Seth C. Brody, M.D.a; Russell Harris, M.D., M.P.H.b; Kathleen Lohr, Ph.D.c
The authors of this article are responsible for its contents, including any clinical or treatment recommendations. No statement in this article should be construed as an official position from the Agency for Healthcare Research and Quality, or the U.S. Department of Health and Human Services.
Address correspondence to: Seth Brody, MD; Department of Obstetrics and Gynecology; University of North Carolina; Wake Area Health Education Center; 3024 New Bern Avenue, Suite 306; Raleigh, NC 27610-11255; E-mail: sbrody@med.unc.edu
Select for copyright information. The USPSTF recommendations based on this review are online.
The summaries of the evidence briefly present evidence of effectiveness for preventive health services used in primary care clinical settings, including screening tests, counseling, and chemoprevention. They summarize the more detailed Systematic Evidence Reviews, which are used by the U.S. Preventive Services Task Force (USPSTF) to make recommendations.
Abstract
Epidemiology
Materials and Methods
Results
Discussion
Acknowledgments
References
Notes
Objective: To systematically review the evidence for screening for gestational diabetes mellitus (GDM).
Methods: We established eligibility criteria for relevant studies. We systematically searched MEDLINE® and the Cochrane Collaboration Library for studies meeting eligibility criteria; we supplemented this search with further studies identified from reference lists of reviews. Two reviewers examined each article for eligibility. A single reviewer abstracted relevant data from the included articles; a second reviewer checked the abstractions. We graded the quality of the articles according to criteria developed by the U.S. Preventive Services Task Force (USPSTF).
Results: No well-conducted randomized controlled trial (RCT) provides direct evidence for the health benefits of screening for GDM. The evidence is unclear about the optimal screening and reference diagnostic test for GDM. The impact of hyperglycemia on adverse maternal and neonatal health outcomes is probably continuous. Although insulin therapy decreases the incidence of fetal macrosomia for those women with more severe degrees of hyperglycemia, the magnitude of any effect on maternal and neonatal health outcomes is not clear. The evidence is insufficient to determine the magnitude of health benefit for any treatment among the large number of women with GDM at milder degrees of hyperglycemia. We found limited evidence about the potential adverse effects of screening for GDM.
Conclusion: Because of the lack of high quality evidence concerning critical issues, we are unable to determine the extent to which screening has an important impact on maternal and neonatal health outcomes. An RCT of screening is necessary to answer the many remaining questions.
GDM is defined as glucose intolerance with the onset or first detection during pregnancy.1,2 About 135,000 cases of GDM are diagnosed annually in the United States.2 Important risk factors include higher maternal age, family history of diabetes, and increased pregravid body mass index (BMI).3 The prevalence of GDM in low-risk populations ranges from 1.4 to 2.8 percent;4,5 in high-risk populations, prevalence ranges from 3.3 to 6.1 percent.4
Markedly elevated maternal glucose levels most often occur in women with pregestational diabetes. Pregnant women with pregestational diabetes are at higher risk for multiple complications affecting both the mother and the fetus than those women without diabetes. Current therapy improves outcomes for both mother and neonate.6
The additional risk for adverse health outcomes attributable to the milder degrees of maternal hyperglycemia associated with GDM and the magnitude of the benefit from treating that risk are less certain. No well-designed and conducted RCT of screening for GDM has been completed, and thus the evidence for screening is indirect.
National groups disagree about whether to recommend screening for GDM.2,7-11 Despite no strong recommendations in favor of universal screening from the American College of Obstetricians and Gynecologists (ACOG), 94 percent of Fellows in office-based practices reported performing universal screening for GDM in 1996.12 Fellows performed this screening even though ACOG acknowledged the weakness in the evidence in both 199413 and 2000.2
With continued controversy around the advisability of GDM screening, the RTI-University of North Carolina Evidence-based Practice Center (RTI-UNC EPC) conducted a systematic evidence review to assist the USPSTF in reconsidering its 1996 review, which found insufficient evidence to recommend screening. We restricted this review to screening for GDM after 24 weeks' gestation, thus excluding both women with known pregestational diabetes and those who are discovered by symptoms earlier in pregnancy.
Our review of the literature was guided by key questions and inclusion criteria we developed relevant to the issue of screening for GDM (Table 1). We required RCTs for direct evidence of the efficacy of treatment and the harms associated with treatment. We examined the critical literature from the 1996 USPSTF review and searched MEDLINE® and the Cochrane Library for reviews and relevant studies published in English between January 1, 1994, and August 30, 2002. We supplemented this search by examining the reference lists of pertinent articles and by contacting experts. We also conducted focused searches of MEDLINE® from 1966 through 1994 to identify older articles of interest.
