The use of saliva as an indirect, non-invasive method of theophylline plasma level measurement was evaluated in 23 older men with chronic obstructive pulmonary disease and numerous concurrent medical problems. Simultaneously collected plasma and saliva samples were obtained on two or more occasions and analyzed for theophylline concentration by the Schack and Waxler spectrophotometric method. The mean (+/- SD) plasma:saliva ratio for 84 sample sets was 1.6 (+/- 0.5) with the saliva concentration averaging 64.8% (+/- 20.8%) of the plasma concentration. Multiplying a randomly chosen plasma:saliva ratio from each patient by the saliva concentration of a second randomly selected observation resulted in a predicted plasma concentration that differed by more than 20% of the measured concentrations for 16 of the 23 patients. Thus, use of saliva theophylline measurements obtained by the Schack and Waxler method to adjust dosage regimens cannot be recommended in these patients.