Joint Association of Hydration Status and Sleep Quality with Type 2 Diabetes Mellitus in a Population-Based Study.

Objectives

To examine the joint association of hydration status and sleep quality with prevalent type 2 diabetes mellitus (T2DM) and to quantify their additive and multiplicative interactions.

Methods

This cross-sectional study included 4762 participants from the baseline Hainan Prospective Cohort Study. Logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the four combinations of hydration status and sleep quality. Additive interaction was assessed with the relative excess risk due to interaction (RERI), the attributable proportion (AP), and the synergy index (S); multiplicative interaction was assessed with a product term. Predicted probabilities were estimated by g-computation.

Results

The prevalence of T2DM was 8.13% (387/4762). Compared with non-dehydration and good sleep quality, the combination of dehydration and poor sleep quality was associated with substantially higher odds of prevalent T2DM (adjusted OR 9.92, 95% CI 7.14 to 14.00). All three additive interaction measures were statistically significant (RERI 7.74, 95% CI 4.99 to 10.50; AP 78.04%, 95% CI 69.28 to 86.80; S 7.57, 95% CI 3.49 to 16.42), as was the multiplicative interaction (OR 3.96, 95% CI 2.30-6.90). On the absolute risk scale, the interaction contrast was 14.53 percentage points (95% CI: 11.01 to 18.34), with an attributable proportion of 80.5% (95% CI: 66.3 to 94.6) and a synergy index of 5.14 (95% CI: 2.94 to 17.25). The predicted probability of prevalent T2DM was highest in the combined dehydration and poor sleep group.

Conclusions

The combination of dehydration and poor sleep quality was associated with higher odds of prevalent T2DM than either condition alone, and a statistically significant additive interaction was observed. Because these additive interaction measures were derived from odds ratios for a non-rare outcome, their absolute magnitude may be overestimated on the odds-ratio scale; the risk-difference-scale measures confirmed the interaction. Given the cross-sectional design, reverse causation cannot be excluded, and longitudinal data are needed to clarify the temporal sequence of these associations.