Although it is well known that the tolerance of the liver to external beam irradiation depends on the volume of liver irradiated, few data exist which quantify this dependence. Therefore, a review was carried out of our clinical trial for the treatment of intrahepatic malignancies in which the dose of radiation delivered depended on the volume of normal liver treated. Three dimensional treatment planning using dose-volume histogram analysis of the normal liver was used for all patients. Nine of the 79 patients treated developed clinical radiation hepatitis. None of the patient related variables assessed were associated with radiation hepatitis. All patients who developed radiation hepatitis received whole liver irradiation, as all or part of their treatment, which produced a mean dose greater than or equal to 37 Gy. Dose volume histograms were used to calculate normal tissue complication probabilities based on parameters derived from the literature. The risk of complication was greatly overestimated among patients receiving a high dose of radiation to part of the liver without whole liver treatment. An estimation of model parameters based on the clinical results indicated a larger magnitude for the "volume effect parameter" than the literature estimate (n = 0.69 +/- 0.05 vs 0.32; p less than 0.001). Computation of the normal tissue complication probabilities using the larger value of n produced a good description of the observed risk of radiation hepatitis. These findings suggest that dose volume histogram analysis can be used to quantify the tolerance of the liver to radiation. The predictive value of this parameterization of the normal tissue complication probability model will need to be tested with liver tolerance and dose volume histogram data from an independent clinical trial.