Canadian Medical Association Journal 1992; 147: 435-443
© 1992 Canadian Medical Association
Since that time other screening manoeuvres have been developed. In Britain Lind [2] has reported on random blood glucose measurement, with a cutoff level of 6.1 mmol/L. Schneider and Curet [3] have suggested blood glucose measurement after fasting and 2 hours after eating. Many authoritative groups have recommended a 1-hour 50-g oral glucose challenge test (GCT) at about 28 weeks' gestation [4,5]. Different GCT regimens have been suggested, including universal screening [4-6] and restricting screening to women over 30 years of age and to younger women with risk factors [7]. Abnormal results are confirmed with a 3-hour oral glucose tolerance test (GTT) before gestational diabetes can be diagnosed.
Some investigators [8,9], however, have questioned the soundness of GCT screening. At issue are the validity of the oral GTT as a diagnostic test [9,10] and the po10ntial impact of management, after screening, on perinatal and maternal morbidity [8,9,11].
For this report the task force reviewed the literature on the validity and potential effectiveness of the different screening regimens that have been proposed. It applied its usual rules to grade the quality of evidence and formulated its recommendations according to established criteria (Appendix 1). This process differs from that generally used in consensus conferences and can result in divergence of recommendations between the task force and other groups.
The original work of O'Sullivan and Coustan as well as references from studies identified by MEDLINE were reviewed in detail. To restrict the search to randomized controlled trials and well-designed cohort studies would have required exclusion of almost all published articles on the subject. Thus, in addition to such studies, articles that met the following criteria were considered.
Gestational diabetes is defined as carbohydrate intolerance of variable severity with onset or first recognition during pregnancy [4,5]. This definition was proposed in 1979 by the National Diabetes Data Group [12], endorsed at the Second International Workshop-Conference on Gestational Diabetes Mellitus [4] and reaffirmed in 1986 by the American Diabetes Association [5]. It departs to some extent from the initial concept proposed by O'Sullivan and Mahan [13], who restricted the diagnosis to temporary alterations of glucose tolerance during pregnancy that resolved by 6 weeks after birth. The current definition encompasses a more heterogeneous population. By this definition gestational diabetes occurs in about 3% of all pregnancies [14].
Risk factors for gestational diabetes include obesity, history of miscarriage or fetal death, maternal age of 40 years or more, family history of diabetes, polyhydramnios, history of a premature infant or an infant with macrosomia (birth weight of 4000 g or more) or congenital malformation, pre-eclampsia, excessive weight gain and glycosuria.
According to O'Sullivan and Mahan [13] and most North American groups [4,5] a definite diagnosis is established with the use of a 3-hour 100-g oral GTT. Currently the condition is diagnosed if the plasma glucose level equals or exceeds, on two or more occasions, 5.8 mmol/L (fasting), 10.6 mmol/L (1 hour after eating), 9.2 mmol/L (2 hours afterward) and 8.1 mmol/L (3 hours afterward) [5]. These criteria differ from those that apply to nonpregnant adults, in whom the test is based on a 75-g oral GTT [12].
However, the World Health Organization (WHO) does not use thesame diagnostic criteria. The 2-hour 75-g oral GTT is the gold standard for both pregnant and nonpregnant adults; gestational diabetes is diagnosed if the blood glucose level is 11.0 mmol/L or greater and impaired glucose tolerance if it is between 8.0 and 10.9 mmol/L [15]. With these criteria a diagnosis of gestational diabetes is less likely. Among 1546 pregnant women with risk factors Li and associates [16] found that 216 (14%) had gestational diabetes with the 100-g oral GTT. However, 2 weeks later only 7 (3%) of the 216 women had an abnormal result with the 75-g oral GTT, 51% had normal glucose tolerance, and 45% had impaired glucose tolerance. The women in the last two groups were randomly allocated to treatment and nontreatment groups. There was no difference in the birth weights of their infants, all of which were below 4000 g.
