Diagnostic criteria for gestational diabetes in relation to pregnancy outcome

Journal of Diabetes and Its Complications 17 (2003) 115 – 119

Diagnostic criteria for gestational diabetes in relation to pregnancy outcome Martha S. de Seredaya,*, Mo´nica M. Damianob, Claudio D. Gonza´lezc, Peter H. Bennettd a

Argentine Diabetes Society, Buenos Aires, Argentina Endocrinology and Diabetes Unit, Pedro Fiorito Hospital, Avellaneda, Buenos Aires, Argentina c Department of Pharmacology, School of Medicine, University of Buenos Aires, Buenos Aires, Argentina d National Institutes of Health, Phoenix, Arizona, USA b

Received 23 October 2001; received in revised form 1 March 2002; accepted 7 March 2002

Abstract Objectives: To determine which of the American Diabetes Association (ADA) or World Health Organization (WHO) plasma glucose criteria for gestational diabetes mellitus (GDM) best predicts poor fetal outcome. To determine whether an alternative cut-off point would result in increased predictive value and greater diagnostic effectiveness in pregnancies at high risk for GDM. Research design and methods: A sample of 473 successive apparently normal pregnant women attending the Obstetric Department were screened for GDM using both the ADA and the WHO criteria. Between 26 and 30 weeks of gestation, they underwent, on subsequent days, a screening test with a 50-g oral glucose load and two oral glucose tolerance tests (OGTTs) with 75 and 100 g of glucose according to the WHO and the ADA recommendations, respectively. From this group, we identified 99 women at high risk for GDM, who did not attend their pregnancy follow-up and whose delivery records were recovered at our hospital or in neighbouring hospitals. This unusual situation enabled us to study the natural history and outcome of their pregnancy in spite of not receiving special management usually provided to such women. As macrosomia was expected to be the most frequent undesirable foetal outcome, sensitivity and specificity calculations have been based on this outcome. Results: The study population (n = 99) had a median parity of two and 14% had abnormal results in the 2-h, 75-g load test (WHO) vs. 6% in the 100-g test (ADA). Optimal cut-off points for each test were lower than those recommended for diagnosis by the ADA and the WHO. The optimal sensitivity for the 1-h, 50-g test was 66.7% (cut-off 137 mg/dl), and for the 2-h, 75-g test (cut-off 119 mg/dl). The best specificity and positive predictive value was for this last test with a cut-off point of 140 mg/dl in the second hour. Conclusions: The standard 2-h cut-off value of 140 mg/dl for the 75-g test, as now recommended by WHO, was optimal for predicting macrosomia. Based on the sensitivity and specificity for macrosomia, the 1-h, 50-g screening test had an optimal cut-off point of 137 mg/dl (vs. 140 mg/dl recommended by ADA). The 2-h, 75-g OGTT value using a cut-off point of 119 mg/dl had equivalent sensitivity, specificity, and positive predictive value. In contrast, the 100-g OGTT had much lower levels of sensitivity, but higher specificity and higher positive predictive value. D 2003 Elsevier Science Inc. All rights reserved. Keywords: Gestational diabetes; Diagnostic criteria; Macrosomia; Cut-off points; Predictive value

1. Introduction Gestational diabetes mellitus (GDM) as defined by the American Diabetes Association (ADA) and the World Health Organization (WHO) is ‘‘any degree of glucose

Abbreviations: ADA, American Diabetes Association; WHO, World Health Organization; GDM, gestational diabetes mellitus; OGTT, oral glucose tolerance test; DM, diabetes mellitus; ROC, receiver operating characteristics; BMI, body mass index; AUC, area under the curve * Corresponding author. Rodrı´guez Pen˜a 1838-1
, piso ‘‘4,’’ CP 1014 Capital Federal, Argentina. Tel./fax: +54-11-4813-9535. E-mail address: [email protected] (M.S. de Sereday).

intolerance with onset or first recognition during pregnancy’’ (The Expert Committee on the Diagnosis and Classification of Diabetes Mellitus, 1997; National Diabetes Data Group [NDDG], 1979; O’Sullivan & Mahan, 1964). In spite of defining it in the same terms, the recommended criteria and methods for diagnosis differ. The ADA Expert Committee (1997) recommends a screening test performed with a 50-g glucose load between 24 and 28 weeks of gestation. A glucose value  140 mg/dl after 1 h indicates the need for a 3-h oral glucose tolerance test (OGTT) with 100-g load. The results of the test are considered abnormal if two of the four plasma glucose values exceeded the normal value. These values were proposed by

