Joint diet and activity trajectories linked to multimorbidity risk in Chinese women

A sustained combination of a more inflammatory diet and low physical activity was associated with a 62% higher risk of developing cardiometabolic multimorbidity among Chinese women in a longitudinal analysis of the China Health and Nutrition Survey. During an average follow-up of 8.2 years, researchers identified three joint trajectories linking dietary inflammatory potential and activity from 2004 through 2011, then tracked the women for incident multimorbidity through 2015. Women in the higher-dietary-inflammation and inactive group had an adjusted hazard ratio of 1.62, with a 95% confidence interval of 1.14 to 2.31, compared with women in the lower-inflammation and moderately active reference group.

Cardiometabolic multimorbidity was defined as the first coexistence of at least two of four conditions: hypertension, diabetes, stroke and heart disease. The observational cohort included 3,970 women who were at least 18 years old in 2004 and had no multimorbidity at baseline. Among them, 400, or 10.1%, developed the condition during follow-up. Participants entered the analysis if they had dietary and activity data from at least two survey waves between 2004 and 2011 and at least one subsequent assessment of multimorbidity. Women who were pregnant or breastfeeding at baseline, already had multimorbidity, or reported implausibly low or high total energy intake were excluded.

Dietary intake at each survey wave was reconstructed from three consecutive 24-hour recalls, generally covering two weekdays and one weekend day, supplemented by household food-weighing records. Interviewers used food models, photographs and other visual aids to estimate portion sizes. The researchers calculated a dietary inflammatory index from 24 parameters: 18 nutrients and six foods, including energy, protein, fats, carbohydrates, fiber, cholesterol, several vitamins and minerals, alcohol, garlic, ginger, onion, tea, chili pepper and soy isoflavones. Each parameter was standardized against global reference values, converted to a centered percentile and multiplied by a literature-derived inflammatory effect score. The sum represented the overall index, with higher scores indicating a more pro-inflammatory dietary pattern.

Physical activity was measured through questionnaires covering occupational, household, transportation and leisure-time activity. Participants reported the frequency and duration of activities during the preceding year, and the researchers converted these reports into metabolic-equivalent hours per week. Occupational work was classified as light, moderate or heavy, while five household activities, four commuting modes and several forms of exercise were included. Total activity was the sum across all four domains. This measure therefore captured both structured exercise and everyday movement, although its reliance on self-report limited precision and introduced the possibility of recall bias.

The analysis modeled diet and activity jointly rather than treating them as isolated baseline exposures. Group-based trajectory modeling used four repeated measurements from 2004, 2006, 2009 and 2011, with dietary and activity values standardized within each wave so that the fitted patterns reflected each participant’s relative position. Measurements taken on or after a participant’s first multimorbidity assessment were excluded from trajectory estimation to preserve the exposure-outcome sequence. Models with two through six groups and several polynomial forms were compared using information criteria, posterior probabilities, group size and interpretability. The selected three-group model had an average posterior probability of at least 0.750 and no group smaller than 13.6%.

The resulting trajectories divided the cohort into lower dietary inflammation with moderate activity, comprising 559 women or 14.1%; moderate dietary inflammation with high activity, comprising 541 women or 13.6%; and higher dietary inflammation with inactivity, comprising 2,870 women or 72.3%. The groups remained distinctly separated across the measured waves. The lower-inflammation group averaged a dietary inflammatory index of 1.6 and 150.8 metabolic-equivalent hours per week of activity, while the higher-inflammation, inactive group averaged an index of 3.3 and 103.0 such hours. The latter group was also older, had lower educational attainment, higher blood pressure and generally less favorable nutrient intake, demonstrating that the trajectories captured more than two isolated behavioral variables.

Kaplan–Meier curves showed a faster loss of multimorbidity-free survival in the higher-inflammation, inactive group, with a log-rank p value below 0.0001. Its unadjusted hazard ratio was 1.91, with a 95% confidence interval of 1.36 to 2.69. The association persisted after adjustment for age, residence, education, marital status, body mass index, systolic and diastolic blood pressure, baseline presence of one cardiometabolic condition, smoking and drinking. By contrast, the moderate-inflammation, highly active group had no significant increase in risk in the unadjusted or adjusted models, with a fully adjusted hazard ratio of 1.13 and a 95% confidence interval of 0.70 to 1.84.

Subgroup analyses did not identify significant effect modification by age, education, marital status or baseline component-disease status. The elevated risk among women in the higher-inflammation, inactive trajectory was statistically significant in several subgroups, including women aged 50 or older, those with no more than primary education, those who were married or cohabiting, and those without a cardiometabolic component condition at baseline. Results also remained similar when women who developed multimorbidity within two years were excluded, when death was treated as a competing event, and when the analysis was restricted to women initially free of all four component diseases.

The findings point to long-term behavioral patterns as potential warning signals before multiple cardiometabolic conditions accumulate, but they do not establish that diet or inactivity caused the outcome. The analysis adjusted for several demographic, lifestyle and clinical factors, and sensitivity analyses reduced concern about early disease influencing behavior and about pre-existing component conditions driving the results. Still, the observational design cannot eliminate residual confounding or reverse causation, and unmeasured factors such as medication use, sleep, psychological health, family history and menopausal status could affect the observed association. The self-reported activity and largely self-reported diagnoses may also have introduced measurement error, while incomplete follow-up created the possibility of selection bias. The findings therefore identify an association in Chinese women and require validation in other populations with objective activity measures, repeated clinical measurements and more detailed reproductive histories.

Subject of Research: Longitudinal associations of dietary inflammatory potential and physical activity with cardiometabolic multimorbidity in Chinese women.

Article Title: Longitudinal joint trajectories of the dietary inflammatory index and physical activity and risk of incident cardiometabolic multimorbidity among women: evidence from the China Health and Nutrition Survey

Article References: Li, S., Liu, X., Gong, K., Tian, Y., & Yu, H. (2026). Longitudinal joint trajectories of the dietary inflammatory index and physical activity and risk of incident cardiometabolic multimorbidity among women: evidence from the China Health and Nutrition Survey. Frontiers in Public Health, 14, Article 1916853. https://doi.org/10.3389/fpubh.2026.1916853

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1916853

Keywords: cardiometabolic multimorbidity, dietary inflammatory index, physical activity, Chinese women, longitudinal cohort, dietary inflammation, China Health and Nutrition Survey, incident multimorbidity, trajectory modeling, Longitudinal, joint, trajectories