Volume 83, Issue 6 p. 524-530
Free Access

A population-based risk factor scoring will decrease unnecessary testing for the diagnosis of gestational diabetes mellitus

Eray Caliskan

Corresponding Author

Eray Caliskan

From the SSK Ankara Maternity and Women's Health Teaching Hospital, Etlik, Ankara, Turkey

*Eray Caliskan
Adnan Kahveci cd
Menekse sok
No: 37 Madenler Yapi Koop
A/2 Blok D: 9
Yenikent, Derince, Kocaeli
Turkey
e-mail: [email protected]Search for more papers by this author
Fulya Kayikcioglu

Fulya Kayikcioglu

From the SSK Ankara Maternity and Women's Health Teaching Hospital, Etlik, Ankara, Turkey

Search for more papers by this author
Nilgun Öztürk

Nilgun Öztürk

From the SSK Ankara Maternity and Women's Health Teaching Hospital, Etlik, Ankara, Turkey

Search for more papers by this author
Sevgi Koc

Sevgi Koc

From the SSK Ankara Maternity and Women's Health Teaching Hospital, Etlik, Ankara, Turkey

Search for more papers by this author
Ali Haberal

Ali Haberal

From the SSK Ankara Maternity and Women's Health Teaching Hospital, Etlik, Ankara, Turkey

Search for more papers by this author
First published: 10 May 2004
Citations: 32

Abstract

Background. To determine the effectiveness of a population-based risk factor scoring to decrease unnecessary testing for the diagnosis of gestational diabetes mellitus (GDM).

Methods. We formed a risk factor scoring over five, which questions maternal age, body mass index and first-degree relatives with a diagnosis of diabetes mellitus, a prior macrosomic fetus and adverse outcome during the previous pregnancies. All participants underwent a 50-g glucose challenge test (GCT) followed by a 100-g oral glucose tolerence test (OGTT). We opened the 50-g GCT envelope if the participant had a risk score ≥ 1 and opened the 100-g OGTT envelope if the 50-g GCT value was ≥ 7.2 mmol/l. After all patients delivered we also built other strategies and tested their detection rates.

Results. Fourteen patients (3.3%) were diagnosed as having gestational diabetes mellitus via a 100-g OGTT. None of the patients with a score of zero had gestational diabetes mellitus. Logistic regression analysis revealed that an increase in the score by one caused a three times increase in gestational diabetes mellitus risk (OR = 3, CI = 1.9–5). Compared with the universal screening, our strategy to screen if the risk score was ≥ 1, followed by a 50-g GCT with a 7.2-mmol/l cut-off value, decreased the number of women to be screened by 30% and diagnosed all cases with GDM. Screening the patients with a score ≥ 2 would have decreased the number of women to be screened by 63%, still diagnosing 85% of cases with GDM. Also, risk factor-based screening strategies cause a 50% and 53% reduction in the number of OGTT applied, respectively.

Conclusion. A well integrated, population-based scoring will decrease the number of unnecessary testing but still diagnose 85–100% of GDM cases.

Abbreviations:

  • GDM
  • gestational diabetes mellitus
  • GCT
  • glucose challenge test
  • OGTT
  • oral glucose tolerance test
  • BMI
  • body mass index
  • WHO
  • World Health Organization
  • ACOG
  • American Collage of Obstetrics and Gynecology
  • ADA
  • American Diabetes Association
  • Gestational diabetes mellitus (GDM) is defined as glucose intolerance with the onset or first detection during pregnancy (1). The incidence of GDM varies between 0.6% and 15% according to the population surveyed and diagnostic test used (2). Today, 5% of United Kingdom and 12% of United States total healthcare expenditure is spent on diabetes and its complications (3).

    Screening for gestational diabetes provides the opportunity to identify the women at risk of future non-insulin dependent diabetes and to identify the adolescents at risk of glucose intolerance that were born to diabetic mothers (3,4). Besides many adverse maternal and fetal outcomes published in women with GDM, there is lack of consensus regarding how to screen and the threshold value of the screening test used.

