Article type
Year
Abstract
Background: The rising incidence of gestational diabetes (GDM) contributes to an increasing number of adverse pregnancy outcomes (for example macrosomia and neonatal hypoglycemia). Currently, risk stratification in early pregnancy is based on single medical and/or obstetrical history risk factors. GDM, commonly diagnosed in late pregnancy, is particularly suited for early prediction in pregnancy. Timely recognition and treatment of GDM will offer opportunities to improve pregnancy outcome.
Objectives: To identify and validate existing prediction models for GDM in the first trimester of pregnancy.
Methods: MEDLINE and EMBASE were searched up to December 2014. First trimester non-invasive prediction models for GDM in current pregnancy were included as well as external validation studies thereof. Eligibility was appraised by two reviewers. For each prediction model, data were extracted according to the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS).
Models were validated using data from a large prospective study, the RESPECT cohort (3641 pregnancies of which 171 developed GDM).
Results: Our systematic review identified 20 articles, 16 of which were model development studies and four were external validations. Of the 16 models, 12 could be validated (Figure 1). In developmental studies C-statistics ranged from 0.63 to 0.82. Four models have been externally validated with C-statistics from 0.60 to 0.76. In our validation study, 11 recalibrated models yielded C-statistics of 0.67 to 0.78. At a fixed false positive rate of 10%, sensitivity ranged from 27% to 43%. Most of the models showed acceptable to good calibration (Figure 2). Common predictor variables included in the models were age, body mass index, ethnicity, history of GDM and family history of diabetes mellitus.
Conclusions: Prediction models for GDM can be used to identify high risk pregnancies. This allows early risk stratification and enables customized targeted interventions. Screening will contribute to the development of personalized obstetric care and improve the utilization of health care resources.
Objectives: To identify and validate existing prediction models for GDM in the first trimester of pregnancy.
Methods: MEDLINE and EMBASE were searched up to December 2014. First trimester non-invasive prediction models for GDM in current pregnancy were included as well as external validation studies thereof. Eligibility was appraised by two reviewers. For each prediction model, data were extracted according to the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS).
Models were validated using data from a large prospective study, the RESPECT cohort (3641 pregnancies of which 171 developed GDM).
Results: Our systematic review identified 20 articles, 16 of which were model development studies and four were external validations. Of the 16 models, 12 could be validated (Figure 1). In developmental studies C-statistics ranged from 0.63 to 0.82. Four models have been externally validated with C-statistics from 0.60 to 0.76. In our validation study, 11 recalibrated models yielded C-statistics of 0.67 to 0.78. At a fixed false positive rate of 10%, sensitivity ranged from 27% to 43%. Most of the models showed acceptable to good calibration (Figure 2). Common predictor variables included in the models were age, body mass index, ethnicity, history of GDM and family history of diabetes mellitus.
Conclusions: Prediction models for GDM can be used to identify high risk pregnancies. This allows early risk stratification and enables customized targeted interventions. Screening will contribute to the development of personalized obstetric care and improve the utilization of health care resources.