Statistical analysisBefore statisti

Statistical analysis

Before statistical analysis, the normality of the distribution and the homogeneity of the variances were evaluated using Kolmogorov-Smirnov test and Levene’s test, respectively. Data are presented as means ± standard deviations (SD) or medians with interquartile ranges in the case of a skewed distribution. Statistical analysis were performed with ANOVA followed by a multiplecomparison test for subgroups by LSD. The difference of TSH within normal range between normal and abnormal lipid groups was tested by Mann-Whitney test. The log transformation of lipid parameters was performed to correct the skewed distribution [35]. The relationships between sex, age, history of DM, history of hypertension, drinking, TSH, FT3, FT4, FBG, UA and the log-transformed lipid profile were assessed by simple correlation analysis. Stepwise multiple linear regression analysis was performed to estimate the influence of the above parameters on the lipid profile. Because FT4 (FT3) and TSH are closely correlated, we did not use FT4 and FT3 in the same model with TSH. Model 1 includes TSH and other parameters except FT3 and FT4, while model 2 includes FT3, FT4 and other parameters except TSH. The results are expressed as beta (b)- coefficients with 95% confidence intervals (CI) denoting the SD change in terms of lipid profile markers per unit increase in the TSH level. Backward stepwise logistic regression was used to evaluate the odds ratio (OR) of dyslipidemia in relation to serum TSH. The level of statistical significance was set at P < 0.05, except for the regression analysis, in which P < 0.10 was set as the cut-off point. All the above analyses were conducted with the use of SPSS software (SPSS Inc., Chicago, IL, USA), in version 17.0.