RESUMO
We examined the potential for voice sounds to predict a difficult airway as compared with prediction by the modified Mallampati test. A total of 453 patients scheduled for elective surgery under general anaesthesia with tracheal intubation were studied. Five phonemes were recorded and their formants analysed. Difficult laryngoscopy was defined as the Cormack-Lehane grade 3 or 4. Univariate and multivariate logistic regression were used to examine the association between some variables (mouth opening, sternomental distance, modified Mallampati and formants) and difficult laryngoscopy. Difficult laryngoscopy was reported in 29/453 (6.4%) patients. Among five regression models evaluated, the model achieving better performance to predict difficult laryngoscopy, after a variable selection criteria (stepwise, multivariate) and included a modified Mallampati classification (OR 2.920; 95%CI 1.992-4.279; p < 0.001), first formant of /i/(iF1) (OR 1.003; 95%CI 1.002-1.04; p < 0.001), and second formant of /i/(iF2) (OR 0.998; 95%CI 0.997-0.998; p < 0.001). The receiver operating curve for a regression model that included both formants and Mallampati showed an area under curve of 0.918, higher than formants alone (area under curve 0.761) and modified Mallampati alone (area under curve 0.874). Voice presented a significant association with difficult laryngoscopy during general anaesthesia showing a 76.1% probability of correctly classifying a randomly selected patient. (AU)