Changes in physical activity and adiposity with all-cause, cardiovascular disease, and cancer mortality

Excess adiposity is linked to higher risks of premature mortality and noncommunicable diseases such as cardiovascular disease (CVD) and cancer. Physical inactivity similarly increases risks of adverse health outcomes, including CVD and premature mortality. Prior studies assessing joint associations of baseline physical activity and adiposity with mortality suggest that while physical activity may attenuate mortality risk, it does not always eliminate the elevated risk associated with excess adiposity. Most evidence relies on single baseline assessments, which is problematic because physical activity and adiposity change over time. Single measurements are susceptible to nondifferential misclassification and regression dilution bias, obscuring the distinct effects of each risk factor. Recent analyses (e.g., UK Biobank) indicate that positive changes in one risk factor can attenuate the adverse effects of negative changes in the other. Guideline bodies (WHO 2020; Physical Activity Guidelines for Americans) have highlighted insufficient evidence on longitudinal relationships of changes in physical activity and adiposity with mortality, and on whether these associations vary by ethnicity. This is particularly important in Asian populations, who may have higher cardiometabolic risk at a given adiposity level, and where cancer is the leading cause of death. This study investigates how changes in physical activity and adiposity relate to all-cause, CVD, and cancer mortality in a large Taiwanese cohort.

Baseline-only studies often report either independence of physical activity and obesity effects on mortality or partial attenuation of obesity-related risk by higher activity, but such designs cannot account for within-person changes over time, leading to potential bias. A recent UK Biobank study suggested that improvements in physical activity or adiposity could mitigate the adverse effects of worsening in the other factor. International guidelines (WHO 2020; US guidelines) note insufficient longitudinal evidence, particularly among non-Western populations. Asian cohorts demonstrate higher CVD and certain cancer risks at lower adiposity levels, implying that standard adiposity cut-offs may not optimally capture risk in these populations. This study addresses these gaps by using repeated measures of physical activity and multiple adiposity indicators (BMI, waist circumference, and body fat percentage) in an East Asian cohort.

