Cross-validation of 20 anthropometric prediction equations for appendicular muscle mass in older Brazilian women: a cross-sectional study

Abstract

Objective: To test the cross-validation of anthropometric prediction equations for appendicular muscle mass (AMM) in older Brazilian women. Methods: Sixty-seven older women (69.84 ± 5.95 years old) underwent anthropometric measurements. AMM (kg) reference values obtained by dual-energy X-ray absorptiometry (AMMDXA) were compared to 20 anthropometric equations for estimating AMM in older adults. A paired t-test (p > 0.05), standard error of estimate (SEE < 3.50 kg), and r2 > 0.70 confirmed the validity of the equations. The agreement between predictions and the reference was also verified (Bland-Altman). Results: Four American equations and one Mexican equation were not statistically different from AMMDXA (p > 0.05) but did not present suitable r2 values for validation. The American equation from the National Health and Nutrition Examination Survey (NHANES), AMM (kg) = (-0.04 × age [years]) + (0.46 × calf circumference [cm]) + (0.32 × arm circumference [cm]) + (0.11 × thigh circumference [cm]) – (0.27 × body mass index [BMI, kg/m2]) + (0.07 × waist circumference [cm]) – 13 119) showed the best performance (r2 = 0.64; SEE = 3.24 kg), with minimal mean difference (0.26 kg), no heteroscedasticity for extreme values, and with high agreement with the Brazilian sample (-3.90 to 3.40 kg). Conclusion: When specific equations for a given population are not available, the use of generic equations of greater sample representativeness with scientifically and reliably analyzed data is allowed.