Prediction of insulin treatment in women with gestational diabetes mellitus.

The identification of pregnant women with Gestational Diabetes Mellitus (GDM) who will require insulin therapy, may modify their management to closer monitoring and probable early interventions. The aim of the study was to develop a predictive model for the necessity of insulin treatment in women with GDM. This was a prospective cohort study. Data from 775 women diagnosed with GDM per the IADPSG criteria were analyzed using logistic regression and a machine learning algorithm, the Classification and Regression Trees (CART). Potential predictors routinely recorded at follow-up visits were tested and used for the development of the model. The resultant model was externally validated using the data from two different perinatology clinics. Preconceptional maternal BMI and morning fasting blood glucose levels at baseline and at 1 h during an Oral Glucose Tolerance Test (OGTT) were independent significant predictors for the treatment modality of GDM. Baseline blood glucose greater than 98 mg/dl and preconceptional maternal Body Mass Index (BMI) between 26 and 31 kg/height2 increased substantially the probability of insulin therapy (odds ratio [OR] 4.04, 95% confidence interval [CI] CI 2.65–6.17 and 2.21, 95%CI 1.42–3.43, respectively). The area under the curve (AUC) for the internal and external validation of the predictive model was 0.74 and 0.77, respectively. A simple model based on maternal characteristics and the values of an OGTT can predict the need for insulin treatment with accuracy. Overweight women with an abnormal baseline blood glucose at OGTT are at high likelihood for insulin treatment. Fifteen to 30% of women with Gestational Diabetes Mellitus (GDM) require insulin therapy. Overweight women with baseline blood glucose greater than 98 mg/dl at OGTT are at increased risk for insulin treatment and close monitoring and increased physical exercise are required.