Craniofacial phenotyping for prediction of obstructive sleep apnoea in a Chinese population

• Published in: Respirology (Carlton, Vic.) Vol. 21; no. 6; pp. 1118 - 1125
• Main Authors: Sutherland, Kate, Lee, Richard W. W., Petocz, Peter, Chan, Tat On, Ng, Susanna, Hui, David S., Cistulli, Peter A.
• Format: Journal Article
• Language: English
• Published: Australia Blackwell Publishing Ltd 01.08.2016
• Subjects: Adult; Anthropometry - methods; Asian Continental Ancestry Group; Cephalometry - methods; craniofacial morphology; Face; facial phenotype; Female; Hong Kong - epidemiology; Humans; Male; Middle Aged; Neck; Obesity - complications; obstructive sleep apnoea; photogrammetry; Photography - methods; Polysomnography - methods; prediction; Predictive Value of Tests; Regression Analysis; Risk Factors; Sleep Apnea, Obstructive - diagnosis; Sleep Apnea, Obstructive - ethnology; Sleep Apnea, Obstructive - etiology; Sleep Apnea, Obstructive - physiopathology; prediction; facial phenotype; obstructive sleep apnoea; craniofacial morphology; photogrammetry
• ISSN: 1323-7799; 1440-1843
• DOI: 10.1111/resp.12792

Background and objective Craniofacial morphology is a risk factor for obstructive sleep apnoea (OSA). Facial photography has previously shown utility in predicting OSA in a Caucasian sleep clinic. However, ethnic differences in OSA risk factors may influence these facial predictors. Our aim was to assess phenotypical facial measurements for OSA prediction in a Chinese population. Methods Calibrated frontal and profile facial photographs were taken before polysomnography. Photographs were analysed to derive head, face and neck measurements. Demographical, anthropometrical and facial photographical variables were considered in prediction models for OSA. OSA prediction models were derived using logistic regression and classification and regression tree techniques. Results Two‐hundred subjects were recruited (146 OSA, 54 controls). The OSA group contained more men (77% vs 61%) and were more obese. Logistic regression modelling found cervicomental angle (OR 1.06/degree, 95% CI: 1.03–1.09, P < 0.001) and face width (OR 1.7/cm, 95% CI: 1.1–2.7, P = 0.02) predicted OSA (area under the receiver operating characteristics curve 0.76). Classification and regression tree analysis identified cricomental space area, mandibular width, mandibular plane angle and neck soft tissue area as predictors (area under receiver operating characteristics curve 0.81). Conclusion In a Hong Kong Chinese sleep clinic, facial photographical measurements had predictive utility for OSA. Prediction models had similar accuracy and included variables to a previous Caucasian population. Craniofacial phenotyping by quantitative photography is a simple method for craniofacial assessment, which may have research and clinical applications. Facial photographical variables have previously shown predictive utility for OSA risk in a Caucasian sleep clinic population. Although there are ethnic differences in craniofacial OSA risk factors, facial photography performs similarly in a Chinese sleep clinic population.