This study was approved by the local Research Ethics Committee from State University of Campinas (UNICAMP) (CAAE: 79718417.0.0000.5404). All participants and their parents or legal guardians signed an informed consent form. All of the procedures were conducted per the Helsinki Declaration for Human Studies.
Design of the Study
The study was cross-sectional; the evaluations were conducted in November 2018.
Inclusion criteria consisted of male basketball and volleyball athletes, with ages ranging from 14 to 17 years, who participated for 3 years or more in official competitions promoted by the basketball and volleyball confederations. All athletes had to train over 15 hours a week.
Exclusion criteria consisted of female athletes, athletes that presented injuries at the time of the evaluations, and those that did not participate in one or more stages of the study.
All athletes were submitted to tests related to body composition and bone mass evaluation. Furthermore, the athletes were required to answer a questionnaire related to sedentary behavior and dietary intake, which are described below.
Peak of Height Velocity (PHV)
The PHV was assessed according to the formula (for boys) [22]: PHV = -7.999994+0.0036124*(age*stature)), where R2 = 0.896, and SEE = 0.542.
Body Composition
The body mass (kg) was determined using a digital scale. Stature (cm) was determined (cm) using a vertical stadiometer. Body mass index (BMI kg/m2) was calculated.
Body composition in fat mass (FM) and lean mass (LM) was assessed by dual-energy X-ray absorptiometry (DXA) (iDXA - GE Healthcare Lunar, Madison, WI, USA) and version 13.6 software enCore ™ 2011 (GE Lunar Healthcare). The reproducibility of the variables estimated by the DXA was determined by the coefficient of variation (CV%) and by the technical measurement error (TEM), based on a test-retest study performed with 23 subjects (adolescents and young adults) using the formula:
Where: D is the difference between the two measurements and n is the sample size. CV% were 0.74% and 0.26% for FM and LM, respectively; and TEM were 0.25 kg and 0.25 kg for FM and LM respectively.
Bone Mineral Density (BMD), Bone Mineral Content (BMC) and Geometry Parameters
BMD (cm2) and BMC (g) were obtained using the iDXA equipment, with acquisition, positioning, and outcome analysis performed according to the International Society of Clinical Densitometry [23]. These parameters were obtained by scanning: total body less head (TBLH); lumbar spine (L1-L4); and right femoral neck (neck).
The bone geometry, i.e., hip structural analysis data was obtained with the Advanced Hip Assessment software [version 13.6 in software enCore ™ 2011 (GE Healthcare Lunar)]. This software automatically derives geometric properties of the proximal femur, such as a) Cross-sectional moment of inertia [CSMI (mm2)], which is an estimate of resistance and weight forces directed along the length of the bone in a cross-section; b) Transverse cross-sectional area of the femoral neck [CSA (mm2)], which measures the resistance to loads directed along the bone axis; c) Section modulus [Z (mm3)], which is a measure of the maximum bending strength in a cross-section and, d) Femoral strength index (FSI), i.e., an indicator of the risk of a fracture caused by severe fall on the trochanter in relation to the CSA [24].
Therefore, the parameters BMD, BMC, L1-L4-BMD, L1-L4-BMC, Neck-BMD, neck-BMC, CSMI, CSA, Z, and FSI were obtained.
Sedentary Lifestyle Behavior
The sedentary behavior was determined through a questionnaire developed and validated [25]. The questionnaire determines the average adolescent’s sedentary activities during the week (Adolescent Sedentary Activity Questionnaire - ASAQ) and has been validated for the Brazilian population [26]. The Brazilian version of ASAQ consists of 13 items, divided into five categories, in which participants report their time spent in sedentary activities (sitting time during school, sitting time during other activities such as screen time at home including video games, etc.), on each day of the week and during a typical weekend. For this study, the results were expressed in minutes considering the seven days of the week (i.e. including the weekend behavior as well). The Brazilian National Education Guidelines and Law establishes a maximum of 240 minutes of sitting time in school [27].
Dietary Intake Assessment
Information on all foods and beverages were obtained by a nutritionist through a 24-hour food recall [28] and a semi-quantitative food frequency questionnaire [29,30]. The semi-quantitative FFQ, with usual portions of foods rich in calcium, was used only for the estimation of calcium intake, due to its important role in bone health. The estimation of the other nutrients was based on the 24-hour dietary recall. All the nutrients estimations were calculated using the Nutwin software. Nutrient adequacy was assessed using the recommended dietary guidelines. The daily recommendation was based on intake for ages 14-18 and 19-30 years and are as follows: 0.85g and 0.80g protein/kg [31], 1300mg of calcium [32], 5mcg of vitamin D [32], 1250mg of phosphorus, 1500mg sodium and 410mg and 400mg of magnesium [32].
Statistical analyses
Descriptive statistics are presented as means and 95% credible intervals. A series of multilevel linear regression models were fitted to explore whether there was variation for players' body composition, bone parameters, diary nutrient intake and sedentary behavior by sport. Hence, we assumed players (level-1) nested by sports, i.e., basketball and volleyball (level-2). A varying-intercept null model was used initially to measure the proportion of total variance which fell between-sport (i.e., variance partition coefficient). For all outcomes we observed variance partition coefficients smaller than 0.05, implying that there was no substantial variation between players by sport. Hence, the relationship between body composition and bone parameters with sedentary behavior was explored using single-level linear regression. We standardized all variables, allowing the slope of the linear regression to be interpreted similarly as a correlation or partial correlation. All models were fitted using Bayesian methods, which were implemented using R statistical language, with the “brms” package [33] which calls Stan [34]. Since we standardized all outcomes, we used weakly informative priors for population-level effects and group-level, normal priors (0,1). We ran four chains for 2,000 iterations with a warm-up length of 1,000 iterations for each model. The convergence of the Markov chains was examined with trace plots and validity of the models inspected using posterior predictive checks.
The multilevel models explicitly consider variation between players grouped by sport (level-2 unit). Hence, our estimates account for the potential influence of nesting by sport. Since there was no substantial variation accounted for sport, we report our estimates for the sample.