Joint association of sleep quality and physical activity with metabolic dysfunction-associated fatty liver disease: a population-based cross-sectional study in Western China

MAFLD affects roughly one-quarter of adults worldwide, with rapidly rising prevalence in China. Physical inactivity and poor sleep are widespread and each has been linked to metabolic dysfunction. PA improves insulin sensitivity and weight control, key in MAFLD management, while sleep disturbances (short duration, insomnia, snoring) are associated with adverse metabolic outcomes. PA and sleep are interconnected behaviors that can influence each other and metabolic homeostasis. However, evidence on their joint association with MAFLD is limited. This study aimed to evaluate the combined impact of PA and a comprehensive measure of sleep quality (including daytime napping) on prevalent MAFLD in adults from Western China.

Prior studies demonstrate inverse associations between PA and NAFLD/MAFLD and positive associations between poor sleep characteristics (short duration, insomnia, snoring) and metabolic disorders, including NAFLD. PA can influence sleep duration and circadian rhythm, while sleep patterns can affect time available and energy for PA, indicating bidirectional links. Despite independent associations shown in multiple cohorts, the joint effect of PA and sleep on MAFLD has not been well characterized. Evidence from mortality studies suggests synergistic effects of insufficient exercise and poor sleep, but comparable data for MAFLD were lacking, especially in Chinese populations and with inclusion of daytime napping.

Design and population: Population-based cross-sectional analysis using baseline data from the Population-based Cohort study of Chronic Diseases in Xinjiang (PCCDX), conducted in Urumqi and Korla (July 2019–September 2021) with a two-stage stratified sampling strategy. Of 12,295 adults aged 30–74 years enrolled and ultrasonographed, 2,206 were excluded for missing MAFLD data (n=979), history of liver cirrhosis/resection/cancer (n=13), extreme outliers for waist circumference or BMI (>3 SD; n=582), or missing sleep behavior data (n=632), yielding 10,089 participants. Ethics approval and informed consent were obtained. Outcome: MAFLD diagnosed radiologically (ultrasound-based hepatic steatosis with metabolic criteria; details in Supplementary Methods). Exposures: Sleep behaviors assessed via the Pittsburgh Sleep Quality Index. Dimensions included bedtime (Beijing Time before 0:00, 0:00–1:00, after 1:00), night sleep duration (<7, 7–8, >8 h/day), insomnia, snoring, and excessive daytime sleepiness (never/rarely, sometimes, usually), and daytime napping (0–30 min/day, >30 min/day). A composite healthy sleep score (0–6; higher indicates better) was constructed from six dimensions and categorized, guided by restricted cubic spline results, as good (≥5), intermediate (3–4), and poor (≤2). Physical activity measured by the International Physical Activity Questionnaire; total weekly MET-minutes computed and categorized as low (<600), medium (600–3000), and high (>3000) per PURE study. MVPA was dichotomized according to WHO guideline (meeting vs not meeting: ≥150 min moderate or ≥75 min vigorous or equivalent per week). Covariates: Age, gender, education, marital status, current smoking, alcohol drinking, sedentary time, diet diversity score, BMI status; selection informed by a directed acyclic graph. Missing data handling: Multiple Imputation by Chained Equations. Statistical analysis: Descriptive statistics by MAFLD status; trends via linear regression (continuous) and Mantel-Haenszel chi-square (categorical). Independent associations of sleep quality and PA (categorical) with MAFLD assessed using multivariable logistic regression with mutual adjustment; as continuous exposures, restricted cubic splines with knots at the 5th, 50th, and 95th percentiles examined nonlinearity. Joint analyses: Interaction testing between sleep quality and PA; stratified spline analyses across PA categories; analyses of sleep quality vs MAFLD at different PA/MVPA levels; and combined-category logistic models using nine groups (total PA × sleep) with high PA + good sleep as reference and six groups (MVPA × sleep) with MVPA met + good sleep as reference. Subgroup analyses by gender, age, and metabolic comorbidities; sensitivity analyses using complete cases, excluding participants on medications affecting sleep, and further adjusting for metabolic disorders. Residual confounding assessed via E-values. Software: Stata 16.0 and R 4.2.3. Two-sided P<0.05 considered significant.

