Previously published midazolam-alfentanil response surface model cannot predict patient response well in gastrointestinal endoscopy sedation

Background: A response surface model is a mathematical model used to predict multiple-drug pharmacodynamic interactions. With the use of a previously published volunteer model, we tested the accuracy of the midazolam-alfentanil response surface model during gastrointestinal endoscopy. Methods: We enrolled 35 adult patients scheduled for combined endoscopic procedures. Patients were sedated with intravenous midazolam and alfentanil, and monitored with real-time auditory evoked potential. Sedation Observer's Assessment of Alertness/Sedation (OAA/S) scores were recorded by an independent observer every 2 minutes. Patients with OAA/S scores of ≥4 were designated as "awake". Pharmacokinetic profiles were calculated using the TIVA trainer. The published response surface model was modified to make estimations more reasonable. Patient response (OAA/S score ≥ 4 or <4) was then estimated using the modified version of the model. Results: The average procedural times were 3.3 ± 2 minutes and 6.5 ± 2.3 minutes for esophagogastroduodenoscopy and colonoscopy, respectively. The model poorly predicted patient response during gastrointestinal endoscopic procedure sedation. Accuracy in predicting an OAA/S score of <4 was 6% for the original model and 0% for the modified model. The estimated probability of loss of response ranged from 0.04% to 2.94% at the time of arousal (OAA/S score ≥ 4) and from 0.24% to 15.55% when the patient was asleep (OAA/S score < 4). Conclusion: The model showed significant synergy between midazolam and alfentanil; however, it was inadequate in predicting the response of patients undergoing sedated gastrointestinal endoscopic procedures. Future model parameter adjustments are required.