kidney dose for internal therapy make that a fixed injected activity of 13.32 GBq together with a patient-averaged dosimetry is as good as patient-kinetics specific dosimetry using the MIRD method.