    In this study, we aimed to test whether a prospective application of a risk factor scoring, based on a retrospective case–control study conducted in a Turkish population, can be an alternative screening strategy for diagnosing GDM.

    Materials and methods

    Identifying the risk factors

    In order to decide on the risk factors, which increase the frequency of gestational diabetes mellitus (GDM), a retrospective study was performed (unpublished data). All births from August to December 1999 were surveyed. During this time 143 women (3.1%) out of 4612 parturients were identified to have GDM. The data of GDM patients were than compared with 286 randomly selected controls for frequency of risk factors proposed in the literature. The frequency of recurrent spontaneous abortions (>2 abortions), in utero fetal death at a gestational age ≥ 20 weeks, fetal anomaly despite a normal karyotype, intrauterine growth restriction (< 10 percentile), low birth weight newborn (< 2500 g), macrosomic infant (> 4000 g), pre-eclampsia, chronic hypertension and multiparity were compared in the two groups using the chi-square test. Receiver operator curve characteristics of continuous variables such as maternal age and pregravid body mass index (BMI) were identified to decide on the appropriate cut-off value for diagnosing GDM. Logistic regression analysis using the Forward Wald method did not identify prior intrauterine growth restriction, pre-eclampsia, chronic hypertension, low birth weight newborn and multiparity as significant risk factors.

    Population-based scoring

    As a result of this study, factors that were identified to be significantly more common, in GDM patients in our institution, which remained in the logistic regression model, were included in the risk scoring system. For ease of scoring, adverse obstetric outcome was defined as the presence of any of the following: recurrent spontaneous abortions, fetal anomaly despite a normal karyotype and prior unexplained in utero fetal death at a gestational age ≥ 20 weeks. A score of one was assigned for the presence of each of the five variables: maternal age ≥ 25, BMI ≥ 25 kg/m2, diabetes in first degree relatives, prior macrosomic infant and a history of adverse obstetric outcome.

    Patient selection and setting

    The present study was conducted between May 2000 and July 2000 in the antenatal policlinics of the Social Security Council Maternity and Women's Health Teaching Hospital. Patients having singleton pregnancies between 24 and 28 weeks of gestation without a previous diagnosis of diabetes mellitus were invited to participate in the study. All women with suitable criteria were informed about the study. They were offered 24-h access to personal cellular phones of the authors and pregnancy follow-up without a need for booking. One hundred percent acceptance was maintained and informed consent was obtained from 425 pregnant women.

    All participants received a 50 g, 1-h glucose challenge test (GCT). This test was applied at any time of the day but mostly in the morning without an overnight fasting. Venous blood was sampled at the 60th minute. The data obtained were stored and also provided to the patient in sealed opaque envelopes with a ‘50-g GCT’ label and the patient's name on it. All patients than had a 100-g oral glucose tolerance test (OGTT) 1 week after the 50-g GCT. For the 100-g OGTT, patients had a 150-g-carbohydrate diet per day for 3 days and venous blood was sampled in the morning after an overnight fasting. Patients received 100 g of glucose in 200 cc water and venous blood was sampled at 60, 120 and 180 min. The data obtained were stored and provided to the patient in sealed opaque envelopes with a ‘100-g OGTT’ label and the patient's name on it. The third author retained all envelopes. Three cases that could not tolerate a 100-g OGTT were excluded from the study and were noticed to have a 50-g GCT value < 7.2 mmol/l.

    After the 100-g OGTT all patients were scored by the first author over five, according to the population-based scoring. If the patient received a score ≥ 1 the 50-g GCT labeled envelope of the patient was opened. Whenever the value of the 50-g GCT was ≥ 7.2 mmol/l (130 mg/dl) then the 100-g OGTT labeled envelope of the same patient was also opened. GDM was diagnosed according to the criteria of the National Diabetes Data Group if any two measurements were higher than or equal to the cut-off values: fasting 5.8 mmol/l (105 mg/dl), 60th minute 10.6 mmol/l (190 mg/dl), 120th minute 9.2 mmol/l (165 mg/dl) and 180th minute 8.1 mmol/l (145 mg/dl). GDM was managed with diet or insulin therapy according to the maternal glycemic control.

