2.1. The Study Design and Participants
This was a multicenter cross-sectional study aiming to assess assess dietary habits, and cardiovascular risk that occur during university attendance (participants were aged 18–30 years old, and should not have any type of physical or mental disorder). A total of 1330 first-year university students from the University of Castilla La Mancha, Spain, were invited to involve in the study, and 1043 (78.42%) accepted to participate. In this report, we used data from a subsample of 499 university students in which bone variables were measured. The young adults included in the data analysis for this study did not differ in age, sex or parental socioeconomic status from the whole sample of young adults participating in the trial.
The Clinical Research Ethics Committee of The Virgen de la Luz Hospital in Cuenca REG: 2016jPI1116 approved the study protocol, and once participants were informed verbally and in writing, they were asked to sign a consent form as a condition to participate in the study. As there were no participants under the legal age in Spain (younger than 18 years), written informed consent was individually obtained for each participant. The signed informed consent documents were recorded. The Ethics Committee approved the study protocol, including informed consent and permissions documents. The methods used for this research have been previously employed [24
2.2. Study Variables
Anthropometry. Stature (cm), and body mass (kg) were measured by using a wall-mounted stadiometer (Seca-222, Vogel & Halke, Hamburg, Germany) and a scale (Seca-770 scale, Vogel & Halke, Hamburg, Germany) respectively. Body mass index (BMI) was calculated as weight in kilograms divided by the square of the height in meters (kg/m2), using the means of the two measurements of weight and height. Waist circumference (cm) was measured at the end of exhalation in the middle point between the costal margin and iliac crest.
Body composition. A dual-energy X-ray absorptiometry (DXA) scanner was used to measure body composition variables (Hologic Discovery Series QDR, Bedford, MA, USA in Toledo and Lunar iDXA, GE Medical Systems Lunar, Madison, WI 53718, USA in Cuenca). In Toledo, the DXA equipment was calibrated by a lumbar spine phantom following the Hologic guidelines. All the DXA scans were analyzed using Physician’s Viewer, APEX System Software Version 3.1.2. (Bedford, MA, USA). In Cuenca, the analyses were performed using enCoreTM 2008 software version 12.30.008. DXA equipment accuracy was checked daily before each scanning session using the GE Lunar calibration phantom, as recommended by the manufacturer. Participants were scanned in the supine position in the middle of the platform A trained researcher performed all scans at high resolution, following the same protocol. Body fat percentage calculated as total fat mass divided by weight, total lean mass (kg), total bone mineral content (BMC) (which is the amount of bone mineral in a specific area, measured in g) and total areal bone mineral density (aBMD) (which is the bone mineral content divided by the bone scanned area, measured in g/cm2 thus, conceptually it is the ratio of BMC to bone size) were calculated for each individual from the whole-body scan. After that, as two different DXA devices were used to measure body composition variables, in order to achieve measurement ready to be included in a single analysis, z scores according to the device were calculated and used in all analyses, thus controlling the variability due to the measurement device.
Physical fitness variables were assessed after a 4-min warm-up consisting in calisthenic exercises and static stretching, and included the following:
Muscular strength: The handgrip test was used to measure upper body strength using a TKK 5401 Grip- DW digital dynamometer with adjustable grip (Takeya, Tokyo, Japan). The average of four measurements (two with the right hand and two with the left hand) was reported in kilograms.
Cardiorespiratory fitness (CRF): The Course Navette test (20-m shuttle run test) was evaluated. The participants were asked to run between two lines set 20 m apart by following the pace of the audio signals produced from a CD player. The starting speed was 8.5 km∙h−1 and was increased by 0.5 km∙h−1 each minute. The participants were equally encouraged to continue the test until they reached maximal effort. The test finished when the participants stopped due to fatigue or when they failed to reach the line two successive time.
Nutrients: The Food-Frequency Questionnaire (FFQ) [26
] was used to determine the total consumption of proteins, calcium, magnesium, phosphorus, potassium, and vitamin D. This validated questionnaire with 9 levels of intake frequencies (never or almost never, between 1 and 3 times per month, once per week, 2–4 times per week, 5–6 times per week, once per day, 2–3 times per day, 4–6 times per day, and more than 6 times per day), included 137 auto reported items for consumption frequency over last year. Spanish food composition tables [27
] were used to compute nutrient and energy intakes.
2.3. Statistical Analysis
The normality of the distribution of continuous variables was analyzed using both statistical and graphical procedures using Kolmogorov–Smirnov test and normal probability plots, respectively. In order to show the relationships among variables, and considering that age and sex are covariates that influence most of the variables included in our models, we calculated partial correlation coefficients controlling for age and sex among body composition variables (body fat percentage, total lean mass, total BMC and total aBMD), physical fitness parameters (CRF and handgrip strength), and nutrients (calcium, magnesium, phosphorus, potassium, vitamin D and proteins) were calculated.
To identify homogenous groups according to the participants’ body composition and physical fitness, based on the z scores of body fat percentage and handgrip strength, a hierarchical cluster analysis was conducted, which no prior information about the group or cluster membership for any of the individuals, using the Ward’s method, based on a squared Euclidean distance [28
], clusters individuals into a pre-determined number of groups according to the similarity of the values of the selected variables. Because outliers in cluster analysis are recognized as observations belonging to none of the clusters, clusters procedures can be substantially influenced by few outliers; thus, values of more than three standard deviations (+3 SD) above or below the mean were removed (three young adults) before the analysis [29
]. Finally, we included afour-cluster solution with the following categories: (i) Fat Unfit (FU), (ii) Unfat Unfit (UU), (iii) Fat but Fit (FF), and (iv) Unfat Fit (UF) (Figure 1
Subsequently, ANCOVA models tested mean differences in body composition variables, including BMC and aBMD, physical fitness variables, and nutrients related to bone mass. These variables were used as dependent variables in the assessment of the relationship between fat but fit categories and sex (fixed factors) controlling for height, age and sex. Pairwise post hoc multiple comparisons were examined using the Bonferroni test. Finally, a sensitivity analysis was conducted including body fat percentage and CRF levels z scores as cluster variables (Appendix A
). The methodology for this cluster analysis was the same as described above.
Statistical analyses were performed using IBM SPSS Statistics v.24 0 (IBM Corp., Armonk, NY, USA). Statistical significance was set at 0.05.