The original and modified Mallampati tests are commonly used to predict the difficult airway, but there is controversy regarding their accuracy. We searched MEDLINE and other databases for prospective studies of patients undergoing general anesthesia in which the results of a preoperative Mallampati test were compared with the subsequent rate of difficult airway (difficult laryngoscopy, difficult intubation, or difficult ventilation as reference tests). Forty-two studies enrolling 34,513 patients were included. The definitions of the reference tests varied widely. For predicting difficult laryngoscopy, both versions of the Mallampati test had good accuracy (area under the summary receiver operating characteristic (sROC) curve = 0.89 +/- 0.05 and 0.78 +/- 0.05, respectively). For predicting difficult intubation, the modified Mallampati test had good accuracy (area under the sROC curve = 0.83 +/- 0.03) whereas the original Mallampati test was poor (area under the sROC curve = 0.58 +/- 0.12). The Mallampati tests were poor at identifying difficult mask ventilation. Publication bias was not detected. Used alone, the Mallampati tests have limited accuracy for predicting the difficult airway and thus are not useful screening tests.