All searches began by exploding the term "diabetes, gestational" and then proceeded by adding further terms. We retrieved the full text of all articles we thought were potentially eligible. Two reviewers examined each article for eligibility. A single reviewer abstracted relevant data from the included articles; a second reviewer checked the abstractions.
We abstracted all included articles, entered the data into evidence tables, graded the quality of all articles according to USPSTF criteria,14 and resolved disagreements by discussion. We synthesized the evidence into a systematic evidence review; this was subjected to extensive external peer review and revised as appropriate.15 The final systematic evidence review, including the evidence tables, is available on the Agency for Healthcare Research and Quality (AHRQ) Web site (http://www.preventiveservices.ahrq.gov). This article summarizes the evidence from that review.
For the USPSTF to recommend screening for GDM, it must have either direct evidence from a randomized controlled trial (RCT) of screening or indirect evidence that establishes a complete linkage between screening and improved health outcomes. We found no well-conducted RCT that provides direct evidence for the health benefits of screening for GDM. Given this, the USPSTF requires adequate evidence that:
These issues will be examined in the sections that follow.
Determining the existence and magnitude of a causal association between various degrees of GDM and adverse health outcomes is complex. We have only older studies of untreated GDM, at a time when obstetric practice differed from current practice, or more recent studies in which women received some treatment for GDM. Another problem with many studies is that they consider GDM as a dichotomous variable, yet we know that the risk for adverse health outcomes increases with the degree of hyperglycemia among women with GDM; the impact of hyperglycemia on adverse maternal and neonatal health outcomes is probably continuous.16-19 Few studies, however, stratify the risks of GDM by severity of hyperglycemia.
Because the literature is scant and mixed about whether untreated GDM, given optimal obstetric care today, is associated with increased perinatal mortality,19-27 the extent to which GDM is truly associated with perinatal mortality remains unclear.
Macrosomia is an intermediate outcome of GDM. Three recent studies of untreated women with GDM20,22,23 found that the percentage of infants with macrosomia weighing more than 4,000 grams was between about 17 percent and 29 percent; the percentage in the general population is about 10 percent.26 Most infants with macrosomia are born to women without GDM;28 maternal obesity is an important potential confounding factor associated with both GDM and (independently) with macrosomia.
Important adverse neonatal health outcomes linked to macrosomia are brachial plexus injury and clavicular fracture. The best (although minimal) data on untreated women with GDM compared with the non-GDM population reveal no difference in the rate of infant brachial plexus injury or clavicular fracture.22 Recent data suggest that women treated for GDM with more severe degrees of hyperglycemia may have a 2-percent absolute increase in having their infants develop a brachial plexus injury and a 6 percent increase in having their infants develop a clavicular fracture.26,29 While these adverse health outcomes are of concern, the best studies show that 80-90 percent of brachial plexus injuries resolve by 1 year of life,30-33 and more than 95 percent of clavicular fractures heal within a few months without residual problems.34-39
GDM may also be a risk factor for neonatal hypoglycemia. Studies among untreated20 and treated women with GDM have found higher rates of neonatal hypoglycemia among untreated women with GDM. The magnitude of clinically important neonatal hypoglycemia is less clear. Also not clear is whether increased surveillance of infants whose mothers have GDM contributes to the increased finding of hypoglycemia in their infants.
Likewise, the evidence is limited and unclear as to whether GDM is associated with preterm birth or neonatal hyperbilirubinemia, hypocalcemia, or polycythemia.1,20,38,40-44 Because of limited evidence and the increased surveillance given to infants of women with GDM, the magnitude of any associated adverse health effects is uncertain but is likely to be small.
Some have suggested that the diagnosis of maternal GDM may have long-term implications for the offspring, such as an increased risk for impaired glucose tolerance, childhood obesity, and neuropsychological disturbances. No large observational study has followed a group of children whose mothers have GDM and a comparison group whose mothers do not have GDM long enough to demonstrate whether any of these hypotheses are correct.45-48
The diagnosis of GDM can also increase adverse health outcomes for the mother during her pregnancy. Limited data since 1980 reveal total cesarean delivery rates of 22 percent49 to 30 percent22 for unrecognized or untreated women with GDM, compared with a rate of about 17 percent for women without GDM. Although the overall literature suggests an association between GDM and higher cesarean delivery rates,31,50-54 some studies are limited by a lack of adjustment for maternal obesity and by the impact of the diagnosis of GDM on clinical decisionmaking.
