2.8. Statistics
Conditional logistic regressions, taking the matched design into account, were used for evaluation of the associations between maternal exposure to air pollution and the odds of developing T1D. We used IBM SPSS version 21 and Cytel Studios LogXact. The results were considered as statistically significant if the 95% con- fidence interval [CI] did not include 1.
The robustness of the results was evaluated by sensitivity
analyses where we excluded children of mothers who (i) had T1D, (ii) had gestational diabetes, (iii) had been born in another coun- try, or (iv) smoked and (v) children born Large for Gestational Age [LGA]. As an additional sensitivity analysis we excluded cases and controls that were not exactly HLA-matched. We also performed multivariate analyses including the following variables; maternal T1D, gestational diabetes, preeclampsia, maternal country of ori- gin, smoking and LGA.
Furthermore, as case children come from both the DiPiS cohort
and the regional registries, while the controls come only from the DiPiS cohort, we were concerned that there might be some se- lection bias. We therefore investigated the air pollution exposure prevalence including only case children who had participated and been diagnosed with T1D within the DiPiS project, see Fig. 1. We also aimed to repeat all the analyses above in this subsample, but due to the small sample size (case children, n ¼ 49) and the few children with T1D who were exposed to the highest quartile of exposure prenatally (n ¼ 8), models could not be fitted to the data.
2.8. StatisticsConditional logistic regressions, taking the matched design into account, were used for evaluation of the associations between maternal exposure to air pollution and the odds of developing T1D. We used IBM SPSS version 21 and Cytel Studios LogXact. The results were considered as statistically significant if the 95% con- fidence interval [CI] did not include 1.The robustness of the results was evaluated by sensitivityanalyses where we excluded children of mothers who (i) had T1D, (ii) had gestational diabetes, (iii) had been born in another coun- try, or (iv) smoked and (v) children born Large for Gestational Age [LGA]. As an additional sensitivity analysis we excluded cases and controls that were not exactly HLA-matched. We also performed multivariate analyses including the following variables; maternal T1D, gestational diabetes, preeclampsia, maternal country of ori- gin, smoking and LGA.Furthermore, as case children come from both the DiPiS cohortand the regional registries, while the controls come only from the DiPiS cohort, we were concerned that there might be some se- lection bias. We therefore investigated the air pollution exposure prevalence including only case children who had participated and been diagnosed with T1D within the DiPiS project, see Fig. 1. We also aimed to repeat all the analyses above in this subsample, but due to the small sample size (case children, n ¼ 49) and the few children with T1D who were exposed to the highest quartile of exposure prenatally (n ¼ 8), models could not be fitted to the data.
การแปล กรุณารอสักครู่..