ABSTRACT
BACKGROUND: The non-linear mixed amount with zero amounts response surface model can be used to describe drug interactions and predict loss of response to noxious stimuli and respiratory depression. We aimed to determine whether this response surface model could be used to model sedation with the triple drug combination of midazolam, alfentanil and propofol. METHODS: Sedation was monitored in 56 patients undergoing gastrointestinal endoscopy (modelling group) using modified alertness/sedation scores. A total of 227 combinations of effect-site concentrations were derived from pharmacokinetic models. Accuracy and the area under the receiver operating characteristic curve were calculated. Accuracy was defined as an absolute difference <0.5 between the binary patient responses and the predicted probability of loss of responsiveness. Validation was performed with a separate group (validation group) of 47 patients. RESULTS: Effect-site concentration ranged from 0 to 108 ng ml-1 for midazolam, 0-156 ng ml-1 for alfentanil, and 0-2.6 µg ml-1 for propofol in both groups. Synergy was strongest with midazolam and alfentanil (24.3% decrease in U50, concentration for half maximal drug effect). Adding propofol, a third drug, offered little additional synergy (25.8% decrease in U50). Two patients (3%) experienced respiratory depression. Model accuracy was 83% and 76%, area under the curve was 0.87 and 0.80 for the modelling and validation group, respectively. CONCLUSION: The non-linear mixed amount with zero amounts triple interaction response surface model predicts patient sedation responses during endoscopy with combinations of midazolam, alfentanil, or propofol that fall within clinical use. Our model also suggests a safety margin of alfentanil fraction <0.12 that avoids respiratory depression after loss of responsiveness.