Design and cohort: Prospective longitudinal analysis using the Taiwan MJ Cohort. Adults (≥18 years) undergoing routine health screening between 1998 and 2013 were eligible if they had at least one re-examination ≥2 years after baseline and complete data. Exclusions: underweight at baseline (BMI <18.5 kg/m²), prevalent CVD or cancer at baseline or re-examination, and deaths within 2 years after re-examination to reduce reverse causation. Assessments: At each visit, trained staff conducted standardized physical exams; fasting blood samples were collected; questionnaires captured lifestyle/health behaviors (physical activity, diet, smoking, alcohol, sleep). Hypertension and diabetes were identified via medical history or measured criteria (hypertension ≥140/90 mmHg; fasting glucose ≥126 mg/dL for diabetes). Physical activity: Self-reported activities (e.g., gardening, basketball, swimming) were classified by intensity using the 2011 Compendium (light 2.5 METs, moderate 4.5, medium-vigorous 6.5, high-vigorous 8.5). Weekly MET-hours were computed (intensity × duration). Categories around WHO guidelines: inactive (<1 MET-h/week), insufficient (≥1 to <7.5 MET-h/week), sufficient (≥7.5 MET-h/week). Change between baseline and re-exam was categorized as decreased (moved down a category), stable, or increased (moved up a category). To ensure consistency across waves, analyses used assessments with the same questionnaire version. Adiposity indicators: Height and weight measured with a Nakamura KN-5000A auto-anthropometer; waist circumference (WC) measured at the midpoint between the last rib and iliac crest. Body fat percentage (BF%) measured by bioelectrical impedance (InBody Co., Ltd.). BMI categories (WHO Western Pacific): healthy (≥18.5 to <23 kg/m²), overweight (≥23 to <25), obese (≥25). BF% categories based on sex-specific distributions anchored to BMI strata: high (≥25.2% men; ≥34.8% women), moderate (<25.2% to ≥21.6% men; <34.8% to ≥31.0% women), low (<21.6% men; <31.0% women). WC categories (WHO/IDF): high (≥85 cm men; ≥80 cm women), moderate (<85 to ≥75 cm men; <80 to >70 cm women), low (<75 cm men; ≤70 cm women). Adiposity change categorized as decreased, stable, or increased (movement across categories). Joint exposure: Participants classified into nine groups (3 physical activity change × 3 adiposity change) for each adiposity indicator. Outcomes and follow-up: Mortality ascertained via National Death file linkage through October 31, 2019. Time-to-event began at re-examination date. Cause-specific deaths: CVD (ICD-9: 410–414, 420–429, 430–438, 390–392, 393–398, 401–405, 440; ICD-10: I20–I25, I10–I15, I01–I02.0, I05–I09, I27, I30–I52, I70, I71) and cancer (ICD-9: 140–208; ICD-10: C00–C97). Statistical analysis: Cox proportional hazards models estimated HRs and 95% CIs for all-cause mortality (ACM). Fine–Gray subdistribution hazard models estimated HRs for CVD and cancer mortality with competing risks. Independent change analyses used insufficient-stable physical activity and overweight/moderate-stable adiposity as referent groups. Joint analyses used stable physical activity and stable adiposity as the referent, with an interaction term to test effect modification. Models used follow-up time as the timescale and adjusted for baseline and re-exam covariates: age, sex, smoking, alcohol, diet (fruit/vegetable intake), education, sleep. Baseline physical activity and adiposity included as covariates in joint models. Sensitivity and ancillary analyses: E-values for unmeasured confounding; negative control outcome (accidents/sequelae) to assess residual bias; multiple imputation by chained equations (five datasets) for missing covariates; additional adjustment for potential mediators (hypertension, diabetes); mutual adjustment of physical activity and adiposity in separate models; analyses restricted to participants ≥40 years at baseline. Analyses conducted in R (rms, survival packages).

- Cohort and outcomes: 116,228 adults (46.2% female); mean 4.6 years between baseline and re-exam; mean 11.9 years to death/censoring; 1,384,723 person-years; 3,838 deaths (607 CVD, 1,777 cancer). - Physical activity changes (independent analyses): Relative to stable-insufficient activity, increasing activity lowered mortality risk similarly to maintaining sufficient activity. All-cause mortality HRs: inactive→increased 0.85 (0.76, 0.95); insufficient→increased 0.85 (0.74, 0.96); sufficient-stable 0.80 (0.71, 0.92). CVD mortality HRs: insufficient→increased 0.72 (0.55, 0.93); inactive→increased 0.78 (0.63, 0.94); sufficient-stable 0.71 (0.58, 0.88). Decreasing activity raised risk (e.g., insufficient→decreased ACM HR 1.39 [1.18, 1.65]; CVD HR 1.77 [1.21, 2.59]). - Adiposity changes (independent analyses): Relative to stable-overweight/moderate adiposity, increasing adiposity increased risk, especially for CVD (e.g., BMI overweight→increased CVD HR 1.38 [1.13, 1.66]; BF% moderate→increased 1.32 [1.11, 1.53]; WC moderate→increased 1.27 [1.05, 1.49]). Decreasing adiposity attenuated but did not eliminate risk; only maintaining healthy adiposity at both timepoints showed lower risk (e.g., BMI healthy-stable CVD HR 0.75 [0.61, 0.89]). - Joint changes (all-cause mortality): Increased physical activity conferred lower ACM regardless of adiposity change. For WC, increased PA with decreased/stable/increased WC yielded HRs 0.57 (0.48, 0.67), 0.73 (0.66, 0.80), and 0.83 (0.72, 0.97), respectively. Decreasing adiposity attenuated the higher risk from decreased activity (e.g., BF% decreased activity + decreased BF% HR 1.13 [0.95, 1.35] vs reference). - Joint changes (CVD mortality): Strongest benefits when activity increased and adiposity decreased: BMI 0.44 (0.25, 0.76); BF% 0.55 (0.33, 0.89); WC 0.49 (0.29, 0.82). Increased PA mitigated or reversed risks from increased adiposity (e.g., increased PA + increased BF% HR 0.59 [0.38, 0.95]). Stable BMI or BF% did not mitigate risks from decreased activity. - Joint changes (cancer mortality): Similar patterns; increased activity with decreased BMI HR 0.58 (0.44, 0.77) or stable BMI HR 0.81 (0.69, 0.94) associated with lower cancer mortality. - Interaction: Physical activity by adiposity interaction significant for ACM (p<0.01), CVD (p=0.04), and cancer (p=0.03) in joint analyses. - Robustness: E-values suggested substantial unmeasured confounding would be needed to nullify associations (e.g., ≥3.0–4.0-fold). Negative control outcomes did not mimic primary association patterns. Sensitivity analyses (≥40 years only; multiple imputation; adjustment for hypertension/diabetes; mutual adjustment) yielded similar results.

