Research report: Accelerometer-derived moderate-to-vigorous-intensity physical activity: sex differences in health benefits and health equivalence between the sexes: a prospective cohort study

Using accelerometer data from 86,196 middle-aged adults in the UK Biobank, researchers found that the associations between moderate-to-vigorous-intensity physical activity (MVPA) and all-cause, cardiovascular, and cancer mortality differed significantly between the sexes. For cardiovascular mortality, the estimated male-to-female equivalence ratio was 1.37, meaning men needed roughly 37 percent more weekly MVPA than women to achieve the same modeled reduction in risk. For all-cause and cancer mortality the pattern reversed, with ratios of 0.90 and 0.71 respectively, indicating smaller male requirements for those outcomes.

The study, led by researchers at Shengjing Hospital of China Medical University, addressed a gap in current physical activity guidance. The World Health Organization recommends that all adults accumulate 150 to 300 minutes of moderate-intensity or 75 to 150 minutes of vigorous-intensity activity per week, a threshold that is applied uniformly regardless of sex. The authors noted that physiological differences between the sexes have long been recognized, and previous studies have hinted at sex-specific dose-response relationships for coronary heart disease, sarcopenia, bone mineral density, and mortality, but the picture across a broad range of health outcomes remained incomplete.

To measure activity objectively, the team drew on participants who wore an Axivity AX3 wrist-worn triaxial accelerometer for seven days between February 2013 and December 2015. Of 103,617 participants with raw data, 86,196 remained after exclusions for withdrawn consent, recording errors, poor calibration, fewer than three valid wear days, extreme MVPA values, missing covariates, or absent hospital records. MVPA was identified with a machine-learning classifier trained on the CAPTURE-24 dataset, using a random forest to classify 30-second windows and a hidden Markov model to smooth the temporal sequence of behaviors. Behaviors at or above three metabolic equivalents were counted as MVPA, and weekly duration was estimated by multiplying average daily minutes by seven. Participants logging at least 150 minutes per week were classified as active.

Over a median follow-up of 7.93 years, 5,195 deaths occurred. Being active was associated with lower mortality in both sexes, but the magnitudes differed. In men, activity was linked to a 30 percent lower risk of all-cause mortality (hazard ratio 0.70), a 24 percent lower risk of cardiovascular mortality (0.76), and a 22 percent lower risk of cancer mortality (0.78). In women, the corresponding reductions were 19 percent (0.81), 21 percent (0.79), and 12 percent (0.88). Likelihood ratio tests comparing models with and without sex-by-MVPA interaction terms found significant differences for all three mortality outcomes, with P values of 0.032 for all-cause mortality and below 0.001 for cardiovascular and cancer mortality.

Dose-response analyses using restricted cubic splines, with knots placed at the 10th, 50th, and 90th percentiles of weekly MVPA (43.2, 230.4, and 590.4 minutes), showed pronounced gradients for men across all mortality outcomes, with modeled risk reductions reaching 50 percent or higher. Women showed a similar gradient for cardiovascular mortality but only modest gradients for all-cause and cancer mortality. From these curves, the researchers calculated health equivalence: for any given percentage of risk reduction between 5 and 35 percent, they extracted the sex-specific MVPA duration needed and divided the male duration by the female duration. The median ratios were 0.90 (95 percent confidence interval 0.87 to 0.93) for all-cause mortality, 1.37 (1.35 to 1.39) for cardiovascular mortality, and 0.71 (0.68 to 0.74) for cancer mortality. The equivalence curve for cardiovascular mortality remained stable across the risk-reduction range, while the curves for all-cause and cancer mortality declined gradually, with a marked downward shift for cancer mortality at risk reductions of 18 percent or more.