- Sample: 10,089 participants; mean age 47.0 (SD 9.1) years; 51.6% men; MAFLD prevalence 38.2% (n=3,854). Most had intermediate sleep quality (60.7%); 12.3% had high total PA; 79.2% did not meet MVPA guideline. - Independent associations: Sleep quality showed a linear association with MAFLD prevalence (P overall <0.001; P non-linear = 0.640). Compared with good sleep, adjusted ORs for MAFLD were 1.25 (95% CI: 1.11–1.40) for intermediate and 1.45 (1.23–1.70) for poor sleep (P trend <0.001). - Total PA: A dose-dependent inverse association with MAFLD (P overall <0.001; P non-linear = 0.554). Relative to high PA, adjusted ORs were 1.24 (1.07–1.44) for medium and 1.37 (1.15–1.63) for low PA. Not meeting MVPA guideline associated with higher odds of MAFLD: OR 1.37 (1.21–1.54). - Stratified by PA: Within each PA category, worse sleep quality associated with higher MAFLD prevalence. High PA with poor sleep: OR 2.53 (1.59–4.01) vs high PA with good sleep. In medium PA, intermediate vs good sleep: OR 1.20 (1.04–1.38); poor vs good: OR 1.52 (1.25–1.86). - Stratified by MVPA: Even among participants meeting MVPA recommendations, poor vs good sleep associated with OR 2.01 (1.40–2.90); intermediate vs good sleep: OR 1.64 (1.27–2.12). Among those not meeting MVPA, intermediate vs good: OR 1.16 (1.02–1.33); poor vs good: OR 1.33 (1.11–1.60). P for interaction: sleep × total PA <0.001; sleep × MVPA = 0.029. - Joint categories: Using high PA + good sleep as reference, the highest odds were observed for medium PA + poor sleep (OR 2.73, 95% CI: 1.96–3.80), followed by high PA + poor sleep (2.54, 1.59–4.08) and low PA + intermediate sleep (2.49, 1.81–3.42). Compared with MVPA met + good sleep, no MVPA + poor sleep had OR 2.36 (1.81–3.08). - Sensitivity/subgroups: Results were robust in complete-case analyses, excluding medication users, and with additional adjustment for metabolic disorders; consistent across subgroups by sex, age, hypertension, diabetes, and metabolic syndrome. E-value analyses suggested unmeasured confounding unlikely to fully explain associations.

The study demonstrates that sleep quality and PA have independent and synergistic associations with prevalent MAFLD. Poor sleep quality substantially increases the odds of MAFLD across all PA levels, and good sleep provides additional protection even among those with high PA. Notably, achieving or exceeding WHO-recommended MVPA levels does not sufficiently mitigate the adverse association between poor sleep and MAFLD, underscoring sleep as a critical and complementary target alongside PA for MAFLD prevention. The findings align with evidence linking PA to improved metabolic health and sleep disturbances to metabolic dysregulation, potentially via circadian mechanisms and hepatic metabolic regulation. The observed interactions were consistent across demographic and clinical subgroups, supporting generalizability within the study context. Public health strategies should integrate sleep improvement with PA promotion to more effectively address MAFLD risk.

Poor sleep quality is associated with higher MAFLD prevalence, and its adverse association is amplified by insufficient PA. While maintaining PA is important, currently recommended thresholds of MVPA are not sufficient to offset the harms of poor sleep. Interventions that concurrently target improving sleep quality and increasing PA may be more effective for MAFLD prevention and management than focusing on either behavior alone. Future research should include prospective cohorts and randomized interventions to establish causality and elucidate biological mechanisms mediating the joint effects of sleep and PA on MAFLD.

Cross-sectional design precludes causal inference. Sleep behaviors and PA were self-reported, introducing potential recall and misclassification bias (likely non-differential). The study focused on total PA and MVPA thresholds without differentiating specific PA modalities or intensities beyond these categories. MAFLD diagnosis relied on ultrasound rather than histology, which, while appropriate for population studies, may misclassify hepatic steatosis. Although extensive adjustments and E-value analyses were performed, residual confounding cannot be entirely excluded.