Recently, we described a basic model and its more elaborate variants to predict the uptake and distribution of inhalational anesthetics during closed-circuit anesthesia. As an initial clinical validation of the linear, continuous, 14-compartment basic model, the current study examined its predictive performance in 50 patients by comparing quantitatively the predicted and the measured alveolar concentration-time profiles after bolus injections of liquid isoflurane into the closed system during mechanical ventilation. The two versions of the model studied differed in the size of their peripheral shunt, as 0% (version A) and 16% (version B) of the cardiac output. A total of 15,744 alveolar concentrations of isoflurane (one per 10s period) were measured by mass spectrometry. For each measured concentration we used computer simulations of version A and version B to calculate a predicted concentration for both versions. For each patient we calculated the bias (indicating over- or underprediction) and the scatter of the prediction errors (indicating the typical error size). The bias and the scatter of the prediction errors, both given as mean (and standard deviation), were 2.25 (13.59) and 12.51 (5.84)% for version A and 12.00 (14.97) and 14.12 (6.54)% for B. Version A performed better than B: both the bias (P = 0.008) and the scatter (P less than 0.0001) were closer to zero for A. Logistic regression analysis showed for version A that scatter, but not bias, increased with age (P = 0.002). Gender, body mass index (weight x height-2), and number of injections per hour did not influence scatter or bias.(ABSTRACT TRUNCATED AT 250 WORDS)