Previous Abstract | Next Abstract
Printable Version
October 22, 2016
10/22/2016 1:00:00 PM - 10/22/2016 3:00:00 PM
Room Hall F Foyer-Area E
A Prospective Evaluation of Three Multivariate Models for Prediction of Difficult Tracheal Intubation
Gustavo P. Bicalho, Ph.D., Roberto C. Bessa, M.D., Fabiano S. Carneiro, M.D., Marcos C. Cruvinel, M.D., Carlos V. Castro, M.D.
Lifecenter Hospital, Belo Horizonte, Brazil
Disclosures:  G.P. Bicalho: None. R.C. Bessa: None. F.S. Carneiro: None. M.C. Cruvinel: None. C.V. Castro: None.

Airway management problems are still an important cause of anesthesia morbidity and mortality. Single airway prediction tests have been introduced in clinical practice. Unfortunately, these tests frequently fail to correctly discriminate between patients with difficult or easy airways. In order to improve the discrimination power, some multivariate tests have been proposed. The rationale is that the combination of different parameters could lead to better prediction performance.

The Wilson risk score ranges from 0 to 10 and includes six criteria: body weight (BW), head and neck movement (HNM), jaw protrusion (JP), interincisor gap (IG) and the presence of buck teeth or receding mandible.

The El-Ganzouri risk score ranges from 0 to 12 and includes: IG, thyromental distance (TD), modified Mallampati class (MM), HNM, JP, BW and history of difficult intubation.

The Naguib method uses a mathematical formula as follows: l= 0.2262 - 0.4621 x TD + 2.5516 x MM - 1.1461 x IG + 0.0433 x height, in which TD, IG and height is in centimeters and MM value is considered 0 (class I or II) or 1 (class 3 and 4).

The purpose of this research was to prospectively evaluate these three multivariate models in the same patient population in order to identify the one with best prediction performance for difficult intubation during anesthesia.


After Ethics Committee approval, one hundred adults scheduled for surgery requiring general anesthesia with tracheal intubation were studied. An anesthesiologist performed the airway evaluations and then the airway scores were calculated. The values considered as a positive result for difficult intubation were: Wilson score ≥ 2, El-Ganzouri score ≥ 4 and Naguib formula result ≥ 0.

Anesthesia was induced with fentanyl, propofol and cisatracurium. All patients were monitored with a peripheral nerve stimulator. An anesthesiologist, not involved in the airway evaluation, performed the laryngoscopy after the loss of the last muscle twitch in train-of-four stimulation. The laryngoscopy was performed with an English Macintosh blade with the head in full extension and with no external laryngeal manipulation. The laryngeal view was graded according to the Cormack-Lehane classification. Grades 1 or 2 were considered as easy laryngeal view (ELV) and 3 or 4 as difficult laryngeal view (DLV).

A univariate analysis was performed for each parameter to assess an association with DLV. Values of sensitivity, specificity, positive and negative predictive values were calculated for each method. Receiver operator characteristic (ROC) curves were plotted and the area under curve (AUC) were also calculated.


Thirteen patients presented with DLV (13%). The univariate analysis showed an association of a lower IG (4.2 vs. 4.9cm, p < 0.001) and higher MM class (III and IV vs. I and II, p = 0.017) with DLV. The models’ diagnostic profiles are shown in Table 1. The models’ ROC curves are shown in Figure 1. The Naguib, El-Ganzouri and Wilson models AUC were 0.798, 0.749 and 0.688, respectively.


The Naguib model showed the highest overall prediction performance for difficult laryngoscopy with the largest AUC on ROC curve and also the highest sensitivity among the methods studied.

Figure 1
Figure 2

Copyright © 2016 American Society of Anesthesiologists