Childhood obesity is associated with a high degree of metabolic disturbance in children from Brazilian semi-arid region

Ethics statement

The project was approved by the institutional review board of the Fundação Bahiana para Desenvolvimento das Ciências, Salvador, Brazil, where the study was performed (CAAE: 35038914.3.0000.5544). All children consented to participate after meetings to explain the study, and written informed consent was obtained from each parent or guardian. All clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki.

Study design

This is a cross sectional study that evaluated school-aged children, based on public schools in the municipalities of Serrinha, Biritinga, Barrocas and Teofilândia, in the state of Bahia, Brazil, in 2019. Students between 5 and 12 years old, with a minimum school attendance of 70%, were randomly included from different schools in each municipality, maintaining the proportion between rural and urban areas. The students were excluded if (I) they had taken antibiotics or anti-inflammatory medications, (II) those who had experienced acute infections within the 30 days prior to sample collection; (III) individuals who were previously diagnosed with chronic diseases.

Data collection

Two pediatricians and two nutritionists obtained a brief medical history using a standardized questionnaire. All of them were trained before the study started. Height was measured to the nearest millimeter with a wall-mounted stadiometer (Altura Exata^®, Brazil), graduated every 10 cm, with a limit of 2.13 m. All students were instructed to stand in an upright position, feet together and barefoot. Weight was measured with an analytical anthropometric balance that was regularly calibrated by the accredited technical assistance provided by the National Institute of Metrology, Standardization and Industrial Quality (INMETRO) to the nearest 0.1 kg. Body mass index (BMI; in kg/m²) was calculated. Waist circumference (cm) was measured at the level of an imaginary horizontal line in the intermediate region between the margin of the lowest rib and the iliac crest. All anthropometric data were measured in triplicate.

BMI was classified according to the Z-score (BMI-for-Age) from the World Health Organization (WHO), using the WHO AnthroPlus^® tool (version 1.0.4). The classifications were as follows: underweight (Z-score < − 2), normal weight (Z-score between − 2 and + 1), overweight (Z-score >  + 1 and <  + 2), and obese (Z-score ≥  + 2)¹⁵.The criteria established by Taylor et al. were used to classify waist circumference measurements as either normal or altered¹⁰.

For the assessment of laboratory parameters, blood samples were drawn from either the cubital or radial vein, collecting 15 ml after a minimum fasting period of 8 h. The CRP was ascertained using the Turbidimetry method, utilizing the AU680 device from Beckman Coulter^®, with compatible analysis kits from the same manufacturer. The hexokinase method was used for measuring glucose. Total cholesterol (TC), high-density lipoprotein (HDL), and triglycerides (Tg) were quantified using an automated enzymatic colorimetric method with the analysis kits from Biosystems^® and their BS 200 model equipment. The determination of low-density lipoprotein (LDL) was carried out computationally, employing the Friedwald equation ([LDL] = [TC] − ([HDL] + [Tg/5])).

Adaptation of molecular degree of perturbation to examine plasma concentration biomarkers

The Metabolic Disturbance Degree (MDD) is adapted from the Molecular Degree of Perturbation (MDP). This metric offers a standard based on typical values from lipid profile, blood glucose measurements, and C-reactive protein levels, as detailed in Table 1. Its computation was achieved using the mdp package in R version 4.2.2¹⁶. For each test, the MDD score was determined by comparing the concentration variances from the average value for a given test in the reference population and then normalizing by the population’s standard deviation. Essentially, the MDD score quantifies the deviation, in standard deviations, from a healthy control group. The specific equation employed to compute the MDP in this research is presented below:

where: \(\sigma = \sqrt\sum _i=1^n\frac(x_i- \underlinex)n-1\), n = Number of data points, x_i = Each of the value of data, \(\underlinex\) = Mean of the data points, σ = Standard deviation.

Within the scope of this study, we applied the MDD to six markers indicative of metabolic status. Those individuals presenting an MDD value exceeding the third quartile were classified as “Highly perturbed” (Fig. 1; Supplementary Table 1).

Frequency and distribution of MDD levels. Panel (A) displays the frequency distribution of Metabolic Disturbance Degree (MDD) scores among the studied population, with a dotted line indicating the threshold for high MDD cutoff. Panel (B) provides a visual comparison of MDD scores between the low and high MDD groups, using boxplots to highlight the range, median, and potential outliers in each group.

Statistical analysis

The sample size was calculated using WINPEPI (version 11.65; considering a 5% alpha and 90% power for a two-sided 0.25 correlation coefficient between high sensitivity CRP and Z-score, estimating the need for 165 individuals¹⁷. Sociodemographic, clinical, and laboratory characteristics were compared across low and high degrees of MDD. For continuous variables, the Mann–Whitney U test was employed, while Fisher’s exact test was used for categorical ones. To ascertain whether the BMI z-score and age-stratified waist circumference were independently linked to high scores on the Metabolic Disturbance Score, a backward logistic regression was conducted. During this analysis, sociodemographic factors were viewed as potential confounders. Hierarchical cluster analysis (Ward’s method) of z-score scaled data was employed to depict the overall expression profile of indicated biomarkers in the study subgroups. To assess differences between patients’ characteristics within each BMI strata, the Kruskal–Wallis test was employed along with Chi-squared test, with Bonferroni’s adjustment for multiple comparisons.

We further conducted a post-hoc power analysis with the G*Power software (version 3.1.9.7; revealing a two-tailed alpha coefficient of 0.05 and a beta error of 0.17¹⁸. This was in relation to a 23% difference in the MDD score prevalence between the overweight and the normal weight groups. Our study does not have enough power to evaluate differences between the normal weight and underweight, due to a small sample of underweight patients.