Determine the prevalence of gestational diabetes in Ardakan and its related factors

Graphical abstract


Protocol data
Our study showed that the highest prevalence of gestational diabetes in women aged 35-39 years. The odds of gestational diabetes increase significantly with increasing maternal age, body mass index (BMI), gestational diabetes in previous pregnancies and macrosomal child. Planning for effective prevention and intervention at local and provincial levels, with an emphasis on the food culture and customs of each region, can increase the health of mothers and infants.

Description of protocol
Among 3808 pregnant women studied, 286 (7.5%) women were diagnosed with gestational diabetes, the highest prevalence of gestational diabetes in women aged 35-39 years (64, 18.4%). The mean age was 27.1 (5.4) with a range of 14-46 years, and the mean body mass index (BMI) was 25.7 (4.9). The mean gestational age was 10.7 (6.6) weeks with a range of 3-34, 93 women (2.4%) were rural residents and 3715 (97.6%) women were urban residents. The mean number of pregnancies was 2.14 with a range of 1-8. Body mass index was divided into three groups: normal (less than or equal to 24.9) and overweight (25-29.9) and obese (more than or equal to 30). Accordingly, 1716 women were in the normal weight group (45.6%), 1348 women were in the overweight group (35.8%) and 702 women were in the obese group (18.6%). In obese women, GDM (gestational diabetes mellitus) of 85 women (12.1%) was more, and in pregnancy that resulted in abortion, 18 women (9%) had a slightly higher prevalence of gestational diabetes Approximately 1.4% of mothers with a birth weight of over 4 kg had GDM, as well as the average pregnancy rank (2.48%) was higher in the mothers with GDM than other mothers (2.12%). The distribution of GDM incidence based on age groups is presented in Table 1. The results of univariate logistic regression analysis are given in Table 2. In the following, regarding the method of Hosmer-Lemeshow to develop the final logistic regression model, the results are reported in Table 3. Also, Fig. 1 shows Forest plot of adjusted and unadjusted logistic regression models in the   Fig. 2. Using Youden index, the best cut off for multivariate logistic regression models was 0.297 with a sensitivity of 0.649 and feature of 0.648. Finally, Fig. 3 shows the boxplot of the results of the evaluation of multivariate regression model using Giancristofaro method. We can conclude, based on the two boxplots, the validity of multivariate regression model. According  to the analysis of Multivariate Logistic Regression Model, the mothers with a BMI greater than 30 had 62% (significantly) higher odds of GDM. The odds of GDM in pregnancy that leads to a baby above 4 kg is 2.2 times more than other pregnancies (P = 0.05) (95% CI: 0.979-5.137).

Materials and methods
This descriptive-analytic cross-sectional study has been conducted on the information of all pregnant women including 4083 women who were referred to Comprehensive Health Services Centers and Healthcare Centers in Ardakan in 2014 for receiving care of pregnancy. The information used in the present study included demographic and clinical information of the mother (such as BMI in the three groups of normal (less than or equal to 24.9), overweight (25-29.9) and obese (more than or equal to 30), maternal fertility history (successful and unsuccessful pregnancy, abnormal birth weight, pregnancy and delivery rank, history of abortion and stillbirth, and history of complications of childbirth), and the condition of the mother with all types of diabetes [1][2][3][4][5]. The risk factors for gestational diabetes such as over age 30, gestational diabetes history in previous pregnancies, history of pre-diabetes (IFG (impaired fasting glucose) or IGT (impaired glucose tolerance)), history of diabetes in the first-degree family, high blood pressure history, birth history of 4 kg or more, BMI equal to or greater than 30 kg/m 2 before pregnancy, history of spontaneous abortion with no cause twice or more times, history of stillbirth, history of fetal abnormalities in previous pregnancies were cases used in the present study.
Also, gestational diabetes is defined as high blood sugar levels three hours after consuming 75 g of glucose for women with a fasting blood sugar level greater than 92-126 mg/dl and at the 24th to 28th gestational weeks [6]. According to this definition, the mothers with type 2 diabetes were excluded from the study. Also, some of the information of the pregnant mother, such as the end of pregnancy  (newborn, abortion, and stillbirth), delivery type (natural or cesarean), gestational age, birth weight, baby abnormalities, maternal complications (prolonged delivery and delivery bleeding) were also used for the present study. The data were analyzed using software SPSS (version 24). The main consequence of this study was the classification of gestational diabetes, which is as a two-state variable (yes/no), and in addition to a descriptive analysis, univariate and multivariate logistic regression model was used for obtaining modified effects of risk factors as the odds ratio (OR). The criterion used for independent variables in the multiple logistic regression model was also considered to be less than p value of 0.2 in all analyzes, the level of significance was less than 0.05 Software SPSS version 21 was used to analyze the data.

Statistical methods
Descriptive statistics have been reported for quantitative variables including mean and standard deviation, and for qualitative variables including frequency (%). In order to study the unadjusted effect of each of the independent variables (risk factors) on the incidence of disease, one-variable logistic regression model analysis was used. In order to obtain adjusted data for each independent variable according to Hosmer and Lemeshow approach to develop multivariate logistic regression model [7], those variables with a likelihood ratio test of less than 0.2 were introduced into multivariate logistic regression model and adjusted effects' estimation was reported according to the method of Hosmer and Lemeshow. Also, in order to study the goodness of fitting of the regression model, the goodness test of the fit of Hosmer and Lemeshow was used. In order to study the differential power of the logistic regression model, Receiver Operating Characteristic (ROC) curves and the area under the curve (AUC) were used [8]. Meanwhile, in order to obtain the best cut-off point for the estimated possibility values by the final multivariate logistic regression model Youden index was used. Finally, in order to evaluate the validity of the final logistic regression model, the method described in [9] has been used, based on this method, first the data were randomly assigned 100 times to two parts with ratios of 75 and 25 percent, then fitted using the larger part (75%) of the model and the model parameters are estimated. Then, using the smaller part data (25%) and estimated parameters at the previous stage, response variable values were predicted, ROC curves are plotted and AUC was provided for both parts of the data. Also, forest plot was used to show adjusted and unadjusted effects of each of the independent variables (risk factors) on the logarithmic scale [10]. All statistical and analytical calculations were performed by software R version 3.5.0 (32bit) and using packages of ggplot2, CA Tools and pROC.