    Both the 50-g GCT and the 100-g OGTT results of the patients with a score of 0 and the 100-g OGTT results of the patients with a score ≥ 1 but the 50-g GCT value < 7.2 mmol/l remained in sealed envelopes until all 422 women included in the study gave birth.

    Statistical analysis

    For statistical analysis of the data SPSS statistical package for social sciences, release 11.0, Chicago, Illinois (USA) was used. Logistic regression analysis was used to determine the association between the risk score and GDM. Sensitivity, specificity, positive predictive value and negative predictive value of the strategy applied in this study and other empiric strategies were calculated with 95% confidence intervals.

    Results

    The mean maternal age of the study group was 24.9 ± 4.9 (16–40) while the mean BMI was 23.9 ± 3.5(16.9–39.1). Multiparity was observed in 44.7% (n = 189) of the cases. Forty-eight percent of the cases (n = 205) were greater than 25 years of age while 31.5% (n = 133) had a BMI greater than 25. A history of adverse obstetric outcome was found in 3.5% of the cases (n = 15) and 4.5% (n = 19) had a history of a macrosomic fetus. A diagnosis of diabetes mellitus was present in first-degree relatives of 104 patients (24.6%).

    Fourteen patients (3.3%) were diagnosed as having gestational diabetes mellitus. Six patients had one abnormal value during the OGTT (1.4%). The distribution of risk factors with respect to GDM is presented in Table I. All risk factors were significantly more frequent in patients with GDM.

    Table I. The distribution of risk factors with respect to the presence of gestational diabetes mellitus (GDM)
    Risk factor GDM present n = 14 (%) GDM absent n = 408 (%) p
    Maternal age ≥ 25 years 12 (85.7) 193 (47.3)   0.01*
    Body Mass Index ≥25 (kg/m2) 11 (78.5) 122 (29.9) < 0.001*
    Prior adverse obstetric outcome 2 (14.2) 13 (3.1)   0.02*
    Family history of diabetes mellitus 10 (71.4) 94 (23) < 0.001*
    Prior macrosomic fetus 3 (21.4) 16 (3.9)   0.02*
    • Data are presented as numbers and percentages. GDM: gestational diabetes mellitus.
    • * Statistically significant (p < 0.05).

    The distribution of patients with GDM and one abnormal value during OGTT with regard to risk factor scoring is presented in Table II. None of the patients with a score of zero had GDM. Although not the primary outcome of the study, it was interesting to note that, all patients with one abnormal value during the OGTT had a score more than or equal to one. Logistic regression analysis revealed that an increase of the score by one caused a three times increase in GDM risk (OR = 3, CI = 1.9–5).

    Table II. The distribution of patients with GDM and patients with one abnormal value during a 100-g OGTT with regard to the risk factor scoring
    Risk score GDM Absent GDM Present One abnormal value during OGTT
    0 n = 125 (%) 125 (100)
    1 n = 140 (%) 137 (97.9) 2 (1.4) 1 (0.7)
    2 n = 93 (%) 88 (94.6) 4 (4.3) 1 (1.1)
    3 n = 48 (%) 43 (86.9) 3 (6.3) 2 (4.2)
    4 n = 14 (%) 8 (57.1) 4 (28.6) 2 (14.3)
    5 n = 2 (%) 1 (50) 1 (50)

    The screening strategy used in this study and other possible strategies, the number of patients screened and GDM cases diagnosed with the number of 100-g OGTT needed are given in Table III. The sensitivity, specificity, positive predictive value and negative predictive value of each strategy are given in Table IV. The screening strategy used in this study was to apply the 50-g GCT if the pregnant women had one or more risk factors and, apply the 100-g OGTT if the 50-g GCT value was bigger than or equal to 7.2 mmol/l. With this strategy, besides diagnosing all the patients with GDM, a 30% reduction in the number of women to screen and an approximately 5% reduction in the number of OGTT applied was achieved when compared with universal screening with a corresponding cut-off value.