To understand how the diagnostic criteria were established it is helpful to refer to the original work of O'Sullivan and collaborators. The observation that diabetic women had a history of more frequent fetal wastage and of delivery of unusually heavy infants in the years preceding diagnosis of their condition triggered the hypothesis of a prediabetic state that could adversely affect the fetus [17]. The criteria that were to become the basis for diagnosing gestational diabetes were developed by O'Sullivan and Mahan in 1964 [13]. These investigators first established in 752 healthy pregnant women (cohort 1) the mean responses to a 100-g oral GTT after fasting and 1, 2 and 3 hours after ingesting glucose. They then recruited 1013 pregnant women (cohort 2), performed an oral GTT during the pregnancy and followed them to determine whether overt adult diabetes developed in the years after delivery. Life-table techniques were used to project the cumulative incidence of diabetes mellitus. It was estimated that a blood glucose level one standard deviation (SD) above the mean was associated with a 17.2% (SD 4.3%) incidence of diabetes. The figures were 29.0% (SD 4.5%) for two SDs and 60.1% (SD 8.9%) for three. O'Sullivan and Mahan proposed that pregnant women with a mean glucose level more than two SDs above the mean in an oral GTT be considered to have gestational diabetes. Further follow-up showed that 13.5% actually had overt diabetes during the next 17 to 23 years, as compared with 0.6% of women with normal glucose tolerance during pregnancy [18].
It is important to realize that by O'Sullivan and Mahan's definition [13] gestational diabetes was a retrospective diagnosis. It was based on a Gaussian (or statistically normal) distribution and was associated with an increased probability of overt diabetes later in life. The original work did not permit direct evaluation of any association between impaired glucose tolerance and adverse perinatal outcomes.
The generalizability of O'Sullivan and Mahan's work [13] has been questioned. In a thorough review Naylor [10] pointed out major methodologic flaws. The fact that the prevalence of gestational diabetes was 18.8% in cohort 2 raises questions about the selection process. There was certainly a selection bias that increased the probability of including women at risk of diabetes in the cohort. This leads directly to an overestimate of the predictive value of an abnormal test result. There is not enough information to assess the impact of confounding variables, notably obesity, associated with the development of diabetes. Also, the reproducibility of both the GCT and the GTT was not evaluated.
Finally, in the original study O'Sullivan and Mahan [13] used a technique to measure blood glucose that is now obsolete (the Somogyi-Nelson method on whole blood). Equivalencies needed between venous, serum and plasma glucose levels led to the application of several mathematical models developed from nonpregnant adult populations. Schwartz and Brenner [19] pointed out that various translation errors have occurred in setting the current diagnostic criteria. The relation between these criteria and the original whole blood values does not appear to be consistent. The cutoff points are about 0.5 mmol/L lower than the original values.
The GTT, which is the gold standard for the diagnosis of diabetes, has its own limitations. Its predictive value over time was questioned in the classic study by Genuth and colleagues [20], who showed that 50% of adults identified as having abnormal glucose tolerance had a normal GTT result 5 years later.
The reproducibility of the GTT over shorter periods of time may also be low. McDonald, Fisher and Burnham [21] administered a 100-g oral GTT to 334 healthy male volunters on six occasions, 2 months appart. On the first occasion 9% had an abnormal response. Half of those men had five normal test results subsequently, and three-quarters had four. They repeated their study in a female population of 101 volunteers [22]. Only one of the six women with a positive result the first time continued to have a positive result on four subsequent occasions.In addition, O'Sullivan and Mahan [23] reported intraindividual variability in the response to the GTT. This variability was even more marked in pregnant women.
In summary, two sets of diagnostic criteria are still used to establish the presence of gestational diabetes; this leads to two estimates of its prevalence. The diagnostic cutoff points established for the 100-g GTT were originally conceived as identifying a group at risk of overt diabetes later in life and were based on a statistical approach to the definition of normality.
Consequences of gestational diabetes
There is no point in searching for a biochemical abnormality that does not cause significant health problems. The presence of gestational diabetes has been associated with several adverse outcomes: perinatal death, fetal macrosomia with its associated complications, and congenital malformations. To what extent is gestational diabetes responsible for these adverse outcomes and what is the magnitude of its effects?