Limited evidence is available on the rate of third- or fourth-degree lacerations in women with GDM. Some studies have suggested an increase,30 but the only study that found a substantial percentage of women with untreated GDM who had such lacerations included only 16 subjects.26 Another study found equally low rates among women with GDM and women without GDM.22
Overall, observational studies have shown mixed results and are inconclusive as to whether women with GDM have a higher risk for pre-eclampsia than women without GDM.1,39,55-57 Recent data from untreated women with GDM22 reveal a rate of pre-eclampsia (about 9 percent) that is similar to that for treated women and women without GDM.58-61
Mothers identified as having gestational diabetes also have a higher risk for developing type 2 diabetes in the years after delivery.62 Studies investigating the rate of development of type 2 diabetes after the onset of gestational diabetes have suffered from low participation rates, retrospective design, short followup, and variation in definition of both GDM and new diabetes. Although nearly all studies show that women who have GDM face some increased risk for developing diabetes, the degree of risk elevation they experience and the degree of glucose abnormality they develop are uncertain.2 Further, the added benefit of early detection of diabetes in young women with few cardiovascular risk factors is uncertain.63
Before we can determine the accuracy of a screening test, we need a reference diagnostic test for comparison. Unfortunately, no universally agreed on reference test for the diagnosis of GDM exists.
Three competing criteria for diagnostic glucose tolerance tests (GTT) are available (Table 2). Criteria from the World Health Organization (WHO)64 label twice as many women with GDM as do criteria from the National Diabetes Data Group (NDDG).65 Criteria from the American Diabetes Association (ADA)66 give an intermediate prevalence.2,67-69
Abnormal values on any of the three tests are predictive of fetal macrosomia.11,23,70-72 This association is diminished or eliminated when adjustments are made for such potential confounders as pregravid weight, age, parity, and race.
The reliability of any oral GTT is open to question. In 1 of the few studies on this issue, Harlass et al. found that 23 percent of 64 unselected pregnant women who had had a positive screening test for GDM had inconsistent results between two 100-gram oral GTTs performed 1 week apart.73 Other studies have also raised concerns about the reproducibility of the oral GTT in nonpregnant groups.74-76
The thresholds for the reference diagnostic tests do not clearly distinguish women at high risk from women at low risk for adverse maternal or neonatal health outcomes from GDM. Thus, we can evaluate screening tests only against imperfect standards. Most studies on GDM screening strategies compare the results of one test with the results of another test rather than examining how the test predicts adverse health outcomes. Some studies assess the association of the test with intermediate outcomes such as macrosomia rather than health outcomes such as brachial plexus injury.
In the United States, the 50-gram, 1-hour glucose challenge test (GCT) is most commonly used for screening (Table 2). Two groups have proposed different threshold criteria to define a positive screening test. If the GCT glucose value is above either 130 mg/dL77 or 140 mg/dL,65 then the patient is usually given the 100-gram GTT for diagnosis. Using the 130 mg/dl threshold, the GCT is positive for 20-25 percent of all pregnant women, including 90 percent of women with GDM. Using the 140 mg/dl threshold, the GCT is positive for 14-18 percent of all pregnant women, including about 80 percent of women with GDM.67
In the general population, false-positive results for the GCT are common. Fewer than 1 in 5 women with a positive GCT will meet criteria for GDM on a full 100-gram GTT.78 The reliability of the GCT is also problematic.16
In many countries outside North America, clinicians use the WHO screening approach: the 75-gram 2-hour oral GTT as a single-step screening and diagnostic test. As noted above, this approach identifies at least twice as many women as having GDM as the two-step approach, although the evidence is sparse about whether the one-step test is more or less predictive of adverse health outcomes than the two-step approach.68,69
Because glucose intolerance increases during pregnancy, screening for GDM is most commonly conducted during the 24th to 28th week of gestation. However, this timing is not based on any evidence that this is the optimal time to identify women who would benefit most from treatment. Determining the best time to screen involves examining the trade-off between the potential benefits of early screening (i.e., finding fewer women at higher risk and treating them for a longer time) and the potential benefits of later screening (i.e., finding a larger number of women at lower risk and treating them for a shorter time).19 We found no study on this issue.
One suggested approach to improve the efficiency of screening for GDM is to restrict screening to women at higher risk ("selective screening") rather than screening all women ("universal screening"). In the most detailed study of selective screening strategies, Naylor et al. developed a scoring system that excluded nearly 35 percent of women from screening and actually detected more cases of GDM than universal screening.61
In summary, the evidence is unclear about the optimal screening and reference diagnostic test for screening for GDM.
