Pharmacokinetic (PK) data from 34 cancer patients receiving 2 to 10 micrograms of calcitriol subcutaneously (s.c.) were used to develop a limited sampling method for predicting serum calcitriol area under curve (AUC) based on three samples instead of the full complement of 12 to 16 samples. Serum calcitriol levels were measured by 1, 25-dihydroxyvitamin D3-[I125] radioimmunoassay. Individual patient-corrected serum calcitriol AUC0-12 h was calculated by the trapezoidal rule after subtracting the pretreatment serum calcitriol level. PK data were split into "training" and "evaluation" sets based on calcitriol dose and chronological order of enrollment. Linear regression models of log-corrected AUC0-12 h versus individual log calcitriol serum levels in the hour 1, 2, 3, 4, 6, and 8 samples were established using the training data set of 17 patients. The fit was tested on the evaluation data set of 17 patients using mean squared error (MSE) as the fit criterion. The best single time point predictor of log AUC0-12 h was the log serum (calcitriol) at hour 6 (MSE = 0.0061). The best prediction of log AUC0-12 h using two time points was found to involve hour 6 and hour 2 (MSE = 0.0018). The prediction equation for the latter model was as follows: Log AUC = 1.125 + 0.3756.log (calcitriol) at hour 2 + 0.5859.log (calcitriol) at hour 6. This limited sampling method was further evaluated in 83 cancer patients treated with 4 to 38 micrograms of oral (p.o.) calcitriol; observed and predicted calcitriol AUC0-12 h were highly correlated (r > or = 0.90, p = 0.0001). These results show that serum calcitriol AUC0-12 h after s.c. and p.o. calcitriol administration is accurately estimated using pretreatment and 2- and 6-hour blood samples.