Perinatal death: An association between gestational diabetes and perinatal death was first reported in 1973, by O'Sullivan and coworkers [24]. In 259 women the rate of perinatal loss was 1.5% among those with a normal oral GTT result, as compared with 6.4% among those with gestational diabetes. However, the results reached statistical significance only after exclusion of women under 25 years of age (relative risk 4.3). In 1975 Abell and Beischer [25] reviewed the outcome of 2000 pregnancies in women who had undergone a 3-hour 50-g oral GTT. An abnormal result was defined as a 2-hour value above the 95th percentile. They reported a perinatal death rate of 31.7 per 1000 when the result was abnormal, as compared with 9.5 per 1000 when it was normal. Of the 26 perinatal deaths 10 were "predicted" by an abnormal GTT result, for a sensitivity of 38% and a specificity of 85%. In none of these studies was there control for confounding variables (gestational age, parity, maternal age, obesity and socioeconomic status) associated with both perinatal death and gestational diabetes.
Consequently, as Hunter [9] pointed out, the increased risk of perinatal death observed may have been predicted as much by the indications to perform glucose tolerance testing as by the results of the test. Sutherland and Stowers [26], who performed intravenous GTT on 1600 pregnant women, observed that the number of fetal deaths increased eightfold as the number of indications for glucose tolerance testing increased from one to four. But the risk of fetal loss for each condition was not affected significantly by the results of the tolerance test itself.
Macrosomia: Gestational diabetes has been associated with macrosomia and with asymmetric growth, which may cause overdevelopment of the shoulders even if the birth weight is within normal limits. The first case series showed a marked prevalence of macrosomia among infants born to women with gestational diabetes, from 12% to 35% [24,27,28]. Unfortunately, the potential impact of confounding variables was not taken into account.
However, gestational diabetes is not the only risk factor for macrosomia. Maternal age, maternal weight and multiparity are all more strongly associated with macrosomia than with gestational diabetes alone.
In a well-conducted study Spellacy and associates [29] found that only 5.1% of infants weighing more than 4500 g were born to women with gestational diabetes. By comparison 44.6% were born to obese women and 10.8% to women beyond 42 weeks' gestation. With an overall incidence of macrosomia of 1.7 per 1000, the relative risk for gestational diabetes was 3.0, as compared with 25.8 and 6.4 per 1000 for obesity and postmaturity respectively.
The association of gestational diabetes with macrosomia would be of no concern if the latter were not associated with an increased rate of perinatal complications. It has been clearly established that macrosomia is associated with increased maternal and neonatal death rates [26,28,30]. Boyd, Usher and McLean [30] have shown that primary cesarean section for cephalopelvic disproportion and for nonprogression of labour is twice as frequent if the infant has macrosomia (19% v. 8%). Brachial plexus paralysis and clavicular fractures are more frequent in infants weighing more than 4000 g than in smaller infants (2.5% v. 0.01% and 5.5% v. 0.06% respectively). Levine and collaborators [31] retrospectively studied 13 870 singleton live births. The incidence rates of brachial plexus injury and clavicular fracture were 2.6 and 2.0 per 1000 respectively. In the group without birth trauma 5.5% of the infants had macrosomia, as compared with 38.9% of those with brachial plexus injury and 21.4% of those with clavicular fracture. The frequency of gestational diabetes was not given. The limitation of these studies is that birth trauma was considered in relation to macrosomia and not to gestational diabetes. The relative contribution of gestational diabetes to adverse outcomes of newborns, particularly those with disproportionate growth and normal birth weight, cannot be estimated. In a series of infants with shoulder dystocia in a community hospital, Hopwood [32] reported the presence of gestational diabetes in only 5.4% of cases.
Hypoglycemia: Most studies of the potential impact of the treatment of gestational diabetes have not included neonatal hypoglycemia as an outcome. Perhaps this is because screening for hypoglycemia is done routinely in nurseries, and intervention is prompt for all infants.
