A prospective validation and comparison of three multivariate models for prediction of difficult intubation in adult patients

ABSTRACT

Abstract Purpose Several bedside clinical tests have been proposed to predict difficult tracheal intubation. Unfortunately, when used alone, these tests show less than ideal prediction performance. Some multivariate tests have been proposed considering that the combination of some criteria could lead to better prediction performance. The goal of our research was to compare three previously described multivariate models in a group of adult patients undergoing general anesthesia. Methods This study included 220 patients scheduled for elective surgery under general anesthesia. A standardized airway evaluation which included modified Mallampati class (MM), thyromental distance (TMD), mouth opening distance (MOD), head and neck movement (HNM), and jaw protrusion capacity was performed before anesthesia. Multivariate models described by El-Ganzouri et al., Naguib et al., and Langeron et al. were calculated using the airway data. After anesthesia induction, an anesthesiologist performed the laryngoscopic classification and tracheal intubation. The sensitivity, specificity, and receiver operating characteristic (ROC) curves of the models were calculated. Results The overall incidence of difficult laryngoscopic view (DLV) was 12.7%. The area under curve (AUC) for the Langeron, Naguib, and El-Ganzouri models were 0.834, 0.805, and 0.752, respectively, (Langeron > El-Ganzouri, p= 0.004; Langeron = Naguib, p= 0.278; Naguib = El-Ganzouri, p= 0.101). The sensitivities were 85.7%, 67.9%, and 35.7% for the Langeron, Naguib, and El-Ganzouri models, respectively. Conclusion The Langeron model had higher overall prediction performance than that of the El-Ganzouri model. Additionally, the Langeron score had higher sensitivity than the Naguib and El-Ganzouri scores, and therefore yielded a lower incidence of false negatives.