Population-based appendicular skeletal muscle mass equation: Weak association between predicted and bioimpedance-derived changes during stroke and hip-fracture rehabilitation.

BACKGROUND: A population-based appendicular skeletal muscle mass (ASMM) equation was developed against dual-energy X-ray absorptiometry (DXA), primarily in community populations. We examined agreement with repeated clinical bioelectrical impedance analysis (BIA)-derived estimates in older rehabilitation inpatients. METHODS: We analyzed 554 patients after stroke and 296 after hip fracture, without repeat admissions. Standardized BIA was performed within 7 days after admission and within 7 days before discharge; nurse-measured anthropometry was obtained within 7 days of BIA. The equation was applied without refitting, with conventional BIA-derived ASMM as the primary comparator. We examined correlations and mean differences in within-person change and sensitivity analyses for data quality, incomplete follow-up, and comparator definition. RESULTS: Admission correlations were 0.899 and 0.839 after stroke and hip fracture, respectively. Change correlations were 0.198 (95% CI, 0.082-0.309) and 0.227 (0.078-0.388), respectively. Mean differences in change (equation minus BIA) were -0.41 kg (95% CI, -0.53 to -0.28) and -0.14 kg (-0.39 to 0.10). Sensitivity analyses for data quality, incomplete follow-up, and comparator definition yielded similarly weak change associations. Internal-coherence analyses did not establish the reliability of BIA-derived change. CONCLUSIONS: Equation-predicted and BIA-derived changes were weakly associated in these selected rehabilitation patients. Population transportability, measurement-method differences, nonconcurrent assessments, and comparator uncertainty could not be disentangled. These findings do not establish longitudinal inaccuracy of the original DXA-based equation.