The SAS 9.13 statistic software package (SAS Institute, Carey, New Jersey) was used for data evaluation and statistical analysis. For descriptive analysis, mean and standard deviation were used for normally distributed variables. Wilcoxon rank
sum test/Kruskal-Wallis-test were used to compare descriptive parameters between boys and girls, or among the 3 age groups. For multiple comparisons, the Bonferroni method was used. A P value .05 was considered significant. A multivariate mixed model (SAS procedure mixed) was used to relate smoking status to major risk indicators in pediatric diabetology (metabolic control, insulin dose, BMI, blood
pressure, dyslipidemia), adjusting for age, diabetes duration, sex, and insulin therapy, as well as interactions among the risk factors (fixed effects) and center heterogeneity (random effect).
Adjusted means (lsmeans) with corresponding P values are reported to compare smokers and nonsmokers. Because serum triglycerides differ between the fasting and the postprandial state, this variable was also included as a potential confounder into the model with triglycerides as dependent variable.