    Table III. The number of patients screened and GDM diagnosed with different combination of tests
    Strategy Patients screened n= 422 (%) GDM diagnosed n = 14 (%) OGTT applied n = 422 (%)
    100-g OGTT 422 (100) 14 (100) 422 (100)
    Universal 50-g GCT with a 7.8 mmol/l cut-off value 422 (100) 13 (92.8) 109 (25.8)
    Universal 50-g GCT with a 7.2 mmol/l cut-off value 422 (100) 14 (100) 139 (32.9)
    Risk score ≥1 followed by a 50-g GCT with
    7.8 mmol/l cut-off value
    297 (70) 13 (92.8) 95 (22.5)
    Risk score ≥1 followed by a 50-g GCT with
    7.2 mmol/l cut-off value*
    297 (70) 14 (100) 119 (28.1)
    Risk score ≥2 followed by a 50-g GCT with
    7.8 mmol/l cut-off value
    157 (37) 12 (85.7) 59 (13.9)
    Risk score ≥2 followed by a 50-g GCT with
    7.2 mmol/l cut-off value
    157 (37) 12 (85.7) 70 (16.5)
    • * The strategy used in this study. GCT: glucose challenge test, OGTT: oral glucose tolerance test.
    • The data are presented as numbers and percentages.
    Table IV. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of different strategies with 95% confidence intervals (CI) in diagnosing GDM
    Strategy Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI)
    Universal 50-g GCT with a 7.8 mmol/l cut-off value 92.8 (79.3–100) 76.4 (72.5–80.5) 11.9 (5.9–17.9) 99.6 (99–100)
    Universal 50-g GCT with a 7.2 mmol/l cut-off value 100 69.4 (65–73.8) 10 (5.1–14.9) 100
    Risk score ≥1 followed by a 50-g GCT with a
    7.8 mol/l cut-off value
    92.8 (79.3–100) 71 (65.8–76.2) 13.6 (6.8–20.4) 99.5 (98.6–100)
    Risk score ≥1 followed by a 50-g GCT with a
    7.2 mol/l*
    100 62.8 (57.2–68.4) 11.7 (6–17.4) 100
    Risk score ≥2 followed by a 50-g GCT with a
    7.8 mol/l cut-off value
    85.7 (67.4–100) 67.1 (59.4–74.8) 20.3 (10.1–30.5) 97.9 (95.1–100)
    Risk score ≥2 followed by a 50-g GCT with a
    7.2 mol/l cut-off value
    85.7 (67.4–100) 59.4 (51.4–67.4) 17.1 (8.3–25.9 97.7 (94.6–100)
    • * The strategy used in this study. GCT: glucose challenge test.

    Screening the patients who have a risk factor score of 2 or more with the 50-g GCT will further decrease the number of patients screened by 63% and the number of 100-g OGTT applied by 49–55% depending on the cut-off value selected for the 50-g GCT. Also, this selective screening strategy could double the positive predictive value without causing a significant change in specificity and negative predictive value but a slight decrease in sensitivity. This screening strategy would have missed the two cases with a score of one. Case 1 has an age > 25 year old as a sole risk factor and case 2 has a BMI > 25. GDM in both cases were controlled by diet only. Case 1 gave birth to a small-for-gestational age infant, whose birth weight of 2400 g corresponded to the third percentile, via cesarean section for transverse lie. Case 2 gave birth to a large-for-gestational age infant, whose birth weight of 4100 g corresponded to 97th percentile, via the vaginal route. Postpartum follow-up of the newborn of case 1 was uneventful while the newborn of case 2 had a hypoglycemic attack.

    Selective screening of patients with a risk factor score of 2 or more would also identify, 83.3% (n = 5/6) of the cases with one abnormal value during OGTT (Table II). Among these, the case with only one risk factor had a BMI > 25, gave birth to a 3600 g appropriate for gestational age infant. The route of delivery was vaginal and neonatal follow-up was uneventful.

    Discussion

    The earliest and simplest test for gestational diabetes mellitus is taking of a history (5). Asking the presence of several risk factors is the easiest and cheapest screening test to identify pregnant women at high risk of having GDM. However, universal screening for gestational diabetes with a glucose challenge test has been endorsed by the Second and Third International Workshop Conferences on Gestational Diabetes (6,7) and the American Diabetes Association (8).

