A comparison of three computer models for prediction of dose in acute amitriptyline overdose

The pharmacokinetics of amitriptyline in overdose have been reported not to fit conventional compartmental models. In this study, the dose-concentration-time relationships of amitriptyline in overdose were modeled with discriminant analysis, with an evolutionary heuristic search program, and with a decision-tree model based on the entropy of uncertainty of classification. The computer models all used the same data from dogs administered treatment (80 mg/kg), toxic (250 mg/kg), or fatal (500 mg/kg) doses directly into the surgically isolated duodenum. All the models achieved a high degree of success (77 to 93%) in assigning records to the high-, low-, or middle-dose groups. Two of the models gave a probability of the assignment. Results of this analysis suggest that blood amitriptyline and nortriptyline concentrations are most useful in estimating dose in acute amitriptyline overdose.