Body-composition phenotype axes provide additional explanatory information for cardiorespiratory fitness and liver-related metabolic risk within BMI strata.

BACKGROUND: Body mass index (BMI) does not fully characterize heterogeneity in adiposity patterning, lean reserve, and impedance-derived body-composition features. We derived residualized body-composition phenotype axes from routine measurements and examined whether these projected axes provided additional explanatory information for estimated VO₂max and alanine aminotransferase (ALT) within BMI strata. METHODS: Phenotype axes were derived using principal component analysis (PCA) of six variables (BMI, waist-to-height ratio, fat mass index, fat-free mass index, percent body fat, and phase angle) following residualization for age and sex in 23,698 individuals. Robustness and repeatability were assessed. The derivation-defined axes were externally projected into 5,060 adults aged 20-49 years from the National Health and Nutrition Examination Survey. Survey-weighted regression models were used to examine associations between projected phenotype scores and estimated VO₂max from the submaximal treadmill fitness examination (n = 2,721) and alanine aminotransferase (ALT) (n = 4,857) within unified BMI quartile strata, with adjustment for age, sex, and race/ethnicity. Sensitivity analyses included expanded covariate adjustment, false-discovery-rate correction, interaction testing, protocol-adjusted estimated VO₂max models, and ALT clean-sample analyses. RESULTS: Three axes explained 98.6% of the residualized input variance: PC1 represented a composite body-size/adiposity axis, PC2 a phase angle-anchored body-composition contrast axis, and PC3 an FFMI-dominant axis. The phenotype structure was robust, repeatable, and structurally concordant in NHANES. Within BMI strata, selected associations with estimated VO₂max remained after 24-test BH-FDR correction, including PC2 in Q3 (β = 1.82, 95% CI 0.41 to 3.22) and PC3 in Q4 (β = 1.72, 95% CI 0.61 to 2.83). For ALT, higher-BMI strata showed divergent PC2 and PC3 associations that remained after BH-FDR correction; for example, in Q4, PC2 was associated with higher ALT (+6.74%, 95% CI 3.23 to 10.38%), whereas PC3 was associated with lower ALT (-6.41%, 95% CI - 10.15% to -2.52%). Expanded and endpoint-specific sensitivity analyses were generally consistent with the primary interpretation, although selected associations were attenuated. CONCLUSION: Projected body-composition phenotype axes derived from six routine measurements provided additional explanatory information for estimated VO₂max and ALT within BMI strata. This residualized multivariable framework may help characterize within-BMI heterogeneity in clinical nutrition research, but should not yet be interpreted as a clinical classification or decision-making tool.