Improved diagnostic accuracy for pediatric obstructive sleep apnea using an out-of-center sleep test

Abstract

ObjectivePolysomnography (PSG) is considered the gold standard for diagnosing obstructive sleep apnea syndrome (OSA) in children. However, many hospitals do not carry out PSG evaluations, and use out-of-center sleep test (OCST) devices for diagnosis instead. The aim of this study was to confirm the reliability of OCSTs for the diagnosis of pediatric OSA. We also investigated the factors influencing diagnostic reliability of OCST for the severe OSA patients who should be treated earlier. MethodsThis was a retrospective study using the Ota Memorial Sleep Center database. We analyzed the data of children who underwent Type 4 OCST at home and Type 1 PSG in the sleep lab between April 2006 to April 2015. Cephalometric parameters and anthropometric findings such as enlarged tonsils were also evaluated. We compared the 3% oxygen desaturation index (ODI3%) measured by OCST with the apnea-hypopnea index (AHI) measured by PSG. We used Receiver Operator Curve (ROC) to calculate the optimal OCST- ODI3% value to diagnose PSG-AHI ≥10 per hour. In order to determine which factors increase the accuracy of OCST, we calculated the accuracy, sensitivity and specificity in regard to the predicted values using multiple logistic regression analysis. The Ethics Committee of Ota General Hospital approved the study (approval no. 21018). ResultsA total of 191 children were enrolled in this study. The study included 127 boys and 64 girls, with a mean age of 5.4 years (range: 3-8 years), BMI of 15.7 kg/m² (range: 11.5-35.7 kg/m²), PSG-AHI of 17.4 per hour (range: 0.3-89.8 per hour). The sensitivity, specificity and accuracy with an OCST-ODI cutoff of 6.3 per hour were 64.4%, 70.3% and 67.5%, respectively, to detect PSG-AHI ≥ 10 per hour for children with suspected OSA. Multivariable stepwise regression revealed that increases of sensitivity, specificity and accuracy with an OCST-ODI cutoff of 6.3 per hour were independently predicted by facial axis, which is cephalometric angle of 81° or less, and tonsil hypertrophy, which is Brodsky +3 or +4, showing increases to 73.3%, 71.3%, and 72.3%, respectively, whereas age, gender, body mass index, adenoid size and other cephalometric parameters were not significant predictors. ConclusionThe results of the statistical analyses suggest that it would be useful to add the assessment of tonsil size and facial axis as well as OCST to determine whether the threshold of PSG-AHI ≥ 10 per hour has been crossed.