Since patients with chronic liver disease present with greatly varying theophylline disposal reductions, in cirrhotics the acute theophylline dose schedule must be guided by an index of liver function that predicts theophylline pharmacokinetics. We therefore studied 26 patients with severe chronic liver disease to ascertain the efficacy of the routinely used clinical and biochemical liver function tests in predicting theophylline pharmacokinetics. The prealbumin plasma level, recently proposed as a valuable index of liver function, was also considered. With respect to 10 controls, theophylline clearance was found to be significantly reduced (30 +/- 2 vs. 75 +/- 11 ml/kg/h, mean +/- SD, p less than 0.01). However, only 7 patients had a reduction great enough to require a reduced intravenous theophylline dose schedule. An analysis of clinical utility, made on the basis of ROC curves, showed that the albumin/globulin ratio was the most effective index for identifying patients requiring lower doses of theophylline. Prealbumin and albumin were also useful, whereas bilirubin, prothrombin time, pseudocholinesterase, the presence of ascites and Pugh-Child's classification of the severity of liver disease were found to be worthless.