68-3

57/130

Statistical analysisStudent’s t-test (for continuous variables), Fish-er’s exact test (for categorical variables), or Mann- Whitney’s U-test (ordinal scale) were used for comparisons between the two groups. Pearson’s correlation coefficient was used to analyze the rela-tionships between UCo and BMI. Multiple linear regression analysis was performed for assessing the relationships between UCo and urinary concen-trations of the other 17 trace elements and BMI, controlling for possible confounding variables. All covariates were included using the forced-entry method. Analyses of all models were gender-strat-ified. The variance inflation factor (VIF) was employed for checking the multicollinearity problem among the variables. We used Bonferroni’s correc-tion to correct multiple comparisons, such as repeating the statistical tests 36 times (18 measured trace elements for both boys and girls), with a p- value<0.001 (0.05/36), which was considered to indicate a statistically significant difference. All statistical analyses were conducted using IBM Statistical Package for Social Sciences (SPSS), version 27.0 (IBM Corp., New York, USA).The mean BMI of the children was 17.9 kg/m2 and very close between boys and girls (17.8 and 17.9 kg/m2, respectively). The mean log UCo was −0.295 μg/g, and there was no significant differ-ence between boys and girls (mean=−0.308 and Co, Ni, Cu, Zn, As, Se, Sr, Mo, Cd, Ba, and Tl, and ICP-AES (Optima 2100, PerkinElmer, Massachu-setts, USA) was used for measuring Na, Ca, Mg, and K. For the measurement preparation, urine samples were melted at room temperature and mixed with 0.5 % HNO3 with 5-fold dilution in ICP-MS and analyzed by the multi-element stan-dard solution XSTC-13 (SPEX CertiPrep, New Jersey, USA) as the external standard solution. For ICP-AES, the urine samples were diluted 10-fold with 0.5 % HNO3 and analyzed using XSTC-2A (SPEX CertiPrep, New Jersey, USA) as the external standard solution. Measurements were repeated three times, and the average of the three measurements was used for statistical analyses. For instrument calibration throughout the measure-ments, at least 10 % of the analyses were external standards, and 5 % were blank (pure water). For statistical analysis, the values under the limit of detection (LOD) were substituted with half the LOD. To correct for variations in urine dilution, the concentration of every trace element was expressed as a ratio to urinary creatinine concentration. To reduce the influence of outliers and normalize the right-skewed distribution, we used the natural logarithm of the urinary concentration of trace elements in the statistical analysis. Among the 18 trace elements, we focused on cobalt, since several studies showed an inverse association of this element with childhood obesity (Table 1). Figure 1　Study flow chart253Results