Heterogeneous associations between domain-specific physical activity and sarcopenia risk in diabetic adults: A causal forest analysis.

Sarcopenia prevalence among diabetic adults in this sample was 18.46%, with notably higher rates among individuals with obesity, those of Mexican American ethnicity, and those with lower educational attainment.

• Total sample size was 1,230 participants with diabetes.
• Sarcopenia prevalence was 18.46% overall.
• Subgroups with higher sarcopenia rates included individuals with obesity, Mexican American ethnicity, and lower educational attainment.
• The study used a cross-sectional design.

Occupational moderate-to-vigorous physical activity demonstrated an independent inverse association with sarcopenia risk after full covariate adjustment.

• β = -0.020 (95% CI: -0.034 to -0.006, p = 0.007) for occupational moderate-to-vigorous physical activity.
• β denotes multivariable linear regression coefficients for the continuous skeletal muscle mass index (SMI) outcome.
• Negative values indicate lower skeletal muscle mass index and higher sarcopenia risk.
• This association was observed following full covariate adjustment using progressive multivariable linear regression.

Leisure-time moderate-to-vigorous physical activity also demonstrated an independent inverse association with sarcopenia risk after full covariate adjustment.

• β = -0.018 (95% CI: -0.032 to -0.003, p = 0.018) for leisure-time moderate-to-vigorous physical activity.
• This association was independent of occupational physical activity and other covariates.
• Both occupational and leisure-time domains were analyzed separately and found to be independently associated with sarcopenia risk.

Dose-response analyses using restricted cubic splines identified activity thresholds beyond which progressively more favorable associations with sarcopenia risk were apparent.

• Restricted cubic spline analyses were used to characterize dose-response relationships between physical activity and sarcopenia risk.
• Activity thresholds were identified beyond which associations with lower sarcopenia risk became more pronounced.
• The analyses examined both continuous skeletal muscle mass index scores and binary sarcopenia status as outcomes.

Causal forest estimation indicated that total physical activity exhibited the most pronounced inverse association with sarcopenia risk among domain-specific activity measures.

• Mean conditional average association effect (CAAE) for total physical activity was -0.0547.
• Results were reported as conditional average association effects rather than treatment effects, given the observational design.
• The causal forest algorithm was used to estimate individual-level heterogeneity in association magnitude.
• Total physical activity showed a stronger association than either occupational or leisure-time physical activity alone.

Exploratory subgroup analyses suggested potential heterogeneity in association magnitude, with directionally consistent point estimates indicating more pronounced inverse associations in certain subgroups.

• Point estimates were directionally consistent with more pronounced inverse associations among adults aged ≥51 years, females, and individuals with obesity.
• Stratified subgroup analyses used interaction terms to assess consistency across age, sex, body mass index, and race/ethnicity.
• The authors described these patterns as 'preliminary' and 'hypothesis-generating' given the cross-sectional design.
• The authors noted these patterns 'would benefit from prospective verification before more definitive conclusions are drawn.'

The cross-sectional design of the study precludes any inference about temporality or causality between physical activity and sarcopenia risk.

• The authors explicitly state the design 'precludes any inference about temporality or causality.'
• The causal forest findings were reported as 'conditional average association effects rather than treatment effects.'
• The authors called for 'future longitudinal studies and randomized controlled trials' to verify findings.
• Subgroup patterns were characterized as 'hypothesis-generating.'