1056-8727/03/$ – see front matter D 2003 Elsevier Science Inc. All rights reserved. doi:10.1016/S1056-8727(02)00173-3

116

M.S. de Sereday et al. / Journal of Diabetes and Its Complications 17 (2003) 115–119

O’Sullivan and Mahan in 1964 and were converted to plasma values by the NDDG in 1979 (in mg/dl): fasting, 105; 1 h, 190; 2 h, 165; and 3 h, 145. Discrepancies arose because of the interpretation of O’Sullivan and Mahan’s values. Not only the substrate measured switched from whole venous blood to venous plasma, but also the laboratory technique switched from Somogyi– Nelson method to enzymatic ones. Furthermore, the NDDG approach based on O’Sullivan and Mahan’s values were rounded differently from those subsequently calculated by Carpenter and Coustan (1982). To complicate the situation further, the Second and Third International Workshops on Diabetes in Pregnancy advised the revision of the values and the Fourth Workshop recommended the use of the plasma glucose values proposed by Carpenter and Coustan (in mg/dl): fasting, 95; 1 h, 180; 2 h, 155; and 3 h, 140 (Proceedings of the Fourth International Workshop – Conference on Gestational Diabetes Mellitus, 1998). To diagnose GDM, the WHO states that a standard OGTT with 75 g of glucose should be performed (Definition, Diagnosis and Classification of Diabetes Mellitus and Its Complications, 1985) using the same criteria as for nonpregnant women. Furthermore, the recent 1999 WHO recommendations (Definition, Diagnosis and Classification of Diabetes Mellitus and Its Complications, 1999) indicate that women meeting either criteria for diabetes or impaired glucose tolerance be considered to have gestational diabetes. There has been much criticism of both criteria (McCance et al., 1995; Naylor, 1989; Pettitt, Bennett, Hanson, Narayan, & Knowler, 1994; Sacks, Abu-Fadil, Greenspoon, & Fotheringham, 1989), but so far there is not enough evidence to either support or change them. We studied 99 pregnant women from the Obstetric Department of Hospital Fiorito, Avellaneda, Argentina to determine which of the two criteria — ADA or WHO — better predict poor foetal outcome. We also investigated whether an alternative cut-off point would result in increased predictive value and greater diagnostic effectiveness in pregnancies at high risk for GDM.

2. Research design and methods Out of successive pregnant women attending the Obstetric Unit during a 3-month period, 473 apparently normal Table 1 General characteristics of the study population (n = 99)

Age (years) BMI (kg/m2) Gestational age at test (weeks) Number of children born alive Fasting plasma glucose (mg/dl) 1-h, 50-g test (mg/dl) 2-h, 75-g test, (mg/dl) 1-h, 100-g test, (mg/dl) 2-h, 100-g test, (mg/dl) 3-h, 100-g test, (mg/dl)

Mean

S.D.

28.2 30.8 27.4 2.2 73 134.2 115.1 160.4 125.9 88.7

6.1 5.6 5.9 2.8 12.6 36.9 32.1 45.1 33.1 24.3

Table 2 Foetal outcomes Number Perinatal mortality Macrosomia Foetal malformations Premature birth Caesarian section Gestational age at birth (weeks) Birth weight (g) Length (cm)

