Investigations using logistic regression models on the effect of the LMA on morphine induced vomiting after tonsillectomy

The effect of intraoperative airway management on postoperative vomiting after tonsillectomy is unknown. Logistic regression was used in a retrospective study to investigate the effect of the laryngeal mask airway (LMA) on a morphine dose-vomiting response curve. Charts were reviewed in 351 children in whom the airway was managed with either a LMA (n=177) or a tracheal tube (n=174). A mean perioperative morphine dose of 0.10 mg.kg(-1) (SD 0.09) was used in 248 children and a further 103 children were given no opioid. One hundred and eighteen of these 248 children vomited (47.6%) compared to 14 of 103 children given no morphine (13.6%). The probability of vomiting was related to morphine dose using logistic regression with both a linear and an E(max) model. Both the calibration (Hosmer-Lemishow goodness of fit chi-squared test lambda(2), P=0.81) and discrimination (area under the receiver operating characteristic plot, AUC ROC=0.67) of the E(max) model were better than the linear model (lambda(2), P=0.49; AUC ROC=0.64). Pharmacodynamic parameter estimates for the Emax model were P(0) (the baseline probability of vomiting) 0.139, P(max) (the maximal probability of vomiting due to morphine) 0.96, ED(50) (morphine dose that induces an effect equivalent to 50% of the logit P(max)) 0.09 mg.kg-1. The probability of vomiting was 50% after morphine 0.125 mg.kg-1. The use of the LMA had no effect on this dose-response curve. A covariate analysis investigating propofol for induction or isoflurane for the intraoperative maintenance of anaesthesia, however, showed that both drugs shifted the curve to the right. The probability of vomiting was 50% after morphine 0.17 mg.kg(-1) and 0.21 mg.kg(-1) for the isoflurane and propofol use curves, respectively. The concomitant use of propofol and isoflurane, but not the use of the LMA, decreases the probability of vomiting due to morphine.