    Recently, Brody et al. (9) in their review of the evidence for screening GDM concluded that there is insufficient data to determine the magnitude of health benefit for any treatment among the large number of women with GDM and, identification of GDM may also needlessly increase the use of non-stress testing or biophysical profile and rates of cesarean delivery. The U.S. Preventive Services Task Force therefore recommended either not to screen at all or to screen only women at increased risk of GDM (10). Recommending either universal or selective screening the Fourth International Workshop on Gestational Diabetes (1), the American Diabetes Association (ADA) (11) and the American College of Obstetricians and Gynecologists (ACOG) (12) as the U.S. Preventive Services Task Force (10) consider women to be at risk of GDM if they are ≥ 25 years of age, have a body mass index ≥ 25, have a family history of diabetes or a personal history of abnormal glucose metabolism, have a history of poor obstetric outcome such as anomalous or a stillborn infant and have a prior macrosomic infant.

    These wide varieties of screening strategies reflect the lack of evidence that, benefits of screening for GDM outweigh the harms. Accordingly, the effort for diagnosing all cases with GDM using more sensitive strategies and consuming fewer resources become even more important. Up-to-date, only a limited numbers of studies have been conducted comparing selective risk factor-based screening with universal screening. Most of these studies suffer from methodological problems and limitations. Some are retrospective in design, which restricts the availability of information researched (13–15). Some failed to incorporate age and pregravid BMI to what they named as ‘traditional anamnestic risk factors’ (16) or ‘historical risk factors’ (17). In one retrospective study, the detection rate of the risk factors was tested on patients with GDM after universal screening with a 50-g GCT and 7.8 mmol/l cut-off value (14). This can at its best be a prediction of an already preselected cases because, up to 20% of patients (1) with GDM were already excluded by a 7.8 mmol/l cut-off value.

    The comparison of selective versus universal screening and different researches on this issue is further hampered by numerous definitions of risk factors and timing of OGTT. OGTT was obtained at 28–32 weeks of gestation in two studies (16,17) one of which found this time period to be a late diagnosis with poorer pregnancy outcome (17). The cut-off value of age was chosen to be 35 years (15), 30 years (13,18) and 25 years (1,10–12) to screen for GDM. A prior macrosomic infant was defined to be a birth weight of > 4000 g by some (13,14) and > 4500 g by others (15–17). Our study differs in that, we rather identified the risk factors and cut-off values from our population than applying the risk factors proposed by others. The importance of the issue is that, the different frequency of each risk factor in different populations may change the number of women exempt to screening. The different incidence of GDM in different races was also mentioned by previous studies (16,18).

    Maternal age can be a good example of this proposal. Age is the most frequent variable that classifies the mother to be at risk of GDM. The proportion of pregnant women < 25 years of age in the studied population is the main determinant of the proportion of exempting women as they need to be screened only if, they have other less frequent risk factors mentioned in this study. Today, more women are postponing their pregnancies to their late twenties and early thirties in industrialized countries. Supporting this, the proportion of pregnant women < 25 years of age was 51.5% in our study, which was higher than 17–26% in other studies (13,14,16). Increasing the age threshold to 30 years for GDM screening is not a solution because of several findings: (i) age ≥ 25 years identifies approximately 25% more patients with GDM than a threshold age ≥ 30 years (14); (ii) screening only women ≥ 25 years would detect 85% of GDM cases in our study, which is in agreement with the 90% figure in the literature (14,16); and (iii) youth defined as < 25 years of age is the most consistent protective factor against GDM in several studies (14,16). Therefore, it is not surprising that the higher the proportions of women delivering over 25 years of age in a population, the lower the efficiency of risk factor-based screening.

