ACCURACY OF GLYCOSURIA, RANDOM BLOOD GLUCOSE AND RISK FACTORS AS SELECTIVE SCREENING TOOLS FOR GESTATIONAL DIABETES MELLITUS IN COMPARISON WITH UNIVERSAL DIAGNOSING

Abstract

Objective Despite the short-term and long-term health implications of gestational diabetes mellitus (GDM), opinions are divided on selective vis-à-vis universal screening. We validated the accuracy of screening tests for GDM. Research design and methods Pregnant women (n=491) were recruited to this prospective, blind comparison with a gold standard study. We did selective screening between 13 and 20 weeks using reagent-strip glycosuria, random capillary blood glucose (RBG) and the presence of ≥1 risk factor(s). Between 20 and 34 weeks, we did universal screening following the ‘one-step’ approach using glycated hemoglobin (HbA1c), fasting venous plasma glucose (FPG), and the 1-hour and the ‘gold standard’ 2-hour oral glucose tolerance test (OGTT). Tests accuracy was estimated following the WHO and the National Institute for Health and Care Excellence (NICE) diagnostic criteria. Overall test performance was determined from the area under the receiver operating characteristic curve (AUC). Results GDM prevalence per 2-hour OGTT was 9.0% for the WHO criteria and 14.3% for the NICE criteria. Selective screening using glycosuria, RBG and risk factors missed 97.4%, 87.2% and 45.7% of cases, respectively. FPG threshold ≥5.1 mmol/L had the highest clinically relevant sensitivity (68%) and specificity (81%), but FPG threshold ≥5.6 mmol/L had higher positive predictive value. Although sensitivity of 1-hour OGTT was 39.5%, it had the highest accuracy and diagnostic OR. Regarding test performance, 1-hour OGTT and FPG were very good (AUC>0.8), RBG was poor (AUC≈0.60), whereas HbA1c was invaluable (AUC<0.5). Conclusions Selective screening using glycosuria and random blood glucose is unnecessary due to its low sensitivity. Fasting glucose ≥5.1 mmol/L could be applicable for screening at the population level. Where 2-hour OGTT is not available, FPG ≥5.6 mmol/L, complemented by the presence of risk factors, could be useful in making therapeutic decision.