Associations of polysomnographic measures of obstructive sleep apnea, and nocturnal oxygen saturation with incident type 2 diabetes mellitus in middle‐aged and older men

Obstructive sleep apnea (OSA) has been associated with incident type 2 diabetes mellitus (T2DM); however, few prospective epidemiological studies have accounted for important T2DM predictors including pre-diabetes status and testosterone. Participants in the longitudinal Men Androgens Inflammation Lifestyles Environment and Stress (MAILES) study, who underwent eight-channel home-based polysomnography (PSG) in 2010-2011 (n = 824) and were free of diabetes at baseline were included in the analysis (n = 682). From 2015 to 2021, 78.6% (n = 536) completed at least one follow-up assessment. Incident T2DM was determined by self-reported doctor diagnosis, diabetes medications, plasma glucose (fasting ≥7.0 mmol/L or random ≥11.0 mmol/L) or glycated haemoglobin ≥6.5%. Conservative hierarchical Poisson regression models adjusted associations of PSG metrics (categorical and continuous) for age, waist circumference, baseline fasting glucose and testosterone concentrations. In all, 52 men (9.7%) developed T2DM over a mean (range) of 8.3 (3.5-10.5) years. Significant age- and waist circumference-adjusted association of incident T2DM with rapid eye movement (REM) sleep apnea-hypopnea index (AHI) ≥20 events/h (incidence rate ratio [IRR] 1.5, 95% confidence interval [CI] 0.8-2.8; p = 0.23] and highest quartile of delta index (IRR 2.1, 95% CI 0.95-4.6; p = 0.066) were attenuated after adjustment for baseline glucose and testosterone, and the association with the lowest quartile of mean oxygen saturation persisted (IRR 4.2, 95% CI 1.7-10.3; p = 0.029). Categorical measures of AHI severity, oxygen desaturation index, and hypoxia burden index (HBI) were not independently associated with incident T2DM. Associations with T2DM were similar when continuous PSG variables were used; however, HBI was significant (IRR 1.015, 95% CI 1.006-1.024; p = 0.007). In a sub-sample with OSA treatment data (n = 479), these significant associations persisted after excluding adequately treated OSA (n = 32). Understanding underlying OSA endotypes generating hypoxaemia may identify opportunities for diabetes prevention.