    This argument can also be made for BMI. Ehrenberg et al. (19) in their study found an increasing prevalence of obesity complicating pregnancy over the past 15 years. This increase was statistically significant after being controlled in multivariate analysis for socioeconomic status and race. The prevalence of obesity in prepregnant women was reported to range from 17% to 26.1% in European populations compared with 18.5% to 30% in the USA when the BMI > 25 cut-off value is used (20). The prevalence of pregravid BMI > 25 was 31.5% in our study population, consistent with the 28.5–33% prevalence reported (14,16) but higher than 16.5–22.6% in other studies (13,18). The eating habits, acceptance of an overweight body shape and ethnic background of the studied population will also influence the frequency of overweight women in a given population and the number of women the risk factor-based screening was applied to.

    Another important finding in our study is the increasing risk of GDM, with the increasing risk factor score. In their studies, Naylor et al. (18) and Jimenez-Moleon et al. (13) also found that the incidence of GDM increases as the number of risk factors present increase. To say it another way, our scoring system points out the importance of screening women < 25 years of age if they have other proposed risk factors. This suggestion is further supported by the work of Khine et al. (21) who found that the majority of 11 adolescents < 19 years with GDM did have risk factors for GDM other than age; 82% had a pregravid BMI > 25 and 36% had a family history of diabetes.

    After identifying the women at risk of GDM, one should decide on the screening test and cut-off value to be used. The 75-g/2-h test recommended by the World Health Organization (22) labels more women as having GDM than does the 50-g screening test (9). A 50-g screening test requires a 100-g OGTT for confirmation. The sensitivity of the 50-g screening test increases from 88% to 96% with a 7.8 mmol/l cut-off value to 90–98% with a 7.2 mmol/l cut-off value (23,24). In their study, Espinosa de los Monteros et al. (23) found significantly lower blood glucose values in pregnant women < 25 years of age, compared with those ≥ 25 years. Also, a 7.8 mmol/l cut-off value yielded a slightly better sensitivity in pregnant women ≥ 25 years of age (23). Our finding of similar sensitivities of 7.2 mmol/l and 7.8 mmol/l cut-off levels for diagnosing GDM in women with a ≥ 2 risk score is partially supported by the results of Espinosa de los Monteros et al. (23). It can be stated that, in women with risk factors, higher glycemia levels occur and the difference in the sensitivities of 7.2 versus 7.8 mmol/l cut-off levels became less pronounced.

    Another argument is the cost-effectiveness of each screening strategy. Considering 79% sensitivity and 87% specificity for a 50-g screening with a 7.2 mmol/l cut-off level and 100% sensitivity and 85.5% specificity for the 75-g/2-h test, Poncet et al. (25) found out that, screening pregnant women, who have risk factors for GDM, via a 50-g GCT produce the most favorable cost- effectiveness ratio when compared with universal screening with 50-g or 75-g/2-h tests. Without doubt, our risk scoring system with 85–100% sensitivity and 97–100% specificity will decrease the above calculated cost further as both the number of women screened via a 50-g GCT and the number of OGTT applied for confirmation is reduced.

    Although not the primary outcome of the present study, we noticed that the risk factor scoring based screening will also enable the diagnosis of all cases with one abnormal value during OGTT. One abnormal value during OGTT was classified as ‘mild gestational hyperglycemia’ occurring in 2.6–3% of the pregnant women (26,27). One abnormal value at any hour after a 100-g OGTT was shown to increase the odds ratio of adverse maternal and perinatal outcome compared with patients with all normal OGTT values (26,27). The prevalence of one abnormal OGTT value in our study was 1.4% and was lower than the prevalence in other studies (26,27). Nevertheless, this limited number of cases in our study renders a detection rate analysis difficult. Further studies with larger numbers of patients must be conducted to find out the diagnostic accuracy of risk factor-based screening on mild gestational hyperglycemia cases.

    In conclusion, compared with the universal screening strategy, a well integrated risk scoring system, based on the evidence of the population, still diagnose 85–100% of GDM cases and also cause a 30–63% reduction in GCT and a 50–53% reduction in the number of OGTT applied in our population. Our study demonstrates that, where ever the health care resources of the population is low, screening women with a ≥ 2 risk score with a 7.8 mmol/l cut-off value for GCT provide a high sensitivity and a minimum intervention approach for the diagnosis of gestational diabetes mellitus.