Chinese Adults Show Lower Depressive Symptoms with Moderate Physical Activity

A large-scale cross-sectional analysis of Chinese adults suggests that engaging in low to moderate levels of physical activity is associated with reduced continuous depressive-symptom scores compared to inactive individuals, though the data does not establish a definitive optimal dose or a stable linear dose-response relationship.

The research, published in Frontiers in Psychology, analyzed data from the 2024 Chinese National Sports and Society Survey (CNSSS), encompassing 10,663 adults across 31 provincial-level regions of mainland China. The study aimed to investigate the association between total physical activity and depressive symptoms, specifically looking for potential deviations from a simple linear dose-response pattern. Depressive symptoms were measured using a five-category, ten-item score ranging from 10 to 50, derived from a questionnaire that assessed experiences over the preceding week. This methodology involved fitting complex sampling-weighted estimating equations, incorporating observation-level sandwich variance, and employing categorical models and a two-part restricted cubic spline model to evaluate the relationship across different activity levels.

The categorization of physical activity volume was based on MET-min/week, calculated using intensity-specific coefficients: Total MET-min/week was defined as $8.0 imes ext{vigorous weekly min} + 4.0 imes ext{moderate weekly min} + 3.3 imes ext{weekly walking min}$. The resulting volume categories were inactive (< 1 MET-min/week), low (1–< 600 MET-min/week), moderate (600–< 3,000 MET-min/week), and high ($ ext{≥}3,000$ MET-min/week). Adjusting for numerous covariates—including nonlinear age, sex, education, household registration, marital status, income, and chronic disease—the study found that relative to inactivity, adjusted symptom-score differences were significantly lower in both low and moderate activity groups. Specifically, low activity showed a difference of $-3.11$ (95% CI $-5.94$ to $-0.27$), and moderate activity showed a difference of $-3.41$ (95% CI $-6.13$ to $-0.69$).

A key finding regarding the dose-response relationship was that low and moderate activity did not differ clearly from one another; the difference between moderate and low activity was $-0.30$ (95% CI $-1.38$ to $0.78$). While the primary nonlinear-term Wald test in the restricted cubic spline model yielded a P-value of $0.034$, the evidence for a stable nonlinear pattern weakened considerably when the model was adjusted with three knots (P = $0.309$) or when activity levels above the 95th percentile were excluded (P = $0.063$). This suggests that the observed potential for curvature is highly dependent on the specific modeling choices and the distribution of the data used in the analysis. Furthermore, the high activity group, defined as $ ext{≥}3,000$ MET-min/week, showed an adjusted difference of $-2.20$ (95% CI $-4.96$ to $0.57$) relative to inactivity, but its interval included zero, meaning high activity was not shown to be more harmful than inactivity in this sample.

The study emphasized that activity volume, as measured by total MET-min/week, is distinct from activity intensity, and crucially, inactivity is not a direct measure of prolonged sitting. The primary analysis utilized a complete-case sample of 10,663 adults, with 722 cases excluded due to missing or invalid data across activity, symptom, or covariate variables. The selection of covariates was strategic, encompassing demographic, socioeconomic, and health differences such as age (modeled with restricted cubic splines at 21, 38, 54, and 76 years), sex, education, urban-rural registration, marital status, annual personal income (transformed as $ ext{ln}(1 + ext{income})$ after capping), and chronic disease status.

The use of the restricted cubic spline model allowed the researchers to explore the dose-response shape, which is critical in public health research where simple linear assumptions may be insufficient. The active curve derived from this model declined across lower volumes and rose across parts of the upper distribution, indicating complex relationships. However, the authors were careful to note that no stable numerical minimum, optimal dose, harmful high-activity effect, or causal mechanism could be inferred from this cross-sectional design. The cross-sectional nature of the data inherently prevents the establishment of temporal order or causality, leaving reverse causation and self-selection as plausible considerations that cannot be ruled out by the survey data alone.

The limitations of the study are substantial and must be considered when interpreting the findings. The self-reported nature of both physical activity and depressive symptoms is susceptible to recall and social-desirability bias. Moreover, the five-category symptom format and its study-specific recoding—which was not a validated conventional CES-D-10 protocol—introduce measurement uncertainty, as alternative recoding methods changed the exploratory screening percentages materially. The inability to separate occupational, transport, domestic, and leisure-time activity domains means the total MET-min/week measure is a composite that does not distinguish between different types of physical exertion. Furthermore, crucial variables such as smoking, alcohol use, and direct general social support were absent from the supplied questionnaire, limiting the scope of confounding adjustments.

Despite these limitations, the findings contribute valuable exploratory evidence from a large Chinese population. The primary conclusion remains that low and moderate activity levels are associated with lower continuous depressive-symptom scores than inactivity, without a clear differentiation between the low and moderate groups. The non-linear pattern observed in the spline analysis is highly sensitive to the number of knots and the exclusion of upper-tail observations, meaning a precise optimal dose cannot be claimed. The high-activity estimates do not demonstrate greater harm than inactivity. Therefore, the authors stress that these results do not support a new, specific activity prescription, but rather reinforce the importance of maintaining baseline physical activity levels within the context of existing public health guidelines. Future longitudinal and intervention studies are necessary to address temporality, activity domains, and measurement validity to determine if any observed nonlinear patterns are reproducible.

Subject of Research: Association between total physical activity volume and depressive symptoms in Chinese adults using cross-sectional survey data.

Article Title: Physical activity and depressive symptoms among Chinese adults a cross sectional analysis

Article References: Xu, M., Yuan, K., & Zhang, X. (2026). Physical activity and depressive symptoms among Chinese adults a cross sectional analysis. Frontiers in Psychology, 17, Article 1989343. https://doi.org/10.3389/fpsyg.2026.1989343

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1989343

Keywords: Depressive Symptoms, Physical Activity, Cross-sectional Study, Chinese Adults, Dose-response, Public Health, CNSSS, Mental Health, Physical, activity, depressive, symptoms