Random Blood Glucose Measurement at Antenatal Booking to Screen for Overt Diabetes in Pregnancy: A retrospective study

A venous plasma RBG measurement at antenatal booking is included in a universal screening program in our hospital (4). Women with a booking RBG >7.0 mmol/L or past history of gestational diabetes (GDM) are offered a 75-g oral glucose tolerance test (OGTT) using either venous or capillary sampling. All women without identified GDM are screened again at 26–28 weeks with a 50-g oral glucose challenge test, with OGTT if the plasma glucose is >7.7 mmol/L. Additional OGTTs are offered where clinically indicated (e.g., macrosomia).

Retrospective demographic, obstetric, glucose, and neonatal data from births in 2004–2008 were obtained from hospital records within an approved service evaluation into the value of RBG screening. Screening RBG was defined as those requested between 0 and 20 weeks’ gestation to reflect current practice. If more than one RBG was identified, the highest value was recorded. OGTTs performed at any time during gestation were included. A diagnosis of ODIP by OGTT was defined based upon current World Health Organization plasma glucose criteria for diabetes (fasting ≥7.0 mmol/L and/or 2 h ≥11.1 mmol/L), grouping venous and capillary data. Women with known preexisting diabetes recorded were excluded.

Samples were collected using standard fluoride-containing tubes and analyzed in the hospital laboratory using a hexokinase-glucose-6-phosphate dehydrogenase method (Dimension RXL MAX Clinical Chemistry System; provided by Siemens Healthcare Diagnostics, Deerfield, IL).

Receiver operator curves (ROCs) of sensitivity plotted against 1-specificity were constructed. The area under the curve (AUC) of each ROC was calculated with 95% CIs to assess the diagnostic ability of the RBG to predict ODIP. Optimum RBG cutoff points were evaluated using R software (R Foundation for Statistical Computing, Vienna, Austria). All other analyses used SPSS (PASW for Windows, rel. 18.0.3, 2010; SPSS Inc., Chicago, IL).

Records were obtained for 26,369 live births. No maternal data could be matched in 506 cases. There were 17,852 screening RBG data, 18.6% of whom had RBG >7.0 mmol/L. Twelve women without OGTT had RBG ≥11.1 mmol/L; 3,007 women had an antenatal OGTT at some point during the pregnancy. The clinical characteristics of women with and without all data are shown in Supplementary Table 1.

The first ROC (Fig. 1) included all 17,852 RBG data and assumed that women without a positive OGTT did not have ODIP; 67 women had ODIP. The AUC was 0.86 (0.80–0.92) with a negative predictive value (NPV) of 0.999 and a positive predictive value (PPV) of 0.020. The best RBG cutoff was 7.31–7.40 mmol/L (corresponding to a sensitivity of 0.78 and specificity of 0.85).

To estimate the maximum diagnostic value of the RBG, a second ROC (not shown) was constructed assuming that those with no/incomplete OGTT and an RBG <11.1 mmol/L did not have ODIP, but that the 12 women with no OGTT but RBG ≥11.1 mmol/L had ODIP. The AUC was 0.88 (0.83–0.93), NPV was 0.999, PPV was 0.028, and the best RBG cutoff was 7.51–7.59 mmol/L (sensitivity 0.80; specificity 0.88). Although this uses a test (RBG) to predict itself, in clinical practice, RBG can be so high that no OGTT is performed. However, among the 87 women with RBG ≥11.1 mmol/L and an OGTT, only 30% had ODIP. To estimate the minimum diagnostic value of the RBG, a third ROC (not shown) was constructed among only those with both RBG and OGTT; 67 had ODIP. AUC was 0.72 (0.64–0.79), NPV was 0.988, PPV was 0.052, and the best RBG cutoff points were 8.60–8.70 mmol/L (sensitivity 0.60; specificity 0.75).

RBG sampling is convenient in high volume antenatal clinics and appears to be a useful test at booking to detect ODIP. Early studies suggested that RBG screening at 28–32 weeks was efficient for detecting GDM (5). However, OGTTs were not performed among women with RBG below their thresholds, thereby enhancing test utility. (In contrast, our study included many women with booking RBG ≤7.0 mmol/L who had an OGTT later in pregnancy.) Subsequent studies suggested that third-trimester RBG had limited predictive power to detect GDM (6,7). However, our study aimed to identify ODIP, rather than GDM, comparable to RBG and HbA_1c being poor at detecting impaired glucose tolerance but much better at identifying those with OGTT-defined diabetes in the nonpregnant state (8).

Our retrospective data collection has limitations. We assumed that OGTTs were performed according to the standard hospital protocol, including confirmation of fasting and ideal sample handling/timing (9). Standard preservatives might not inhibit glycolysis completely (10). Some OGTT used capillary plasma (e.g., for ease of collection), and capillary results are slightly higher than venous glucose (11). The lack of homogeneity in blood sample handling is a limitation but may also reflect routine clinical practice.

The assumption that a woman without a positive OGTT did not have ODIP probably resulted in the exclusion of a few women. To assess the impact of significant hyperglycemia warranting immediate treatment, the second ROC assumed that all women with RBG ≥11.1 mmol/L had ODIP; the AUC effect was negligible. Bias may exist in the ROC as some women with booking RBG ≤7.0 mmol/L did not have an OGTT; this may overestimate the NPV of the ODIP testing strategy.

We cannot account for the low proportion of women with RBG ≥11.1 mmol/L and diabetes by OGTT, unless it reflects the nonfasting state and limitations in the reproducibility of RBG/OGTT. This supports the necessity for clearly defined criteria for ODIP diagnosis.

RBG at antenatal booking may provide a sensitive screening tool for the detection of ODIP. Although the RBG cutoff >7.0 mmol/L was reasonable, we do not feel our retrospective observational data are adequate to recommend changing the RBG action limit to proceed to OGTT. We recommend further studies and comparison with fasting glucose and HbA_1c.

D.S. acknowledges the support of the National Institute for Health Research Cambridge Biomedical Research Centre.

No potential conflicts of interest relevant to this article were reported.

D.C. contributed to study design and methods, collected and collated data, and wrote the manuscript. D.H. contributed to methodology and reviewed and contributed to the manuscript. C.M. collated data and wrote the manuscript. R.A.P. contributed to methods, collated data, performed statistical analyses, and interpreted data. H.R.M. reviewed and advised on significant modifications to the manuscript. D.S. conceived the study design and contributed to methodology, made significant contribution to data analysis, and wrote the manuscript.

Parts of this study were presented in abstract form at the Diabetes UK Annual Professional Conference 2011, London, U.K., 30 March–1 April 2011.

The authors acknowledge Emma Te Braake, Kerry Stubbington (The Rosie Hospital, Addenbrooke’s Hospital), and Michaela Wilson (Wolfson Diabetes and Endocrinology Clinic) for their contributions to data collection and Andi Thornton and Adam Herrick (Information Technology Services, Addenbrooke’s Hospital) for their contributions to data extraction.