Congenital malformations: Although it has been suggested that gestational diabetes is associated with an increased risk of congenital anomalies, the study by Chung and Myrianthopoulos [33] from the Boston Collaborative Project demonstrated no increased risk of congenital anomaly for this condition. Becerra and colleagues [34] seem to have refuted this fact. They found a relative risk of 20.6 (95% confidence interval 2.5 to 168.5) for women with gestational diabetes who required insulin in the third trimester. Cases involved children born with congenital malformations between 1968 and 1980. The survey was conducted by telephone interview in the early 1980s. These results must be interpreted with caution because of the potential for a strong recall bias and differences in methods used to diagnose gestational diabetes between the case and control groups. The biologic plausibility of such an observation is also questionable. How can an event that occurred toward the end of the pregnancy affect organogenesis?
Testing for glycosuria
The renal threshold for glucose varies from one individual to another and may be quite low. About half of all healthy pregnant women excrete an increased amount of glucose in their urine [35]. In a recent study of primary care patients urine glucose measurement was reported to have a sensitivity of less than 30% [36].
The observation that the fasting blood glucose level correlated poorly with values observed subsequently during the 100-g oral GTT led to the search for a screening manoeuvre that would correlate better with the results of the oral GTT than with the fasting glucose level and be less cumbersome to administer [13,27].
The 50-g GCT
The oral GCT was developed: 50 g of glucose is administered orally, and then the blood glucose level is measured 1 hour later. The validity of the test was evaluated in 752 women who underwent both the oral GTT and the oral GCT in the third trimester [37]. The work was reproduced by Carpenter and Coustan [38], who set the plasma cutoff point for the GCT at or above the cutoff point for the GTT (7.8 mmol/L).
Two issues are still being debated among advocates of the GCT: What should the cutoff point be at which to prescribe an oral GTT, and who should be screened? The GCT's reported sensitivity was 79% to 83% and the specificity 87% when the cutoff point was established at a plasma glucose level of 7.8 mmol/L. Marquette and coworkers [39] have suggested increasing the screening threshold to 8.3 mmol/L. In their study such modification decreased the sensitivity of finding an abnormal oral GTT result by only 0.7% while increasing the specificity to 93%. Unfortunately no consideration was given to a possible association with adverse outcomes. However, others have advocated a more stringent cutoff point [28,40,41]. A follow-up study involving women who had a plasma glucose level above 7.2 mmol/L with the GCT but a normal GTT result revealed that they were at higher risk of macrosomia than those with a level below 7.2 mmol/L [28,40]. Langer and associates [41] also reported that the presence of only one abnormal value with the GTT was associated with an increased risk of macrosomia. More research is needed to evaluate the significance of these observations that raise questions about current screening and diagnostic criteria.
The question of who to screen is also debated in relation to the yield of screening compared with the results of the oral GTT. The American College of Obstetrics and Gynecology recommends screening women over 30 years of age [7], since O'Sullivan and coworkers [24] reported an association between adverse outcomes and poor glucose tolerance only after excluding women under 25 years old. However, since some studies have shown that up to half of women with gestational diabetes can be missed if age criteria are used [42], other groups have argued in favour of universal screening of pregnant women [4-6].
Fasting and random blood glucose measurement
If the correlation between the fasting blood glucose level and the response to the GTT appears to be low in pregnant women,the maintenance of euglycemia after fasting is the factor associated with a low prevalence of perinatal death in most studies of gestational diabetes. For this reason Mortensen and collaborators [43] and Schneider [3] have recommended measuring the fasting blood glucose level. In Britain Lind [2] has argued for the use of random blood glucose measurement. Neither manoeuvre has been adequately evaluated.
Finally, glycohemoglobin is a poor predictor of an abnormal oral GTT result. Its sensitivity in pregnant women has been found to be about 26% [44].
To conclude, the 50-g oral GCT is the only screening manoeuvre that has been evaluated in relation to the 100-g oral GTT in order to provide reliable estimates of its sensitivity and specificity. Its test characteristics in terms of the WHO diagnostic standard (the 2-hour 75-g oral GTT) are unknown.
Measuring the blood glucose level after fasting and 2 hours after eating might be an acceptable alternative to the GCT. A study is under way that will compare the sensitivity and specificity of fasting and random blood glucose measurements and of the GCT in detecting gestational diabetes as defined by the oral GTT. The study will also compare their value in predicting several
No analytic study, either a randomized controlled trial or a case-control study, has evaluated the impact of a screening program for gestational diabetes on perinatal death and illness.
