Diagnostic value of the Friedman tongue position and Mallampati classification for obstructive sleep apnea: a meta-analysis.

Objective: To assess the association between the Mallampati classification and Friedman tongue position for obstructive sleep apnea severity as determined by apnea-hypopnea index and to determine which method is most closely correlated with prediction of obstructive sleep apnea severity.

Data Sources: English-language searches of PubMed, MedLine, and the Cochrane database. Reference sections of identified studies were examined for additional articles.

Review Methods: Databases through December 2011 were searched, combined with review of relevant article bibliographies, and assessed by 4 reviewers. Systematic review and random-effects meta-analysis of studies evaluating tongue position and obstructive sleep apnea severity were performed. Outcomes were reported as correlations.

Results: Ten studies met inclusion criteria and had data for pooling (2513 patients). Friedman tongue position and Mallampati classification were significantly associated with obstructive sleep apnea severity, with a correlation of 0.351 (0.094-0.564, P = .008). Analysis of the correlation of tongue position with obstructive sleep apnea severity reveals correlations of 0.184 (0.052, 0.310, P = .006) and 0.388 (0.049, 0.646, P = .026) for the Mallampati classification and Friedman tongue position, respectively. Publication bias does not yield a significant Egger regression intercept; however, 4 imputed values to the right of the mean were found using Duval and Tweedie's trim-and-fill method, yielding an overall correlation of 0.498 (confidence interval = 0.474-0.521).

Conclusion: The Mallampati classification and Friedman tongue position assessment techniques are significantly correlated with predicting obstructive sleep apnea severity. Publication bias does not significantly affect our results. The strength of this correlation is higher for Friedman tongue position, although 95% confidence intervals for the respective correlation coefficients overlap.