Sleep medication statistically links activity and diabetes in Azar cohort

In a population-based cohort of 13,259 adults initially free of type 2 diabetes, higher physical activity was associated with a lower risk of developing the disease, while sleep medication use accounted for a modest statistical share of the relationship. The Azar Cohort Study analysis, published in BMC Public Health on 30 September 2026, examined whether sleep medication use helped explain the observed association between physical activity and incident diabetes.

The cohort included adults aged 35–70 years, and researchers compared moderate or lower physical activity with high or very high activity. Physical activity was expressed as metabolic equivalent task hours per week, or MET-h/week. Modified Poisson regression with robust variance was used to estimate risk ratios. In the unadjusted analysis, moderate or lower activity was associated with greater risk of incident type 2 diabetes compared with high or very high activity: the risk ratio was 1.49, with a 95% confidence interval of 1.03–2.14. These findings describe an association and do not establish that changing physical activity would directly cause a change in diabetes risk.

Adjustment using a directed acyclic graph reduced the estimated risk ratio to 1.31, with a 95% confidence interval of 1.01–1.70. An expanded model produced a further attenuation, yielding a risk ratio of 1.23 and a 95% confidence interval of 1.01–1.50. Across these models, the estimates consistently associated moderate or lower physical activity with a higher occurrence of incident type 2 diabetes than high or very high activity, although the strength of the association varied with the covariates included.

Sleep medication use provided a possible statistical explanation for part of the activity–diabetes association. Participants with moderate or lower physical activity were more likely to report sleep medication use than those with high or very high activity. The adjusted odds ratio was 1.93, with a 95% confidence interval of 1.23–3.02. The researchers conducted mediation analyses using logistic and Poisson regression within a potential-outcomes framework to estimate the indirect association through sleep medication use.

The estimated indirect association was statistically significant in both the directed-acyclic-graph-adjusted model and the expanded model. It corresponded to a risk ratio of 1.09, with a 95% confidence interval of 1.01–1.16, in the former and a risk ratio of 1.04, with a 95% confidence interval of 1.01–1.06, in the latter. The estimated proportions attributed to the indirect association were 37.4% and 20.8%, respectively. Thus, the proportion classified as mediated was smaller after broader covariate adjustment, even though the indirect estimate remained statistically significant in both models.

When physical activity was examined on a continuous scale rather than divided into the activity categories used for the main comparison, the analysis showed an inverse association with incident type 2 diabetes. The researchers reported no significant evidence of nonlinearity, indicating that the continuous association did not provide clear statistical support for a sharp change at a particular activity level. Together, the categorical and continuous results were consistent with greater physical activity being associated with a lower risk of developing type 2 diabetes.

The central limitation concerns the timing of the measurements. Physical activity and sleep medication use were assessed at baseline, and the analysis therefore could not establish that sleep medication use occurred after an activity difference and subsequently contributed to the development of diabetes. The authors explicitly characterized the mediation findings as statistical rather than temporally established causal mediation. The results also concern people who were free of type 2 diabetes at baseline and do not directly describe the risk among adults who already had the condition.

Overall, the study supports an inverse association between physical activity and incident type 2 diabetes in the Azar Cohort and identifies sleep medication use as a potential statistical mediator of that relationship. The magnitude of the indirect association was modest in the fully expanded model, and its interpretation depends on the adjustment set. Because the design measured both physical activity and medication use at the start of the cohort, the analysis cannot determine whether sleep medication use is a causal pathway, a marker of other factors, or a consequence of unmeasured influences. The findings therefore narrow an epidemiological association without claiming that sleep medication has been shown to cause diabetes or that changing medication use would prevent the disease.

Subject of Research: Statistical mediation of physical activity, sleep medication use, and incident type 2 diabetes in the Azar Cohort Study.

Article Title: Physical activity and incident type 2 diabetes: the potential mediating role of sleep medication use in the population-based Azar Cohort

Article References: Haghshenas, R., Somi, M. H., & Faramarzi, E. (2026). Physical activity and incident type 2 diabetes: the potential mediating role of sleep medication use in the population-based Azar Cohort. BMC Public Health. https://doi.org/10.1186/s12889-026-29678-y

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29678-y

Keywords: type 2 diabetes, physical activity, sleep medication, statistical mediation, cohort study, incident diabetes, risk ratios, Azar Cohort, population health, Physical, activity, incident