Study design and sampling
Data from this cross-sectional study resulted from LabMed Physical Activity Study (Longitudinal Analysis of BioMarkers and Environmental Determinants of Physical Activity).
For the present study we assessed a sub-sample of 250 adolescents who were willing to participate, aged between 13 and 18 years, from schools in Braga district, with urinary excretion data collected across two time blocks, September 2012–April 2013 and September 2013–April 2014, excluding the warmer months of the year.
Quality control was used by calculating 24-hour urinary creatinine excretion in relation to body weight according to age group  and incomplete urine collections were repeated; subjects that had felt ill, had reported renal problems or took drugs in the day of collection were also not included for the present analysis (rejected n = 50, 20 %). Therefore, the final sample comprised 200 adolescents (82 boys) with valid urine collection and corresponding dietary recall.
The study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human participants were approved by the Ethical Commission of University of Porto. Written informed consent was obtained from all participants and caregivers.
Urine sampling and hydration markers
Participants and caregivers received oral and written instructions on how to collect complete 24-hour urine samples. All participants were instructed to discard the first morning void and to collect all urine over the following 24-hour including the first void on the following morning, the time of the start and finish collection was asked to be recorded in a questionnaire. During the collection period, subjects were asked to store collected urine in a cool place. All samples were sent to a certified laboratory to be analyzed sodium and potassium by indirect ion-selective electrodes methodology (Siemens Advia 1800), creatinine by Jaffé reaction (Siemens Advia 1650) and osmolality by sum of solute particles (exocinase method and Siemens Advia 1650/1800 equipment). Urine samples were analyzed for urinary creatinine (mg/day), and urinary sodium (mEq/day); sodium excretion was reported in mEq/day, however, for comparative purposes, it was converted to mg/day by using the molecular weight of sodium. Estimated salt intake was calculated from analyzed (24 h urine) sodium excretion (1 g salt = 393 mg sodium).
Hydration status was assessed using urinary markers, namely 24-hour urinary volume (mL), 24-hour Uosm (mOsm/kg), and FWR (ml/24 h) as determined and described previously . Since concentration ability decreases not until age of 20 [6, 22], 830 mosm/1000 g is the mean maximum urine osmolality used to establish FWR in adolescents . Positive values of FWR indicate euhydration, negative values the risk of hypo-hydration . Risk of hypo-hydration correspond to the Uosm mean - 2 SD of maximum Uosm, and euhydration to Uosm between the mean −2 SD of maximum Uosm and the mean + 2 SD of minimum Uosm .
A 24 h dietary recall referring to the day of the urine collection was collected by trained interviewers using photo book and household measures to quantify portion sizes . Energy and nutritional intake were estimated using an adapted Portuguese version of the nutritional analysis software Food Processor Plus (ESHA Research Inc., Salem, OR, USA). The nutrient content of basic food was taken from standard nutrient tables, whereas the content of commercial food, e.g. pizza, ready-to-eat-food was derived from labelled ingredients and nutrients. Water from solid and fluid foods (total water in g per day), recorded from the 24 h dietary recall, was calculated using data from the Food Processor Plus (ESHA Research Inc., Salem, OR, USA).
Height was measured to the nearest millimeter in bare or stocking feet with the adolescents standing upright against a stadiometer (Crymych, Pembrokeshire, UK). Weight was measured to the nearest 0.10 kg, with adolescents lightly dressed using portable electronic weight scale (Tanita Inner Scan BC 532, Tokyo, Japan). Body mass index (BMI) was calculated as weight (kg) divided by square height (m2), and participants were classified according to World Health Organization (WHO) BMI reference values , in normal weight, overweight, and obesity. Underweight subject (n = 1) was combined with subjects in the normal weight category, due to the fact that represented a very small proportion of the sample.
Physical activity and participation in sports were measured by means of a short self-report questionnaire that was administered individually . The answers were coded from 1 to 3, 1 representing inactivity or very low activity, 2 moderately intensive or frequent activity, and 3 frequent or vigorous activity. The physical activity questionnaire consisted of questions concerning frequency of physical activity, intensity of physical activity, frequency of vigorous physical activity, hours spent on vigorous physical activity, average duration of a physical activity session, and participation in organized physical activity. After coding, a sum index of physical activity was calculated.
As an indicator of the socioeconomic status of the household, the Family Affluence Scale (FAS) was used (which ranged from 0 to 9 points, being higher socioeconomic status corresponding to highest score) . The FAS is a four-item questionnaire that helps students report their family income objectively: It evaluates the sum of scores regarding whether the family owns a car, whether the student has his/her own bedroom, the number of family vacations during the past 12 months, and the number of computers the family owns. FAS was used as a continuous variable as well due by other authors in a number of analyses focusing on health gradients [26–28].
The Kolmogorov-Smirnov test was used to assess the assumption of normality. Independent samples T-test or Mann–Whitney U test were performed to compare continuous variables and the χ2 test was used for categorical variables to assess differences between sample characteristics, dietary and nutritional data and urinary data stratified by sex.
Receiver operating characteristic (ROC) curves were used to analyse the potential diagnostic accuracy of sodium excretion to identify adolescents with low hydration status and to find the best trade-off between sensitivity and specificity. The area under the ROC curve (AUC) represents the ability of the test to correctly classify the participants with euhydrated status and risk of hypo-hydration. AUC values range between 1 (a perfect test) and 0.5 (a inadequate test).
Spearman's rank correlation coefficient was performed to assess the relationship between sodium excretion (mg/d) and urinary volume (ml/d), Uosm (mosm/kg), and FWR (ml/d). Kruskal–Wallis one-way analysis of variance and Mann–Whitney U test were used to identify differences for sodium excretion grouped by quartiles and by below or at above the upper limit recommendation (2000 mg/d), respectively.
Linear regression was used to estimate the association between the 24-hour urinary sodium excretion and the FWR. There were no interactions for sex x FWR (p-value for interaction = 0.420), however data from girls and boys were analyzed separately based on the existing sex differences in Uosm . The following variables were considered as potential covariates of FWR: BMI, energy intake (kcal/d), carbohydrate intake (% energy), fat intake (% energy), protein intake (% energy) and total water intake (g/d resulted from beverages and solid foods ingested), socio-economic status and physical activity. All variables were initially tested simultaneously, and after only those variables that significantly predicted the FWR (p < 0.05) and substantially modified the coefficient of sodium excretion (mg/d) by 10 % were included in the models. The crude model – Model 1, included FWR as dependent continuous variable and sodium excretion as the independent variable. For boys the adjusted model – Model 2, included FWR as dependent variable and sodium excretion, energy, total water intake. For girls the adjusted model – Model 2 included FWR as dependent variable and sodium excretion, energy, fiber (g/1000 kcal), carbohydrate intake % energy and total water intake as independent variables.
Data were analysed using IBM Statistics for Windows, Version 21.0 (Armonk, NY: IBM Corp) and Med Calc software v.10.4.5 (MedCalc Software, Mariakerke, Belgium). A p-value <0.05 was considered to indicate statistical significance. In this report, descriptive analysis is presented in terms of median and interquartile range, unless otherwise stated.
Sample size was calculated a priori for linear regression model considering 4 predictors. For a power sample ≥ 80 %, a medium effect size (f2 = 0.15) and α = 0.05 we had to enrol 85 subjects. Additionally since in boys the sample was lower than 85, we performed a post-hoc test to assess the power sample. According to the linear regression model results and α = 0.05 power sample was higher than 80 %.