Effect of obesity on sleep apnea pathogenesis differs in women versus men: multiple mediation analyses in the retrospective SNOOzzzE cohort

Abstract

There are multiple mechanisms underlying obstructive sleep apnea (OSA) development. However, how classic OSA risk factors such as body mass index (BMI) and sex portend to OSA development has not been fully described. Thus we sought to evaluate how obesity leads to OSA and assess how these mechanisms differ between men and women. The San Diego Multi-Outcome OSA Endophenotype (SNOOzzzE) cohort includes 3,319 consecutive adults who underwent a clinical in-laboratory polysomnography at the University of California, San Diego, sleep clinic between January 2017 and December 2019. Using routine polysomnography signals, we determined OSA endotypes. We then performed mediation analyses stratified by sex to determine how BMI influenced the apnea-hypopnea index (AHI) using OSA pathophysiological traits as mediators, adjusting for age, race, and ethnicity. We included 2,146 patients of whom 919 (43%) were women and 1,227 (57%) were obese [body mass index (BMI) > 30 kg/m²]. BMI was significantly associated with AHI in both women and men. In men, the adjusted effect of BMI on AHI was partially mediated by a reduction in upper airway stiffness (β_standardized = 0.124), a reduction in circulatory delay (β_standardized = 0.063), and an increase in arousal threshold (β_standardized = 0.029; P_{boot-strapped,all} < 0.05). In women, the adjusted effect of BMI on AHI was partially mediated by a reduction in upper airway stiffness (β_standardized = 0.05) and circulatory delay (β_standardized = 0.037; P_{boot-strapped,all} < 0.05). BMI-related OSA pathogenesis differs by sex. An increase in upper airway collapsibility is consistent with prior studies. A reduction in circulatory delay may lead to shorter and thus more events per hour (higher AHI), while the relationship between arousal threshold and OSA is likely complex.NEW & NOTEWORTHY Our data provide important insights into obesity-related obstructive sleep apnea (OSA) pathogenesis, thereby validating, and extending, prior research findings. This is the largest sample size study to examine the relationships between obesity and gender on OSA pathogenesis. The influence of obesity on sleep apnea severity is mediated by different mechanistic traits (endotypes).

Keywords: endotypes; gender; mediation analysis; obesity; sleep apnea.

Graphical abstract

Figure 1.

Estimation of endotypical traits from polysomnography flow signals. Representative example of quantifying pathophysiological obstructive sleep apnea (OSA) traits from polysomnography signals during non-rapid eye movement (NREM) sleep. Top: shows electroencephalogram power (EEG) with probability of arousal (green). Middle: flow from nasal pressure. The purple boxes represent respiratory events (apneas or hypopneas). Bottom: breath-by-breath the observed ventilation (purple), estimated chemical drive (black), and estimated ventilatory drive (i.e., chemical drive + drive due to arousals; green). The ventilatory drive preceding event-related arousals is taken as the arousal threshold (i.e., events without associated arousals do not contribute arousal threshold measurements), the difference between chemical drive and overall ventilatory drive during event-related arousals reflects the ventilatory response to arousals, while the time from drops in ventilation during events to the subsequent rise in chemical drive reflects chemoreflex delay (principally the circulation time between the lung and chemoreceptors, i.e., circulatory delay). Ventilation relative to eupnea at specific ventilatory drive levels is taken to assess the propensity to collapse during passive (V_min, ventilation during lowest decile of drive; V_passive, ventilation at eupneic drive) and active conditions (V_active, ventilation at the arousal threshold) when upper airway dilator muscles are relative hypotonic or activated, respectively. Loop gain is the magnitude of increase in chemical drive relative to a preceding, event-related drop in ventilation. The median values of all these measurements (taken repeatedly throughout NREM sleep) are used to represent each person’s trait, which can be summarized in an endogram (right). For more details, please see the original Refs –.

Figure 2.

Study flowchart. “Break the glass” denotes patients in the medical record whose personal health information is considered particularly sensitive (e.g., employees, celebrities, prisoners). BMI, body mass index. UCSD, University of California, San Diego.

Figure 3.

Multiple mediation model examining potential physiological mechanisms through which body mass index (BMI) leads to obstructive sleep apnea [OSA; as measured by the log-apnea-hypopnea index (AHI)] in women vs. men. The total effect of BMI on log-AHI consists of indirect effects (BMI → mediator → log-AHI) and a direct effect (BMI → log-AHI). Standardized beta coefficients (β_S) above the arrows indicate bootstrapped structural path coefficients, with blue indicating a positive relationship and red indicating an inverse relationship (e.g., in men, BMI reduces upper airway stiffness, but increases unstable ventilatory control). Statistical significance is denoted by asterisks and solid arrows (*P < 0.05), e.g., in men an increase of the BMI by 1 SD was significantly associated with a 0.21-SD decrease in upper airway stiffness, and 1-SD increase in upper airway stiffness was significantly associated with a 0.56-SD reduction in log-AHI (independent of all other indirect or direct effects). Thus the indirect effect of the BMI through upper airway stiffness is −0.21 × −0.56 = 0.118 (i.e., a 1-SD increase in BMI is associated with a 0.118-SD increase in log-AHI through its effects on upper airway stiffness). Details for direct, indirect, and total pathways including bootstrapped 95% confidence intervals (CI) are provided in Table 3 and Supplemental Table S2. In men, there was a significant total effect of BMI on log-AHI (β_S = 0.38; 95% CI: 0.33 to 0.42), explaining 66% of the variance in log-AHI (r² = 0.66). There was a significant direct effect of BMI (β_S = 0.17; 95% CI 0.13 to 0.21), consistent with partial mediation. The BMI increased the log-AHI (i.e., worsened OSA) by decreasing upper airway stiffness (β_S = 0.118), increasing the arousal threshold (β_S = 0.027), and decreasing circulatory delay (β_S = 0.06). In women, there was a significant total effect of BMI on log-AHI (β_S = 0.18; 95% CI: 0.12 to 0.25), explaining 53% of the variance in log-AHI (r² = 0.53). There was a significant direct effect of BMI (β_S = 0.15; 95% CI: 0.10 to 0.20), consistent with partial mediation. The BMI increased the log-AHI (i.e., worsened OSA) by decreasing circulatory delay (β_S = 0.041). BMI decreased the arousal threshold, but a lower arousal threshold was associated with a lower log-AHI, consistent with competitive mediation (β_S = −0.019; 95% CI: −0.04 to -0). In women, upper airway stiffness was inversely related to log-AHI, but BMI was not associated with upper airway stiffness; thus there was no mediation through upper airway stiffness. The effect of BMI on log-AHI through upper airway stiffness and arousal threshold was significantly modified by sex (P_adjusted < 0.05).