Early Prediction Model For Gestational Diabetes Mellitus Combining Demographic Characteristics And Clinical Characteristics

Background:Gestational diabetes mellitus(GDM)is defined as the degree of glucose intolerance which is first recognized during pregnancy,and studies have shown that GDM was related to a couple of adverse outcomes during pregnancy.GDM is diagnosed by the raised blood glucose in mid-trimester of pregnancy(24 to 28 weeks,but hyperglycemia has already made a negative impact on the mother and fetus at that time.In recent years,with the improvement of life standards and the adjustment of childbirth policy in our country,people are paying more attention to prenatal care and pre-pregnant examination.To create early prediction models for GDM combining demographic characteristics and clinical characteristics and identify high risk group of GDM is gaining more importance.Methods:A retrospective study on 18000 singleton women who were recruited in Obstetrics and Gynecology Hospital of Zhejiang University from January 2017 to December 2017.Our exclusion criteria includes:not singleton;induced abortion;type 1 or type 2 diabetes mellitus,complicated with cancer,primary hypertension,poly cystic ovary syndrome;not Han population;some details missing.And 14307 women were included in our study.Demographic characteristics and clinical characteristics were collected.And they were divided into the observation group(80%)and the validation group(20%)randomly.Multivariable logistic regression analysis was performed to develop a risk prediction model in the observation group and the validation group evaluated it by the value of the area under the curve(AUC)of receiver operating characteristic(ROC).Results:1.Age,pre-pregnant BMI,family history of diabetes mellitus,gravidity,parity,previous history of GDM and macrosomia,previous history of spontaneous abortion,career(except group"others")were significant predictors of future GDM(P<0.05.However,previous history of abnormal fetus(P=0.735),residential place(P=0.954),education level(P<0.05 for all groups)were not related with GDM significantly.2.1779 pregnant women(15.5%)developed GDM in the observation group.These significant factors above were all explored through multivariable logistic regression.And age,pre-pregnant BMI.family history of diabetes mellitus,gravidity,parity,previous history of GDM and macrosomia were incorporated into our model to make a simple clinical scoring system eventually.At the multivariate level,we can know:1)The history of GDM is the strongest risk factor for GDM.Women with previous GDM history are 7.3 times more likely to have GDM than women without this history(95%CI:5.8-9.3,P<0.001);2)Age is another important risk factor for GDM.Compared with women younger than 25 years old,the risk of GDM in women aged 26-30 years,aged 31-34 years,aged 35-40 years and over 41 years increases to 1.9 times(95%CI:1.5-2.5),2.8 times(95%CI:2.1-3.6),4.0 times(95%CI:3.0-5.3),6.4 times(95%CI:4.3-9.5),respectively(P<0.001);3)When compared with the normal pre-pregnancy BMI group(18.5-23.9 kg/m2),the probability of GDM in overweight group(24-27.9 kg/m2)and obese group(≥28 kg/m2)was 1.4 times(95%CI:1.2-1.6)and 2.6 times(95%Cl:1.9-3.5),and P<0.001,while that of the lean group(<18.5 kg/m2)was 0.7 times(95%CI:0.6-0.9,P<0.001);4)The second gravidity,third gravidity is 1.3 times(95%Cl:1.1-1.5,P<0.001),1.4 times(95%Cl:1.2-1.7,P<0.001)more likely to suffer GDM than first gravidity women;5)The incidence of GDM in multiparous is 1.7 times(95%CI:1.5-2.0,P<0.001)higher than that in primiparous;6)The incidence of GDM in those with a history of macrosomia was 1.8 times(95%Cl:1.4-2.3,P<0.001)higher than those without;7)The incidence of GDM in patients with a family history of diabetes mellitus was 2.3 times(95%CI:1.9-2.9,P<0.001)higher than those without.3.2809 pregnant women in the validation group and 404 of them(14.4%)developed GDM.They each got a score ranging from-0.36 to 5.58 points and the AUC was 0.659.Thus,our clinical scoring system has certain predictive value for GDM.4.To find the best threshold score,we set different thresholds at the interval of 0.5 points.If the score was equal to or greater than the threshold score,the patient was assumed to be GDM by our clinical scoring system.In conclusion,the scoring system worked best when the threshold score is 1.5 points,the sensitivity reached 56.2%(95%CI:51.2-61.1%),while the specificity got 66.6%(95%CI:64.6-68.4),the positive predictive value was 22.0%(95%CI:19.5-24.7),and the negative prediction value is 90.0%(95%CI:88.5-91.4).What’s more,the positive predictive value reached 81.0%(95%CI:57.4-93.7%),the negative predictive value reached 86.5%.1%(95%CI:84.8-87.4%)at the threshold of 4.5 points.5.At the same time,we used the similar method to establish a prediction model for elderly women.Age,pre-pregnancy BMI,history of macrosomia,family history of diabetes mellitus,previous GDM history and residential place were included in the model.The AUC was 0.628(95%CI:0.60-0.65,P<0.001).Therefore,elderly women were not necessary to have a separate prediction model.Conclusion:GDM can be predicted by demographic characteristics and clinical characteristics like age,gravidity,parity,pre-pregnant BMI,family history of diabetes mellitus,previous history of GDM and macrosomia before the pregnancy started or during the first trimester,so that we can start life-style intervention as soon as possible to reduce the morbidity.