The team then extended the approach to disease incidence through a phenome-wide association study covering 447 disease phenotypes drawn from electronic health records using the Phecode Map v1.2, excluding congenital and pregnancy-related codes and retaining only phenotypes with at least 120 events. Here the picture was more cautious. Thirty-five phenotypes showed nominally significant sex-by-MVPA interactions, but none survived correction for the false discovery rate. Among those 35, MVPA was significantly associated with 12 conditions in women only and 4 in both sexes. The four shared associations favored women: activity was linked to a 27 percent lower risk of coronary atherosclerosis in women versus 16 percent in men, a 33 percent versus 18 percent lower risk of other unspecified chronic ischemic heart disease, a 51 percent versus 34 percent lower risk of chronic airway obstruction, and a 23 percent versus 17 percent lower risk of gastroesophageal reflux disease.

Exploratory health-equivalence analyses for these four conditions yielded median male-to-female MVPA duration ratios of 2.02 for coronary atherosclerosis, 2.59 for other unspecified chronic ischemic heart disease, 1.87 for chronic airway obstruction, and 1.73 for gastroesophageal reflux disease, all within the 5 to 35 percent risk-reduction range. In men, risk of coronary atherosclerosis and chronic ischemic heart disease decreased continuously with increasing activity, while risk of chronic airway obstruction and reflux disease declined and then stabilized near 300 minutes per week. In women, the curves for coronary atherosclerosis and chronic airway obstruction were J-shaped, with a turning point near 400 minutes per week. The authors stressed that these incidence-based equivalence estimates rest on dose-response curves for phenotypes whose sex differences were only nominally significant and should be treated as exploratory.

Sensitivity analyses added important nuance. When participants who experienced the analyzed mortality outcome within the first year of follow-up were excluded to limit reverse causation, hazard ratios remained below 1 in both sexes and lower in men, but the sex-by-MVPA interaction remained significant only for all-cause mortality, not for cardiovascular or cancer mortality. In analyses using multiple imputation for missing covariates, all hazard ratios stayed below 1, though the association with cancer mortality in women lost statistical significance and none of the sex-by-MVPA interactions remained significant. Alternative risk-reduction ranges of 5 to 20 percent and 10 to 35 percent produced similar median equivalence ratios, keeping them below 1 for all-cause and cancer mortality and above 1 for cardiovascular mortality.

The findings align with earlier work the authors cite, including a study reporting that men needed 530 minutes of weekly MVPA to achieve a 30 percent reduction in coronary heart disease risk while women needed only 250 minutes, and research on vigorous intermittent lifestyle physical activity showing dose-response relationships in women but not men for major adverse cardiovascular events, myocardial infarction, and stroke. Possible biological explanations include estrogen-related modulation of vascular responses to exercise and sex differences in metabolic adaptations, though the study did not measure these pathways directly.

The authors and outside readers should weigh several limitations. The UK Biobank cohort consists primarily of older British adults of relatively high socioeconomic standing, limiting generalizability. The observational design cannot rule out residual confounding or reverse causation, and most covariates were recorded at recruitment in 2006 to 2010, years before the accelerometer assessment. A single week of accelerometry may not capture habitual activity or long-term changes. The turning points in the dose-response curves were not formally estimated, and potential effect modification by age was not examined. The equivalence estimates, being model-predicted weekly durations at matched hazard-ratio levels, are not fixed conversion factors between men’s and women’s exercise needs. The researchers concluded that replication in independent and more diverse cohorts is needed before the results can inform sex-specific physical activity recommendations.

Subject of Research: Sex differences in the associations of accelerometer-measured moderate-to-vigorous physical activity with mortality and incident diseases in the UK Biobank

Article Title: Accelerometer-derived moderate-to-vigorous-intensity physical activity: sex differences in health benefits and health equivalence between the sexes: a prospective cohort study

Article References: Zheng, Z., Liu, H., Ge, X., Zhang, S., Wang, Y., Wei, Y., Dong, P., Gao, M., Gao, S., & Wei, Y. (2026). Accelerometer-derived moderate-to-vigorous-intensity physical activity: sex differences in health benefits and health equivalence between the sexes: a prospective cohort study. Frontiers in Public Health, 14, 1966467. https://doi.org/10.3389/fpubh.2026.1966467

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1966467

Keywords: physical activity, MVPA, sex differences, UK Biobank, mortality, accelerometry, cardiovascular mortality, cancer mortality, PheWAS, health equivalence, prospective cohort, public health