The relationships between serum and saliva concentrations were evaluated in 18 patients after single doses of three theophylline preparations. Simultaneous multiple saliva and serum samples were obtained after theophylline ingestion and were analyzed for theophylline content by the modified Schack and Waxier spectrophotometric method. Significant variability was observed in the data for those saliva/serum ratio sets obtained within 30 min of dosage. The mean correlation coefficient for saliva/serum sample sets obtained at and after 30 min was 0.86. Pearson product moment correlation coefficients determined for saliva/serum ratios at 2,4,6, and 8 hr ranged from 0.84 to 0.89, leaving 26% of the variance unexplained. Serum concentrations were calculated from a randomly selected saliva/serum ratio for each patient. Predicted serum concentrations were ± 2 μg/ml of measured concentrations in only 143 of the 196 determinations (73%). Therefore, the use of saliva concentrations to predict serum values has substantial error in specific patients and may lead to incorrect dosage adjustments.