Explaining Discrepancies Between Total and Segmental DXA and BIA Body Composition Estimates Using Bayesian Regression.

INTRODUCTION/BACKGROUND: Few investigations have sought to explain discrepancies between dual-energy X-ray absorptiometry (DXA) and bioelectrical impedance analysis (BIA) body composition estimates. The purpose of this analysis was to explore physiological and anthropometric predictors of discrepancies between DXA and BIA total and segmental body composition estimates. METHODOLOGY: Assessments via DXA (GE Lunar Prodigy) and single-frequency BIA (RJL Systems Quantum V) were performed in 179 adults (103 F, 76 M, age: 33.6 ± 15.3 yr; BMI: 24.9 ± 4.3 kg/m2). Potential predictor variables for differences between DXA and BIA total and segmental fat mass (FM) and lean soft tissue (LST) estimates were obtained from demographics and laboratory techniques, including DXA, BIA, bioimpedance spectroscopy, air displacement plethysmography, and 3-dimensional optical scanning. To determine meaningful predictors, Bayesian robust regression models were fit using a t-distribution and regularized hierarchical shrinkage "horseshoe" prior. Standardized model coefficients (ß) were generated, and leave-one-out cross validation was used to assess model predictive performance. RESULTS: LST hydration (i.e., total body water:LST) was a predictor of discrepancies in all FM and LST variables (|ß|: 0.20-0.82). Additionally, extracellular fluid percentage was a predictor for nearly all outcomes (|ß|: 0.19-0.40). Height influenced the agreement between whole-body estimates (|ß|: 0.74-0.77), while the mass, length, and composition of body segments were predictors for segmental LST estimates (|ß|: 0.23-3.04). Predictors of segmental FM errors were less consistent. Select sex-, race-, or age-based differences between methods were observed. The accuracy of whole-body models was superior to segmental models (leave-one-out cross-validation-adjusted R2 of 0.83-0.85 for FMTOTAL and LSTTOTAL vs. 0.20-0.76 for segmental estimates). For segmental models, predictive performance decreased in the order of: appendicular lean soft tissue, LSTLEGS, LSTTRUNK and FMLEGS, FMARMS, FMTRUNK, and LSTARMS. CONCLUSIONS: These findings indicate the importance of LST hydration, extracellular fluid content, and height for explaining discrepancies between DXA and BIA body composition estimates. These general findings and quantitative interpretation based on the presented data allow for a better understanding of sources of error between 2 popular segmental body composition techniques and facilitate interpretation of estimates from these technologies.

Semantic and cognitive tools to aid statistical science: replace confidence and significance by compatibility and surprise.

BACKGROUND: Researchers often misinterpret and misrepresent statistical outputs. This abuse has led to a large literature on modification or replacement of testing thresholds and P-values with confidence intervals, Bayes factors, and other devices. Because the core problems appear cognitive rather than statistical, we review some simple methods to aid researchers in interpreting statistical outputs. These methods emphasize logical and information concepts over probability, and thus may be more robust to common misinterpretations than are traditional descriptions. METHODS: We use the Shannon transform of the P-value p, also known as the binary surprisal or S-value s = -log2(p), to provide a measure of the information supplied by the testing procedure, and to help calibrate intuitions against simple physical experiments like coin tossing. We also use tables or graphs of test statistics for alternative hypotheses, and interval estimates for different percentile levels, to thwart fallacies arising from arbitrary dichotomies. Finally, we reinterpret P-values and interval estimates in unconditional terms, which describe compatibility of data with the entire set of analysis assumptions. We illustrate these methods with a reanalysis of data from an existing record-based cohort study. CONCLUSIONS: In line with other recent recommendations, we advise that teaching materials and research reports discuss P-values as measures of compatibility rather than significance, compute P-values for alternative hypotheses whenever they are computed for null hypotheses, and interpret interval estimates as showing values of high compatibility with data, rather than regions of confidence. Our recommendations emphasize cognitive devices for displaying the compatibility of the observed data with various hypotheses of interest, rather than focusing on single hypothesis tests or interval estimates. We believe these simple reforms are well worth the minor effort they require.

Predicting Adaptations to Resistance Training Plus Overfeeding Using Bayesian Regression: A Preliminary Investigation.

Relatively few investigations have reported purposeful overfeeding in resistance-trained adults. This preliminary study examined potential predictors of resistance training (RT) adaptations during a period of purposeful overfeeding and RT. Resistance-trained males (n = 28; n = 21 completers) were assigned to 6 weeks of supervised RT and daily consumption of a high-calorie protein/carbohydrate supplement with a target body mass (BM) gain of ≥0.45 kg·wk-1. At baseline and post-intervention, body composition was evaluated via 4-component (4C) model and ultrasonography. Additional assessments of resting metabolism and muscular performance were performed. Accelerometry and automated dietary interviews estimated physical activity levels and nutrient intake before and during the intervention. Bayesian regression methods were employed to examine potential predictors of changes in body composition, muscular performance, and metabolism. A simplified regression model with only rate of BM gain as a predictor was also developed. Increases in 4C whole-body fat-free mass (FFM; (mean ± SD) 4.8 ± 2.6%), muscle thickness (4.5 ± 5.9% for elbow flexors; 7.4 ± 8.4% for knee extensors), and muscular performance were observed in nearly all individuals. However, changes in outcome variables could generally not be predicted with precision. Bayes R2 values for the models ranged from 0.18 to 0.40, and other metrics also indicated relatively poor predictive performance. On average, a BM gain of ~0.55%/week corresponded with a body composition score ((∆FFM/∆BM)*100) of 100, indicative of all BM gained as FFM. However, meaningful variability around this estimate was observed. This study offers insight regarding the complex interactions between the RT stimulus, overfeeding, and putative predictors of RT adaptations.