1 12 0 1 22 39.3 ± 1.4 3462.9 ± 517.9 50.6 ± 2.2

pregnant women were selected following this exclusion criteria: younger than 18 years of age; personal record of gestational or clinical diabetes; presence of hepatic, renal, infectious or endocrine disease. These women were mainly of low economic and educational level. The group was predominantly Caucasian and 7% had a quarter of mixed Indian origin. Between 26 and 30 weeks of gestation, they underwent on three alternative days over a period of 10 days a screening test with 50 g of glucose load and two OGTTs with 75 and 100 g of glucose according to the WHO and the ADA recommendations (Carpenter & Coustan, 1982; The Expert Committee on the Diagnosis and Classification of Diabetes Mellitus, 1997), respectively. The order of the OGTTs over the examination days was according to increasing glucose load (i.e., first screening test: 50 g, second: 75 g OGTT, and third: 100 g OGTT). Glucose in venous plasma was measured by glucose oxidase method. Two hundred sixteen women out of the 473 were considered at high risk for GDM on the basis of the following risk factors: history of macrosomia, malformation, or foetal death; obesity — body mass index (BMI)  30 kg/m2 (WHO criteria) — in previous pregnancy; multiparity — three or more deliveries; family history of DM in first- or second-degree relatives. From this group, we were able to identify 99 women at high risk for GDM who did not attend their pregnancy follow-up and whose delivery records were recovered at our hospital or in neighbouring ones. This unusual situation enabled us to study the natural history and outcome of their pregnancy in spite of not receiving special management usually provided to such women. The 99 pregnant women were between 26 and 30 weeks gestation. The record for each woman included: demographic data, family history of diabetes in first- and second-degree relatives, personal history of obesity and hypertension, personal obstetric history number of pregnancies, obesity, and morbidity in previous pregnancies, abnormalities of the outcome (foetal macrosomia, congenital abnormalities, stillbirth, and perinatal morbidity and mortality), and characteristics of the current pregnancy (date of last menstrual period, trimester of pregnancy, height, initial, and current weight). As macrosomia was expected to be the most frequent undesirable foetal outcome, sensitivity and specificity

M.S. de Sereday et al. / Journal of Diabetes and Its Complications 17 (2003) 115–119 Table 3 ROC curves optimal cut-off points

1-h, 50-g test (mg/dl) 2-h, 75-g test (mg/dl) 2-h, 100-g test (mg/dl) 3-h, 100-g test, (mg/dl) Mother’s age (years) Mother’s BMI (kg/m2) Parity (children born alive)

Cut-off point

AUCa ± S.E.

P value

137.0 119.0 126.0 88.0 30.0 30.0 3 or more

0.747 ± 0.06 0.703 ± 0.08 0.589 ± 0.11 0.533 ± 0.12 0.676 ± 0.08 0.637 ± 0.06 0.726 ± 0.07

.0004 .02 .418 .786 .03 .06 .0007

End-point macrosomia. a AUC: Area under the curve of each ROC curve.

117

Table 5 WHO 75-g second hour glycemia and multiparity and macrosomia. Results of the logistic regressiona Variable

b coefficient ± S.E.

Adjusted OR (95% CI)

P value

WHO second hour (  120 mg/dl) Multiparity (three or more children) Obesity (BMI  30) Obstetric personal history (yes/no)

1.960 ± 0.828

7.10 (1.40 – 36.1)

.020

1.689 ± 0.747

5.42 (1.25 – 23.5)

.026

1.190 ± 0.855 0.004 ± 0.785

3.29 (0.61 – 17.6) 1.00 (0.23 – 4.44)

.996 .995

a Multiple logistic regression (maximum likelihood, quasi-Newton technique); explained variance: 21.0%; Hosmer – Lemeshow test P = .612.

calculations have been based on this outcome. We considered macrosomia a birth weight  4000 g between 38 and 42 weeks gestation. 2.1. Statistical analysis Quantitative variables are presented as mean ± standard deviation and the median value was also determined for parity. Considering macrosomia as the main foetal outcome, receiver operating characteristics (ROC) analysis were performed to determine optimal cut-off points for the glucose levels corresponding to the 1-h, 50-g test value, the 2-h, 75-g test value, and the 2- and 3-h, 100-g test values, maternal age, BMI, and parity. The significance was assessed calculating the corresponding area under the curve (AUC) for each ROC curve. A stepwise logistic regression analysis was carried out to assess the best predictive model for macrosomia (loss function: maximum likelihood; estimation method: quasi-Newton). The Hosmer – Lemeshow test (1989) was performed to check the goodness-offit within the logistic model. Significance was considered at a P level below .05 (two-tailed).

