Predictive Effect Of Thyroid Hormone In First Trimester Of Pregnancy On Gestational Diabetes Mellitus

Objectives To describe the associations of maternal thyroid hormone levels in the first trimester and blood glucose levels.To identify thyroid function makers have significant influences on the risk of GDM in the second trimester,and to establish an effective GDM predictive model combining with conventional variations in order to identify high risk individuals at early stage.MethodsThe participants of this study were from the Ma ’anshan Birth Cohort(MABC),which includes 3474 pregnant women whose first antenatal check-up are before 14 weeks.After excluding women with a history of family or personal thyroid disease,pre-pregnancy diabetes,extrauterine pregnancy,embryo damage or miscarry during pregnancy,lack of GDM and thyroid function makers data,3 072 pregnant women were finally included in this analysis.All participants were asked to fill out a series of questionnaires to gather information like age,education,household income,history of the metabolic diseases,gravidity and so on.Pregnancy-related information and blood glucose testing data were extracted from clinical records.Their blood samples were also collected for further usage.Serum concentrations of T3,T4,FT4,TSH,TPOAb,and TGAb were measured using electrochemiluminescence immunoassay(ECLIA)on Cobas Elecsys 411(Roche Diagnostics Gmb H,Mannheim,Germany).Missing data,such as age and pre-BMI,are imputed by multiple interpolations.Pearson’s chi-square test and independent samples ttests were used to compare baseline characteristics between different groups.The association between maternal thyroid hormone levels and blood glucose levels was analyzed using multiple linear regression.The thyroid hormone levels were grouped into four quartiles,from Q1 to Q4.The association between different thyroid hormone levels and the risk of developing GDM was analyzed using binary logistic regression.Significant variables were selected by the logistic stepwise regression,to establish a predictive model for GDM in early pregnancy.Using Med Calc,the ROC curves were drawn to exhibit the sensitivity and specificity of different models and were compared,to judge their differentiation.Hosmer-Lemeshow goodness of fit test and calibration diagram was used to judge the calibration degree of the GDM predictive model.ResultsA total of 3072 pregnant women were included in the study and 12.6% of them were diagnosed with GDM.TPOAb and TGAb positive rate were 12% and 5.5%,respectively,however,no significant difference was found between GDM and Non-GDM.The linear regression analysis showed a significant positive association among T3,TSH,T3/FT4,and blood glucose levels in the second trimester.FT4 was significantly positively correlated with FPG in the first trimester,while was negatively associated with OGTT 1h PG and2 h PG in the second trimester,and T4 showed similar associations.The risk of GDM was higher among individuals with higher level of T3 and T3/FT4 compared with the lower level,while the risk of GDM was lower in the Q4 group of FT4 when the Q1 group was used as a reference.After adjusting for maternal age,pre-BMI,family history of diabetes,educational experience,parity,residence,household average monthly income,nature of work,conception season,smoking,and drinking experience.Logistic stepwise regression analysis showed that the main variables remarkably affecting the risk of GDM in the second trimester were T3,FT4,age,pre-pregnancy BMI,FPG in the first trimester,gestational week of detection,conception season,household average monthly income,and family history of diabetes.The order of AUC of the models with single biomarker from large to small was FPG in the first trimester(0.605)> T3/FT4(0.602)> T3(0.589)>FT4(0.542),and the corresponding critical values were 4.84 mmol/L,138.18,2.49nmol/L and 17.88 pmol/L,respectively.The best multivariate predictive model of GDM including T3,FT4,maternal age,pre-pregnancy BMI,FPG in the first trimester,gestational week of detection,conception season,and family history of diabetes(AUC= 0.725,95%CI: 0.697～0.753).And Hosmer-Lemeshow goodness of fit test and the calibration diagram both indicated that the calibration capability of this model is good.ConclusionsT3 and FT4 in the first trimester were positively and negatively associated with the risk of GDM in the second trimester,respectively.The predictive performance of the early predictive models for GDM constructed using the two markers along with conventional variables are at a medium level,which indicate the necessity of thyroid function detection in early pregnancy for the earlier detection of the high-risk population of GDM.However,the external applicability of the model needs to further verification.