Assessing body composition in persons with obesity, and in particular, the excess of fat
mass and the possible reduction of muscle mass, is important to define the phenotypic
manifestation of obesity (estimating the risk of dysmetabolic, cardiovascular, and
functional complications), and to determine a better treatment approach. Dual X-ray
absorptiometry (DXA) is a mature technology for assessing body composition with major
advances in the technology over the past three decades. DXA is a validated tool to
investigate body composition phenotypes, as it reliably assesses whole-body and regional
bone mineral content, fat mass and lean mass. Unfortunately, it is not always available
in all settings where instead Bio-Impedance Analysis (BIA) (which has lower costs and
greater convenience of use) is commonly used to estimate body composition starting from
electrical resistance and reactance data.
Regrettably, the two methods often give non-superimposable results and studies have been
carried out to predict, from BIA, values commonly obtainable only with DXA. In
particular, different studies estimated the appendicular lean mass from BIA, which
represents an important parameter for the evaluation of sarcopenia and is correlated with
its functional limitations. For example, a post hoc analysis of the PROVIDE study was
aimed in particular at assessing the level of agreement between BIA- and DXA-derived soft
tissue ratios as indicators of limb tissue quality and at developing and cross-validating
new BIA equations for predicting appendicular soft tissue [fat mass (FM) and appendicular
lean mass (ALM)] in older Caucasian adults with physical function decline using both the
Hologic Horizon and GE Lunar DXA systems as reference methods.
METHODS:
This study is based on baseline data (anthropometric, BIA, and DXA) collected in
pre-existing datasets. In particular
the Sapienza dataset which derived from a study aimed at investigating the
association between markers of insulin sensitivity and SO defined by three novel
body composition models will be used to develop BIA equations predicting
appendicular soft tissue masses;
datasets from different studies and in particular from the BIA International Dataset
Project will be used to validate the BIA equations assessing the agreement between
BIA- and DXA-derived soft tissue estimation
STUDY PARAMETERS:
-Anthropometry: anthropometric parameters should have been measured in accordance with
validated and standardized methodologies.
The anthropometric parameters of interest are body mass, stature, waist circumference,
calf circumference, arm circumference, and triceps skinfold thickness, limb length.
-Dual energy X-ray absorptiometry: all participants should have been scanned using a fan
beam whole body DXA device (Hologic Bedford, Massachusetts, USA; Lunar Prodigy, GE
Healthcare). Daily calibration of the densitometers should have been performed following
the instructions provided by the manufacturer.
Since measurements vary among instruments from different manufacturers, calibration
equations will be used to address these issues and improve the agreement between devices.
The body components of interest are total fat mass (FM), total lean mass (LM), ALM (sum
of the lean mass in the limbs), FM (sum of the fat mass in the limbs), and the ratio of
ALM to FM.
-Bioelectrical impedance analysis: After overnight fasting and bladder voiding,
bioelectrical impedance analysis should have been performed with participants lying
supine (with their limbs slightly away from their body; active electrodes should have
been placed on the right side on conventional metacarpal and metatarsal lines, recording
electrodes in standard positions at the right wrist and ankle) or in vertical position
(barefoot, stepping onto the electrodes embedded into the scale and grasping the
electrode-embedded handles). At each location, a whole-body tetrapolar BIA device
operating at a weak alternating electrical current of 500 µA to 1 mA and a single
frequency of 50 kHz should have been used to measure the voltage drop across body
tissues.
The electric parameters of interest are resistance (R: restriction of current flow),
reactance (Xc: capacitance of cell membranes and tissue interfaces), and phase angle
(PhA).
The information about BIA devices will be recorded since raw R and Xc values may not be
not comparable.
Due to the significant differences found in different studies when comparing vertical to
supine position, the results obtained with the two methodologies will be analysed
separately.
With reference to the limitation reported by the PROVIDE study authors (i.e. the absence
of a direct measurement of extracellular water), the raw data detected through
multifrequency bioimpedance devices will also be used, where available. Specifically, the
values of impedance and resistance measured at a frequency of 5 kHz will be included;
furthermore, where available, it would be optimal to analyze data measured at the
following frequencies; 1, 2, 5, 10, 50, 100, 200, 250 and 500 kHz.
STATISTICS:
Data will be analyzed by using IBM® SPSS® Statistics version 25. The data will be
presented as frequency (percent) and mean ± SD for qualitative and quantitative
variables, respectively. The Shapiro-Wilk test will be used to evaluate if the data are
normally distributed. Comparison of continuous variables will be performed using
parametric or non-parametric tests depending on whether the distribution is normal or
not. The chi-square test will be used to check whether the frequencies occurring in the
sample differ significantly from the expected frequencies. The cut-off for statistical
significance will be set at p<0.05.
Preliminary equations, using DXA-derived appendicular lean and fat mass as the dependent
variables, and age, gender, BMI, weight, impedance index, and reactance as independent
variables, will be developed using a stepwise multiple linear regression approach. Only
significant regressors of appendicular soft tissue masses will be considered in the
equations.
Model performance fit will be assessed using multiple correlations (R2) and standard
errors of the estimate (SEE). For each of the appendicular soft tissue components, the
model with the lowest standard error of the estimate will be used in the cross-validation
analysis.
The individual and body composition data from the cross-validation samples will be
imputed into the developed equations to assess their accuracy. The statistics for
cross-validation includes mean difference, limits of agreement, and root mean squared
error.
Additionally, the agreement between ALM_BIA estimated in our sample, ALM_SERGI,
ALM_Provide, and ALM_KYLE will be assessed using regression analysis.
Finally, the agreement between the ALM/FM-ratios estimated by DXA and by BIA will be
evaluated using Bland and Altman analysis.