values of plasma glucose for each and the gestational age and baby size at birth are presented in Table 1. There were 22 caesarian sections (Table 2), a rate similar to that of the general population at our hospital (20%). There were no foetal malformations and macrosomia was the most frequent foetal outcome occurring in 12 women. Optimal cut-off points for the prediction of macrosomia applied to each GDM diagnostic criteria, mother’s age, BMI, and parity are shown in Table 3. The optimal cut-off points for each test were lower than those recommended for diagnosis by the ADA and the WHO. Considering the ADA, the WHO, and our own cut-off points (Table 4), the optimal sensitivity — 66.7% — was for the 1-h, 50-g test (cut-off 137 mg/dl) and for the 2-h, 75-g test (cut-off 119 mg/dl). The best specificity and positive predictive value was for this last test with a cut-off point of 140 mg/dl in the second hour. By logistic regression, the best predictive model for macrosomia is shown in Table 5. The model explained 21.0% of the variance. Multiparity (three or more) and 2-h, 75-g test glycemia (above 120 mg/dl) were significantly

3. Results The study population (n = 99) had a median parity of two and 14% had abnormal results in the 2-h, 75-g load test (i.e., 2-h value; WHO) vs. 6% in the 100-g test (ADA). The mean

Table 4 Sensitivity, specificity, and positive predictive value (PPV)

1-h, 50-g test (cut-off 140 mg/dl) 1-h, 50-g test (cut-off 137 mg/dl) 2-h, 75-g test (cut-off 140 mg/dl) 2-h, 75-g test (cut-off 119 mg/dl) 100 g, at least one abnormal value 100 g, at least two abnormal values Multiparity (three or more children) Multiparity (four or more children)

Sensitivity (%)

Specificity (%)

PPV (%)

58.3 66.7 41.7 66.7 36.4 27.3 66.6 50.0

67.8 63.2 90.8 64.4 80.2 96.5 72.4 81.6

20.0 20.0 38.5 20.5 19.0 50.0 25.0 27.2

End-point macrosomia applying the different recommended and optimal cut-off points.

Fig. 1. Probability of macrosomia according to second hour glucose values following a 75-g OGTT (WHO) and parity.

118

M.S. de Sereday et al. / Journal of Diabetes and Its Complications 17 (2003) 115–119

associated with macrosomia; the effect of multiparity is also shown in Fig. 1.

4. Discussion Screening for GDM is controversial. In some countries, screening is not done and it is not generally agreed whether it meets the criteria for a beneficial screening activity as set forward by Wilson and Jungner (1968). Moreover, even if screening is regarded as beneficial, there is controversy on how to proceed. Ideally, GDM screening should detect persons at risk for undesirable outcomes, so that intervention could take place at an early stage to prevent foetal complications. The aim of the present study was to determine which of the oral glucose tests recommended by the ADA and WHO for screening for GDM is more effective in identifying women at high risk of unfavourable pregnancy outcome, and to determine optimal glucose values for identifying such women, with the main outcome measure being macrosomia. In the present study, we determined the pregnancy outcome in the 99 women who did not received specific treatment even when diagnosed as having GDM by OGTT (by either current ADA or WHO criteria). Overall, if the tests are considered as single entities, the 50-g test had the highest sensitivity when administered to the present study sample. However, specificity was greatest with the 75-g glucose tolerance test, whereas the 100-g test had lower sensitivity. While sensitivity and specificity have been optimised mathematically in this analysis, it is important to consider the consequences of being a false-negative or a false-positive. A low rate of false-positives is desirable, because of the social and cost impact of treating people who are not likely to develop foetal complications, and thus achieve favourable cost effectiveness. A slightly higher false-negative rate than used in this analysis may be desirable as there are other means, such as ultrasound, of identifying features such as macrosomia that are not dependent on glycemic testing. For practical reasons, a single test to identify women at high risk of GDM is desirable, as many may not return for follow-up if a multistep screening approach is adopted, and thus such women may fail to be identified as having a high-risk pregnancy. On these grounds, the 75-g test, as currently recommended by WHO, would be preferable. The second consideration is whether the standard 2-h, cut-off value of 140 mg/dl for the 75-g test was optimal for predicting macrosomia. Using a ROC curve analysis, we have shown that a cut-off at 120 mg/dl would maximise both its sensitivity and its specificity. However, even with this optimisation, sensitivity and specificity were not very high. Based on the sensitivity and specificities for macrosomia, the 1-h, 50-g initial screening test had an optimal cut-off point of 137 mg/dl (vs. 140 mg/dl recommended by ADA). The 2-h, 75-g OGTT value using a cut-off point of 119 mg/dl had an equivalent sensitivity, specificity, and positive pre-