In 1990 Santini and Ales [45] published a study involving women who did (774) or did not (533) undergo screening for gestational diabetes according to the practices of their obstetricians. The study took place at the New York Hospital-Cornell University Medical Center, where in the early 1980s some obstetricians screened nearly all their patients whereas others did not. Screening failed to decrease the rate of macrosomia (10.5% in the unscreened group v. 11.2% in the screened group) and was associated with more intensive surveillance during pregnancy and a higher rate of primary cesarean section (21.0% v. 26.7%; p <0.01). However, the presence of risk factors was not evenly distributed between the two groups of patients. Screened women were more likely to be obese. Since obesity is the most potent risk factor for macrosomia this imbalance of prognostic variables is a major threat to the validity of the study.
The potential impact of screening must be estimated from studies of the treatment of gestational diabetes once it has been diagnosed. The case series and uncontrolled and randomized trials that have dealt with this issue are reviewed here.
Five observational studies [46-50] have suggested that the maintenance of euglycemia in women with gestational diabetes is associated with a decreased perinatal death rate. In three investigations the patient's previous pregnancy served as a historical control [47-49]. All three showed a marked decrease in the perinatal death rate (8.3% to 0%, 27.5% to 1.8% and 5.2% to 0%).
Coustan and Imarah [50] analysed retrospectively the outcome of 445 pregnancies involving women with gestational diabetes managed between 1975 and 1980. Three therapeutic interventions were compared: a classic prenatal diet not specific to diabetes, a strict diabetic diet and fixed insulin therapy in addition to the diabetic diet. Assignment to each group depended on the gestational age at which diabetes was diagnosed, the degree of hyperglycemia and the patient's willingness to follow a strict therapeutic regimen. Adverse outcomes were less frequent in the insulin group than in the untreated group and the diet-alone group (macrosomia 7% v. 18.5% and 17.8% respectively, operative delivery 16.3% v. 30.4% and 28.5%, and birth trauma 4.8% v. 13.4% and 20.4%).
Gabbe and colleagues [46] compared the perinatal death rate among the infants of 261 women with gestational diabetes followed up between 1970 and 1972 with the rate in the general population of the same hospital. The rates were 32 and 19 per 1000 respectively. The former rate is considerably higher than those observed in other studies [24,47-49].
In none of these case series were confounding variables taken into account. Even more important is how the diabetes was identified. In all studies a woman had to have at least two of the classic risk factors for gestational diabetes. Recognizing these risk factors led to screening for diabetes with the 3-hour oral GTT. All women with an abnormal result were considered to have gestational diabetes. Thus, in none of the observational studies were women identified through the proposed mechanism of screening all women with a 1-hour GCT regardless of the presence of risk factors.
Two experimental studies of the effectiveness of treatment of gestational diabetes are relevant. The first was conducted by O'Sullivan and coworkers [51] between 1954 and 1960. All women attending the prenatal clinic of the Boston City Hospital underwent a 1-hour 50-g GCT. Women with an abnormal result and those with two risk factors for gestational diabetes (independent of the results of the GCT) were offered the opportunity to undergo an oral GTT. The 615 women thus identified as having gestational diabetes were randomly allocated to either a routine prenatal care group or an insulin-plus-diet treatment group. In addition, a random sample of 328 women with no glucose intolerance was followed. The incidence of macrosomia was decreased in the insulin-treated group (3.7% v. 13.1%) and was comparable to the rate observed in the group of women with normal tolerance (4.3%). No difference in the perinatal death rate was observed. O'Sullivan and coworkers did not report the impact of treatment on the rates of birth trauma and operative delivery.
