Risk Factors Of Neural Tube Defects And Its Prediction Models

Objective:We explore the main influence factors of neural tube defects in Shanxi Province,and we use Logistic regression model and artificial neural network model to establish a risk assessment model to evaluate NTDs and identify high-risk groups.We also can provide scientific basis for the primary prevention of neural tube defects,and put forward reasonable prevention and suggestions.Last the prevention of neural tube defects should be carried out.Methods:Our study was a 1:1 case-control design.We recruited women who gave birth to children with neural tube defects from 2010 to 2014.These people came from monitoring hospitals or maternal and child health institutions in various counties included in the "peak cutting project" in Shanxi Province.Our diagnostic criteria are based on the Monitoring program for birth defects in China,411 cases were investigated.Case group(including live birth,stillbirth,and neural tube defects diagnosed in prenatal diagnosis).NTDs types included: single neural tube defect are 320 cases(77.86%);Multiple neural tube defects are 46 cases(11.19%);Neural tube defects were associated with other birth defects are 45 cases(10.95%).We matched and selected the control group according to the 1:1 case-control design method.We select the control group women who deliver a normal baby from the maternal and child health system for investigation,in the end,we investigated 411 cases.The matching conditions were as follows:(1)living in the same county(or district)in before and after pregnancy time;(2)the nationality is same;(3)the difference in the last menstrual period was within 3 months,and deliver a normal baby(the baby without any birth defects).Finally,a questionnaire survey was conducted forthem.We use SPSS22.0 software analyze the data,and establish the Logistic regression prediction model and multilayer perceptron neural network prediction model of neural tube defects.Results:1.In the case group,there were 320 patients with single neural tube defect,196 with spina bifida,accounting for 47.69%;79 with anencephaly,accounting for 19.22%;45with encephalocele,accounting for 10.95%.There were 46 patients with multiple neural tube defects,including 28 patients with anencephaly and spina bifida,accounting for6.81%;13 patients with spina bifida and encephalocele,accounting for 3.16%;4 patients with anencephaly and encephalocele,accounting for 0.97%;1 patient with anencephaly and spina bifida and encephalocele,accounting for 0.24%.There were 45 patients with neural tube defects and other birth defects.2.Single factor analysis suggest us that: elderly parturient women,the educational level of women and her husband,BMI,fetus number,gravidity,unplanned pregnancy,standardized antenatal examination,multiple pregnancy,whether to accept pre-marital or pre-pregnancy guidance,whether to accept the guidance of the wedding date,hyperpyrexia,catch a cold,pregnant vomit,take antibiotics,take antipyretic and analgesic drugs,take cold medicine,take a folic acid supplement,take meat,eggs,dairy,fresh vegetables,fresh fruit,beans and its products,passive smoking,drinking tea,working time in front of the computer,pets,bedroom kitchen division,fuel type,pregnancy in the winter,mental stimulation,husband smoking,husband drinks wine,husband drinks other alcoholic drinks,husband is exposed to toxic metals.The distribution of these factors was statistically different between the patients and the control group(P < 0.05).3.Multivariate Logistic regression analysis of NTDs influencing factors showed that: BMI is overweight(OR=2.101,95%CI: 1.305~3.382),multiple pregnancy(OR=9.682,95%CI: 2.271~41.282),number of gravidity ≥3(OR=2.645,95%CI:1.508~4.639),unplanned pregnancy(OR=2.030,95%CI: 1.210~3.405),have a cold(OR=3.089,95%CI:1.776~5.373),working time in front of the computer ≥20 hours/week(OR=5.722,95%CI: 1.229~26.632),passive smoking > 6 times/week(OR=3.289,95%CI:1.317~8.213)are risk factors for neural tube defects,folic acid supplementation(OR=0.386,95%CI: 0.253~0.588),standardized prenatal examination(OR=0.297,95%CI: 0.137~0.643),health guidance for marriage or the women before pregnancy,(OR=0.365,95%CI: 0.217~0.615),don’t keep pets(OR=0.519,95%CI: 0.324~0.834)are protective factors for neural tube defects.4.The AUC=0.881 and P<0.001,of the multivariate conditional Logistic regression model of neural tube defect established in this paper was significant for the diagnosis of NTDs,and the prediction effect of this model was relatively good.The area under the ROC curve(AUC)of the MPL neural network model is 0.850,P<0.001.The model is meaningful for NTDs diagnosis,and the prediction accuracy is 76.80%.The prediction effect of the model is relatively good.The prediction performance of the multi-layer perceptron neural network model established in this study is worse than that of the conditional Logistic regression model,but there is no statistical difference between the two models.Conclusions:1.Based on the analysis of the influencing factors of neural tube defects,it is concluded that peri-conception women should take folic acid under the direction of doctor’s order,avoid unplanned pregnancy,avoid the blind pursuit of multiple pregnancy,avoid exposure to passive smoking,don’t keep pets,don’t catch a cold and contact other high-risk factors,which can reduce the risk of NTDs.2.The prevention of neural tube defects in Shanxi Province should pay more attention to the early pregnancy women before and during pregnancy,and strengthen the health education and prenatal counseling for women during pregnancy.3.The conditional Logistic regression and multi-layer perceptron neural network prediction models of neural tube defects constructed in this study have some reference significance for the prediction of neural tube defect risk,which can be used as an initialtool for NTDs risk assessment,and also provide a good foundation for the realization of more accurate models in future research.