Bayesian and threshold probabilities in therapeutic drug monitoring: when can serum drug concentrations alter clinical decisions?

The use of Bayesian and threshold probabilities is examined with respect to the range of probabilities of toxicity for which obtaining a patient's serum drug concentration leads to information that is potentially useful in altering a clinical decision. For the situation of potential drug-induced toxicity, three threshold probabilities are needed to characterize the decision process: the decision threshold for deciding between continuing and discontinuing the drug regimen, the testing threshold that separates the decision to continue the regimen without testing the serum drug concentration from the decision to test before deciding, and the companion testing threshold that separates the decision to discontinue the regimen without testing from the decision to test before deciding. For digoxin, theophylline, aminoglycosides, vancomycin, and phenytoin, three prototypical decision threshold probabilities, 0.33, 0.2, and 0.1, are used, along with published true-positive and false-positive rates, to calculate serum concentration testing thresholds for each drug. Practitioners can be more effective in their use of serum drug concentration data when a Bayesian approach to probability assessment is used in conjunction with testing thresholds.