Study design and participants
This is a cross-sectional study that used data from two centers, part of the Study of Cardiovascular Risks in Adolescents (ERICA 2013–2014). ERICA is a school-based, national, multicenter and cross-sectional study carried out in rural and urban contexts. The design of the ERICA study has been published previously (2015) [14]. Briefly, 73,624 students aged 12–17 years were enrolled from private and public schools, located in one of the 273 Brazilian municipalities with more than 100,000 inhabitants [15]. For the present study, all female students from Porto Velho-RO (PVh) and Porto Alegre-RS (PoA), who participated in all research stages of ERICA and had already had menarche, were included.
This study was approved by the Research Ethics Committee (REC) of the Federal University of Rondônia, Federal University of Rio Grande do Sul and the Institute of Studies in Collective Health of the University of Rio de Janeiro (Protocol 45/2008), and was conducted according to the principles of the Helsinki declaration [16]. Written informed consent was obtained from each student and from his or her parents. The present study included a subsample of the students residing in two capitals: Porto Velho (PVh) and Porto Alegre (PoA), respectively, located in the Southern and Northern regions of Brazil.
Data collection
A self-administered questionnaire using a personal digital assistant (PDA, model LG GM750Q) was administered. Data regarding sociodemographic, behavioral and diet characteristics were obtained. The economic status was defined according to the Brazilian Association of Companies and Research (ABEP in the Portuguese acronym), as A1 (the highest social class), A2, B1, B2, C1, C2, D and E (the lowest social class) [17] data were grouped into 3 categories: A, B and C/D. Age was collected in full years and further grouped (12–13, 14, and 15–17 years. Ethnicity was defined by skin color as white and non-white (black, mixed or indigenous) [18, 19]. Smoking and alcohol consumption were assessed according to whether participants had already experimented them or not [19,20,21]. Physical activity was categorized as inactive (students with no leisure-time, physical activity, or who exercised less than 300 min/week), or active for those who exercised from ≥ 300 to 1200 min/week) [22]. Recommended screen time was up to 2 h per day, and not recommended was more than 2 h per day, according to the American Academy of Pediatrics guidelines [23]. Menarche was assessed according to age (age at menarche 9–12, or 13–16 years) and time since menarche, that is, peri-menarche: less than 2 years since menarche, and post-menarche: 2 years or more since menarche. This classification considered the expected time for maturation of the reproductive axis, after menarche, accepted as being more than 2 years of occurrence of the menarche event [24].
As for nutritional status, body mass index (BMI) was used. Height was assessed using a portable and demountable stadiometer, Alturexata® [25]. Body weight was assessed using a digital scale from Leader, model P150m, capacity of 200 kg and precision of 50 g. BMI was defined by weight (kilograms) divided by square of the height (meters). The girls were stratified by overweight and obesity (z-score > 1) and normal weight groups (z-score ≤ 1), according to BMI-for-age z-scores from the World Health Organization child growth standards [26]. Waist circumference was measured to the nearest 1 mm using a fiber glass anthropometric tape, with millimeter resolution and length of 1.5 m (Sanny®, São Paulo, Brazil). WC classification followed the International Diabetes Federation (IDF) guidelines, which uses the 90th percentile as a cutoff point for girls up to 16 years old and 80 cm for those over 16 years old [27].
Dietary intake was assessed using a 24-h recall performed by trained interviewers. The food and drinks consumed were recorded in all meals and snacks before the interview in the dietary assessment software, ERICA-REC 24 h [28]. Portion size estimation was obtained by showing photographs included in the software. Nutritional composition was calculated using the software database consisted of 1626 food items based on data from a Dietary National Survey carried out from 2008 to 2009 [29]. Energy and nutrients were estimated using the IBGE table [30].
Specifically for this study, nutritional characterization followed the dietary reference intake (DRI) [31] and presented total energy intake (Kcal), percentage of trans fatty acid (TFA), and ratio of omega-6 to omega-3 fatty acids. The three components were characterized according to mean intake in PVh and PoA.
Outcome assessments
Blood samples were collected after 12 h overnight fasting. Glucose was measured by the hexoquinase method; triglycerides, by enzymatic kinetics and HDL-cholesterol by enzymatic colorimetric assay (ADVIA 2400, Siemens). LDL- cholesterol was calculated by the Friedewald equation. Insulin was determined by chemiluminescence method (Modular Analytics-Roche) [32]. IR was calculated using the model of insulin homeostasis, HOMA-IR Index as follows: insulin (mU/L) × (glucose (mg/dL) × 0.0555)/22.5, as proposed by Matthews et al. [33]. The ≥ 3.16 cutoff point, according to the first guidelines for the prevention of atherosclerosis in childhood and adolescence [31], was used in our analysis. Fasting insulin, with a cutoff of ≥ 15 mU/mL was also assessed as an additional marker of IR [34].
Statistical analysis
Demographic, nutritional, anthropometric, and biochemical variables were expressed as a percentage and 95% confidence interval (CI). The differences between the cities were assessed using the student t test for continuous variables and the chi-square test for dichotomous variables.
All factors with IR associated were converted to categorical variables to enable the comparability of prevalence ratios (PRs). In the evaluation of unadjusted and adjusted measures of effect in the multivariate models, Poisson regression with robust variance was used. The adjusted analysis followed a conceptual model defined a priori [35]. Variables that were associated with outcomes at a significance level of ≤ 20% in the unadjusted analysis were included in the multivariate model as potential confounders. At level I, the most distal level of determination, sociodemographic variables were included; at level II, the reproductive and behavioral ones; and at level III, the most proximal, the nutritional status variables. Finally, variables with a p value of ≤ 0.05 were considered associated with the outcomes, that is, IR (insulin levels and HOMA-IR). Due to the collinearity between WC and overweight/obesity (Ow/Ob), these two variables of nutritional status were entered into different models, model 1 and model 2, respectively (shown in Fig. 1).
In addition, the multivariate-adjusted Poisson regression model with robust estimates was used to assess the association among the subgroups of time since menarche (peri-menarche and post-menarche) in each city with demographic, lifestyle, anthropometric factors and IR.
Statistical analyzes were performed using the statistical software STATA, version 14 (Stata Corporation, College Station, TX, USA).