Obesity and Laryngopharyngeal Reflux as Risk Factors of Lingual Tonsil Hypertrophy in Patients With Sleep-Disordered Breathing: Validity of the Statistical Approach

Abstract

To the Editor I read with interest the study by Sung et al 1 for97 patients with sleep-disordered breathing. They concluded that laryngopharyngeal reflux and body mass index were independent risk factors of lingual tonsil hypertrophy (LTH). I present some statistical concerns about their study. First, they described that multiple stepwise regression analyses were performed for the risk assessment. But they presented significant and insignificant factors simultaneously in their Table 3. I suppose that they used forced entry method instead of stepwise selection method.Second, nonnormal distribution of independent variables in multiple regression analysis is recommended to be converted to normal distribution to keep the residuals being normally distributed. Because the apnea-hypopnea index (AHI) presents nonnormal distribution in general, their conclusion that there is no association between respiratory parameters and LTH should be handled with caution. Please consider the transformation of the AHI as an independent variable of multiple regression analysis.Third, the adjusted R 2 value (the coefficient of determination) was 0.161. This means that under one-sixth of the total variance of LTH could be explained by their statistical model. Fourth, 97 patients were included in the multiple regression analysis, and this model requires at least 10 individuals per independent variable to show appropriate estimates.2 Sung et al selected 9 independent variables for their analysis, and more patients are required for keeping statistical validity (power) in their analysis. The authors quoted a reference describing the significant positive association between LTH and obesity, 3 but the target subjects were children. There was a decreasing tendency of LTH by aging in adults by multiple regression analysis. 1 Further study is needed with special reference to generation. The authors also quoted a reference describing the LTH by laryngopharyngeal reflux and/or obstructive sleep apnea-hypopnea syndrome, 4 but Friedman et al 4 mentioned that laryngopharyngeal reflux was not significantly associated with LTH by multiple regression analysis. This discrepancy should be verified by the follow-up study, which would give information on the causality of the association.