dictive value. In contrast, the 100-g OGTT had much lower levels of sensitivity, but higher specificity and higher positive predictive value. A potential shortcoming of the present study is the fact that the tests were given in a systematic order. The tests were given in order of ascending glucose load, thus probably minimising the effects of the previous test on the subsequent one. Nevertheless, the possibility that the previous test may have influenced the results of the subsequent test cannot be excluded. Furthermore, the fact that these women did not return for prenatal follow-up following the tests raises the question as to whether or not their outcomes are representative of what would have occurred among other women who receive the test if they had not received subsequent care. A combined strategy using clinical risk stratification as an indication to perform the OGTT might be very useful. For instance, using multiparity, mother’s age, and weight as clinical predictors and as indication for the performance of an OGTT could lead to more efficient, sensitive, and reasonably specific means to identify the majority of at risk pregnancies. Nevertheless, the small size and the nature of the present sample limit the ability to test this possibility and reinforce the need of further more extensive studies to determine the best method to predict poor outcomes in GDM.

Acknowledgments We are very grateful to the Advanced Epidemiology Course, Etiologic Factors Group (E. Feskens — Chairperson, C. Baan, C. Foss, A. Buyken, D. Williams, T. de Bruin, N. Wareham), The Netherlands, 1998, for their criticism and comments; to Dr. L Kwiatkowski, Head of the Obstetric Department, Hospital Fiorito, Avellaneda, for his contribution; and to Dr. Jorge Alvarin˜as for the revision of the manuscript.

References Carpenter, M. W., & Coustan, D. R. (1982). Criteria for screening tests for gestational diabetes. American Journal of Obstetrics and Gynecology, 144, 768 – 773. Definition, diagnosis and classification of diabetes mellitus and its complications. (1985). Report of a WHO Consultation. Geneva: WHO. Definition, diagnosis and classification of diabetes mellitus and its complications. (1999). Report of a WHO Consultation. Geneva: WHO. The Expert Committee on the Diagnosis and Classification of Diabetes Mellitus (ADA). (1997). Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Diabetes Care, 20 (7), 1183 – 1197. Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic regression. New York: Wiley. McCance, D. R., Hanson, R. L., Charles, M. A., Jacobsson, L. T. H., Pettitt, D. J., Bennett, P. H., & Knowler, W. C. (1995). Which test for diagnosing diabetes? Diabetes Care, 18 (7), 1042 – 1044. National Diabetes Data Group. (1979). Classification and diagnosis of

M.S. de Sereday et al. / Journal of Diabetes and Its Complications 17 (2003) 115–119 diabetes mellitus and other categories of glucose intolerance. Diabetes, 28, 1039 – 1057. Naylor, C. D. (1989). Diagnosing gestational diabetes mellitus is the gold standard valid? Diabetes Care, 12 (8), 565 – 572. O’Sullivan, J. B., & Mahan, C. M. (1964). Criteria for the oral glucose tolerance test in pregnancy. Diabetes, 13 (3), 278 – 285. Pettitt, D., Bennett, P. H., Hanson, R. L., Narayan, K. M. V., & Knowler, W. C. (1994). Comparison of World Health Organization and National Diabetes Data Group procedures to detect abnormalities of glucose tolerance during pregnancy. Diabetes Care, 17 (11), 1264 – 1268.

119

Proceedings of the Fourth International Workshop – Conference on Gestational Diabetes Mellitus. Diabetes Care, 21 (Suppl. 2), B162 – B163. Sacks, D. A., Abu-Fadil, S., Greenspoon, J. S., & Fotheringham, N. (1989). Do the current standards for glucose tolerance testing in pregnancy represent a valid conversion of O’Sullivan’s original criteria? American Journal of Obstetrics and Gynecology, 161, 638 – 641. Wilson, J. M. G., & Jungner, G. (1968). Principles and practice of screening for disease. Public Health Papers 34. Geneva: World Health Organization.