This article updates the previous review in the Journal regarding theophylline dosing methods. Among the predictive algorithms evaluated, the dose-titration scheme of Weinberger and Hendeles was extensively tested in 1073 asthmatic patients. When the scheme was followed appropriately, 78% of initial serum concentrations and 66% of repeat serum concentrations were within the therapeutic range of 10 to 20 mg/L. Several authors have also demonstrated that the 'condition correction factor' method for estimating theophylline clearance is of limited value. The individualised methods of Chiou, Koup and Vozeh have been evaluated in over 300 patients. In addition, numerous authors have reported the relative predictive performance of Bayesian dosing programs for theophylline. All methods continue to be a rapid and accurate means of individualizing dosing requirements for patients with a diverse range of theophylline disposition characteristics. Overall, the Bayesian predictions have been less biased and slightly more precise than the pharmacokinetics-based dosing methods. The most recent cost-effectiveness data has shown that a pharmacokinetic dosing program resulted in fewer toxic serum concentrations (18.9% vs 37.8%), a shorter mean duration of hospital stay (6.3 vs 8.7 days) and more therapeutic concentrations with subsequent oral therapy (71.1% vs 44.4%) than among control patients receiving dosages prescribed by physicians.