The second experimental study, conducted by Coustan and Lewis [52], was performed at the Naval Regional Medical Center, Oakland, Calif., between 1973 and 1975. All women at risk for gestational diabetes according to their obstetric history, family history, age, parity or obesity were scheduled to undergo a GTT. Seventy-two cases of gestational diabetes were identified. The first 20 were assigned to receive either insulin or no treatment on the basis of gestational age at diagnosis (10 patients in each group). The 52 others were randomly assigned to receive insulin treatment (17 patients), a change in diet alone (11) or no treatment (24). The 72 patients were analysed regardless of the mode of allocation to the treatment groups. It was not explained why the 52 patients were not assigned evenly to each of the three groups. Only 2 of the 11 subjects in the diet-alone group were older than 25 years, as compared with half of those in the two other groups. All women were offered the same prenatal care. A significant decrease in macrosomia was noted in the insulin-treated group (7%), as compared with the diet-alone group (36%) and the no-treatment group (50%). Only one case of Erb's palsy was reported, in the diet-alone group. The rate of primary cesarean section did not differ significantly between the three groups (14.8%, 18.2% and 17.6%), nor did the need for forceps delivery (11.1%, 0% and 8.8%). There was no difference in the perinatal death rate. Coustan and Lewis did not control for maternal weight and postmaturity. The high prevalence of macrosomia points to a selection bias in favour of other known risk factors for macrosomia (e.g., in the three groups half the patients were obese).
Such a dramatic impact on the incidence of macrosomia as the one reported by Coustan and Lewis cannot be expected now from universal screening of pregnant women. Table 1 shows that a screening program in Canada, even if totally effective, would at the most prevent 50 brachial plexus injuries and 109 clavicular fractures. Since many macrosomic infants are delivered by means of cesarean section (over 20%) this is an overrepresentation of the risk of obstetric injury. Furthermore, clavicular fracture is benign, and well over 80% of brachial plexus injuries resolve completely within 3 months [53].
Potential for harm
Since there is no evidence of any benefit from screening asymptomatic adults for diabetes mellitus [54] the identification of a population of women at increased risk of diabetes later in life cannot be considered a valuable outcome of screening for gestational diabetes. There is a danger of labelling, since 70% of women with gestational diabetes will not have overt diabetes [17,18].
It can be argued that such women could be offered anticipatory counselling to prevent the occurrence of overt diabetes. However, nothing more than good nutritional advice, which should be directed to all women, would be given, and most of the women would be unnecessarily worried and potentially penalized in future life insurance or job applications.
Since the indications for the GCT actually define the risk of adverse outcomes a negative test result does not eliminate the risk.
Although of no proven benefit, manipulation of birth weight may be harmful to the 70% of infants who are not destined to have macrosomia. As for other medical interventions, management of gestational diabetes carries with it the risk of overtreatment, as shown by Langer and associates [55]. In their study women with gestational diabetes who maintained a low mean glucoselevel throughout pregnancy (4.8 mmol/L or less) had a significantly higher incidence of small-for-gestational-age infants (20% v. 11% in the control group). Intrauterine growth retardation is not an insignificant perinatal outcome.
Finally, screening can generate an important consumption of health care dollars; if these tests are not covered by insurance, individual expenditures can be considerable [11].
An alternative to the 1-hour GCT is the measurement of the blood glucose level after fasting or 2 hours after appropriate screening mechanisms, their effectiveness has not been evaluated.
These facts therefore support a cautious approach toward screening all pregnant women with a 1-hour 50-g GCT at 28 weeks' gestation. Conversely, women with risk factors for gestational diabetes should be carefully followed throughout their pregnancy, even though the effect of minor elevations in the maternal blood glucose level on neonatal health remains to be established.
Further research is needed to establish the relative risk of neonatal and perinatal illness in relation to various degrees of subdiabetic elevations in the maternal blood glucose level.
The task force recognizes that women have various degrees ofglucose intolerance during pregnancy and that a certain proportion will have adverse outcomes and could benefit from screening. Universal screening with the 50-g GCT has not been demonstrated to be superior, yet it can carry considerable costs. Pending the results of ongoing studies of the sensitivity and specificity of other screening manoeuvres, factors such as clinical judgement (assessment of risk factors and pregnancy evolution) and available resources should be taken into account in choosing a screening strategy.
The task force thanks Drs. David J.S. Hunter and David C. Naylor for their valuable consultation in our review of this condition.
CPG Infobase home page
![CMA Online ]](../gifs/homel.gif){kind=small}
CMA Online