Obstructive sleep apnea (OSA) is a common disease worldwide, affecting over 4% of men and 2% of women [1]. It is characterized by repetitive upper airway collapse during sleep, which induces frequent arousals and oxygen desaturation. OSA is associated with a variety of symptoms and conditions such as fatigue, daytime somnolence, headache, myocardial infarction, arrhythmias, stroke, and an increased incidence of motor vehicle accidents. Therefore, accurate diagnosis and treatment of OSA is recommended to improve quality of life and to reduce the risk of associated morbidity and mortality.

Numerous studies report that OSA is closely related with obesity, which is thought to reduce the diameter of upper airway and increase its collapsibility [2,3]. Several anthropometric measurements like body mass index (BMI), neck circumference (NC), waist circumference (WC), and waist-hip ratio (WHR) have been frequently used to investigate the association between obesity and OSA [4-22]. However, the results have varied based on age, gender, and ethnicity.

Asian patients with OSA are suspected to be affected to a greater degree by obesity than Caucasians [23,24]. However, this finding is not conclusive because large-scale comparisons between different ethnicities are difficult to perform at a single research center. The goal of this study was to provide a systematic and critical appraisal of the available evidence on the anthropometric differences between Asian and Caucasian patients with OSA. Understanding the different obesity phenotypes between Asian and Caucasian may help to elucidate the underlying mechanisms of OSA, provide important clinical tools for screening of OSA, and have an effect on selecting treatment modalities.

This meta-analysis included 19 studies consisting of data from a total of 2,966 patients (803 controls and 2,163 with OSA, Table 1). Males represented 57.9% of controls (range, 8.3% to 90%), and 76.5% of patients with OSA (range, 40.4% to 90%). For controls and OSA group, there was no statistical difference in sex and age between Asians and Caucasians. Sixteen studies (84.2%) defined the presence of OSA with the criteria of AHI≥5, and 17 studies (89.5%) performed standard in-laboratory full-night polysomnography.

There are various pathophysiologic factors associated with OSA such as obesity, unfavorable anatomical features, neuromuscular abnormalities, or genetic problems. Of these, obesity is thought to be one of the most common and important risk factors for OSA. Obesity is associated with increased narrowing and obstruction of the airway due to changes at the level of the soft palate, pharyngeal wall, and tongue base [28]. In addition, obesity may lead to modifications of neuromuscular function resulting in increased airway collapsibility and weakness of neuromuscular compensation [3,29].

The objective of this study was to identify any potential ethnicity-specific characteristics of obesity-related anthropometric measurements (BMI, NC, WC, and WHR). We performed a schematic meta-analysis by stratifying the raw data by ethnicity (Asian vs. Caucasian). The key findings of this study were that: (1) BMI, WC, and WHR were not different between patients with OSA and controls after evaluating for publication bias; (2) Patients with OSA had greater NC than controls in both Asians and Caucasians; and (3) Anthropometric phenotypes (BMI, NC, WC, and WHR) of Asian OSA patients were not different from Caucasian OSA patients. Our study results provide valuable data for clinicians when evaluating patients with different ethnicities with suspected OSA by identifying anthropometric characteristics associated with OSA that can be used to assist with the diagnosis prior to a standard polysomnography.

Previous studies have reported that obesity-related measurements such as BMI, WC, and WHR are significant risk factors for OSA [2,3]. Surprisingly, our meta-analysis found no difference in BMI, WC, and WHR between patients with OSA and controls when publication bias was considered. Only NC was greater and statistically significant among OSA patients regardless of the ethnicity. We used the ‘trim and fill method’ described by Duval and Tweedie [27] to identify and reconcile potential publication bias. This removes (trims) the smaller studies causing funnel plot asymmetry, and uses the trimmed funnel plot to estimate the true center of the funnel. Omitted studies and their missing counterparts are replaced around the center (fill). The results of this analysis are somewhat surprising, but raises the possibility that some of the anthropometric features of obesity, namely BMI, WC, and WHR, should be evaluated more carefully as true risk factors for OSA in future studies. Our findings may indicate that many non obese patients with OSA are still unrecognized or undiagnosed.

Among the four anthropometric measurements evaluated in our analysis, only NC was found to be statistically significant between OSA patients and controls. Although obesity is often regarded as a principle risk factor in the occurrence of OSA, our results suggest that the NC is more closely correlated with OSA without publication bias. Interestingly, we found no difference in NC between Asian and Caucasian patients with OSA and controls. Several studies report that Asians have more severe OSA than Caucasians. Ong et al. [30] evaluated demographic and polysomnographic data between Caucasians and Asians, and found that Asians with OSA had more severe disease than Caucasians when matched for age, gender, and BMI. And Li et al. [31] found that despite having a lower BMI than Caucasians, Asian men had more severe OSA. In this study, we investigated whether differences in obesity phenotypes between Asian and Caucasians were risk factors in OSA. However, we found that most of anthropometric data (BMI, WC, and WHR) lack of confidence due to publication bias and did not show interethnic differences. NC was the most reliable measurement and most likely represents a true risk factor for OSA in both Asian and Caucasian patients.

There are several limitations of this study. The trim and fill method was used to assess if publication bias was present and to estimate the effect when the bias is removed. It is one of the most common statistical methods used to resolve potential publication bias in meta-analysis, and the method has been found to be accurate [32]. However, it is not the only method available, and it requires several statistical assumptions, including that funnel plot asymmetry is caused solely by publication bias. Further studies are required to confirm our results. Finally, sex, age, and severity of OSA were not included in this meta-analysis due to lack of data in the reviewed studies.

In conclusion, obstructive sleep apnea is an airway disease known to be closely related to obesity. However, our study found that the many of the features of obesity such as BMT, WC, and WHR might not be significant risk factors for OSA due to publication bias. In this study, NC demonstrated a strong association with OSA, and this finding was not different between Asians and Caucasians.