ABSTRACT Hemoglobin mass (Hbmass) prediction enhance the accessibility and practicality of athletes’ hemoglobin status monitoring, facilitating better performance. Therefore, we aimed to create prediction equations for Hbmass in well-trained endurance athletes (EA), based on easily obtained measures. The population of 220 well-trained EA (40% females, maximal oxygen uptake = 63.4 ± 8.00 mL·kg·min−1) was randomly split for the models’ derivation and validation in 2:1 ratio. Equations to predict total Hbmass (tHbmass) and Hbmass adjusted to fat-free mass (rHbmass) were developed with multivariable linear regression. The models were stratified for five complexity levels with the inclusion of anthropometric, biochemical, and fitness indices. Models for tHbmass (R2 = 0.87–0.92; root-mean-square error [RMSE] = 60.6–76.5 g) outperform the models for rHbmass (R2 = 0.28–0.58; RMSE = 1.00–1.26 g·kg−1). During internal validation, 9 of 10 of equations accurately predicted tHbmass (0.11 ± 54.7–54.8 ± 45.5 g; p = 0.18–0.99) and only 1 model differed significantly (p = 0.03). There were also no significant differences between observed and predicted values in 8 of 10 of equations for rHbmass (0.1 ± 1.4–1.0 ± 0.1 g·kg−1; p = 0.07–0.65) and 2 models showed significant differences (p = 0.01–0.04). Models present moderate-to-high accuracy. Equations are precise enough to provide complementary data in the epidemiology of diseases with abnormal hemoglobin values, antidoping policy or talent identification. However, they should not substitute direct testing of Hbmass in EA.
Read full abstract