This study demonstrates that changes in physical activity and adiposity jointly influence mortality, with physical activity changes primarily driving the association. Increasing physical activity from inactive or insufficient levels confers reductions in all-cause, CVD, and cancer mortality similar in magnitude to maintaining sufficient activity over time. Conversely, decreasing activity increases mortality risk, and reductions in adiposity alone generally do not offset this elevated risk. Adiposity improvements attenuated, but rarely eliminated, mortality risk compared to stable-overweight/moderate reference categories, whereas maintaining healthy adiposity at both timepoints was protective. The strongest benefits occurred when both risk factors improved, particularly for CVD mortality. Findings suggest that health benefits from increasing physical activity may manifest more rapidly than those from adiposity reduction, consistent with immediate cardiometabolic effects of activity on blood pressure, lipid profile, and other mechanisms, while benefits from adiposity change may accrue over longer periods. Results contrast with baseline-only studies that often show independence between physical activity and adiposity or partial attenuation by activity; by leveraging repeated measures, this study reduces misclassification and regression dilution bias. The absence of an obesity paradox likely reflects exclusions to minimize reverse causation and careful handling of temporal changes. Given higher adiposity-related risk at lower BMI in East Asian populations, current guideline cut-offs may not optimally capture risk; misclassification could attenuate observed adiposity associations. Overall, prioritizing increases in physical activity—alongside adiposity management—appears critical for immediate risk reduction, especially for CVD.

In a large Taiwanese cohort with repeated assessments, increasing physical activity reduced all-cause, CVD, and cancer mortality risks, and largely drove the beneficial joint associations with adiposity changes. Decreasing adiposity attenuated risk but generally did not offset elevated mortality associated with reduced activity; maintaining healthy adiposity consistently conferred protection. Public health and clinical strategies should emphasize increasing and sustaining physical activity, with adiposity reduction as an important ancillary goal. Future research should refine adiposity cut-offs for East Asian populations, incorporate objective activity measures and fitness, and explore longer-term trajectories to clarify latency and causal pathways.

- Observational design precludes causal inference; residual confounding cannot be fully excluded despite adjustments, E-values, and negative control analyses. - Only two measurements of exposures were available, limiting the ability to model detailed trajectories. - Physical activity was self-reported and subject to recall and social desirability biases, likely biasing associations toward the null; misclassification may vary by age, sex, and socioeconomic status. - Cardiorespiratory fitness was not measured, despite its independent association with mortality. - Reverse causation cannot be entirely ruled out, although underweight participants, those with baseline CVD/cancer, and early deaths were excluded. - Some negative control outcome groups had few events, reducing precision of those estimates. - Potential imprecision or suboptimal relevance of adiposity cut-offs for East Asian populations may have attenuated adiposity–mortality associations.