Three factors are important in considering studies that evaluate the impact of tight glycemic control on health outcomes for women with GDM:
Table 3 records data from nine RCTs examining the impact of therapy on a variety of outcomes.20,29,58-60,80-83 The first four RCTs are of women with mild hyperglycemia20,80,81,83 and the last five are of women with severe or very severe hyperglycemia.29,58-60,82
Mild Hyperglycemia. Few studies have examined the effectiveness of intensive compared with less intensive glycemic control among women with GDM who have mild hyperglycemia. An overview of 4 trials that included 612 women with mild hyperglycemia found no difference in adverse health outcomes between the women treated with a modified diet and the women receiving no therapy.84 The Li et al. RCT made a similar comparison and had similar findings.81
Three RCTs compared intensive with less intensive glycemic control (achieving some glycemic separation) among women with GDM who had varying degrees of hyperglycemia but a low mean entry fasting plasma glucose (FPG) or mean hemoglobin A1c (HbA1c).20,80,83 Two studies found statistically significant improvements in intermediate outcomes for those women undergoing intensive glycemic control (e.g., fewer large-for-gestational age [LGA] infants83; lower incidence of neonatal hypocalcemia20); no study found clear differences in health outcomes between glycemic control groups.
Severe Hyperglycemia. Four RCTs examined tight and less tight glycemic control among women with GDM at more severe hyperglycemic levels (Table 3).29,58,60,82 Of these trials, three achieved either small or no difference in glycemic control between groups and found no difference in major outcomes.29,58,60 One trial found a small absolute reduction in chemical abnormalities in the neonate58; another found a reduction in cesarean deliveries that was not explained by fetal size.29
One study achieved a larger glycemic separation between groups (difference in mean glucose, 24 mg/dL).82 The infants of less intensively treated women had a higher mean birth weight plus higher rates of neonatal hypoglycemia and polycythemia. These differences were small and of uncertain clinical importance.
Finally, de Veciana et al. compared tight with less tight control among women with very severe hyperglycemia, some of whom likely had frank diabetes (Table 3).59 They also achieved the largest separation in glycemic control (HbA1c difference of 1.6 percent) and found some of the larger reductions in fetal macrosomia and neonatal hypoglycemia. Given the study population, however, this trial may have little relevance for the great majority of women detected with GDM.
A major issue in all of these trials is that they have too few participants to be able to detect small differences among treatment groups in such uncommon adverse health outcomes as perinatal mortality and brachial plexus injury. They have even less power to determine whether the health benefit is different for women with GDM who have severe hyperglycemia compared with those who have mild hyperglycemia. They provide insufficient evidence to confirm or refute the hypothesis that glycemic control improves health outcomes for women with GDM.
Several observational studies without randomized controls have suggested improved intermediate or health outcomes with more intensive treatment of women with GDM.22,71,85-91 The weakness in these studies is that women in the treatment groups differ from women in the control groups in multiple ways (some known and some unknown) other than glycemic control; most of the known factors are also associated with health outcomes. Thus, observed improvements in health outcomes may be attributable to factors other than glycemic control.
In summary, although insulin therapy decreases the incidence of fetal macrosomia for those women with more severe degrees of hyperglycemia, the magnitude of any effect on maternal and neonatal health outcomes is not clear. The evidence is insufficient to determine the magnitude of health benefit of tight glycemic control among the large number of women with GDM at milder degrees of hyperglycemia.
Various approaches to antepartum surveillance might improve health outcomes among women with GDM. For non-stress testing (NST) or biophysical profile (BPP) to constitute a rationale for GDM screening, evidence would need to show that the use of these tests reduces stillbirth among women with GDM who have no other indication for these tests. This would require a large RCT, as most women with GDM have a low risk for having a stillbirth. No completed study of women with GDM has examined health outcomes among groups randomized to receive or not receive NST or BPP. Observational studies have found that using NSTs or BPPs in women with GDM is associated with either absent or very low rates of stillbirth.92-95 Without appropriate control groups we do not know whether the low rate of fetal death can be attributed to the additional procedures.94 NSTs or BPPs have high false-positive rates,92,95 and they lead to interventions94 that may, on occasion, be unnecessary.
Ultrasound assessment of abdominal circumference to allow improved targeting of insulin therapy in order to decrease fetal macrosomia and birth trauma has been studied. Three RCTs have enrolled women with hyperglycemia into insulin therapy triggered by ultrasound abdominal circumference.29,83,96 These studies have not found any important differences in health outcomes; all three lacked power to detect differences in health outcomes and in none were the obstetricians masked to the intervention group.
Precise evidence on the harms and costs of screening for GDM and early treatment is lacking. Although not well documented, the potential for adverse psychological effects from screening is real; in the general population, more than 80 percent of all positive GCT screening tests are false positives.97 Limited and mixed data suggest that labeling may negatively influence women's perceptions of their health during pregnancy97-100 and that women diagnosed with GDM may have long-term changes in their perception of their own health.98,101 The long-term impact of these changes in perception of health is unclear.
Identification of GDM may also needlessly increase the use of NSTs or BPPs (triggering unnecessary interventions due to false positives) and rates of cesarean delivery (because of a lowered intervention threshold).22,70 Furthermore, additional tests and procedures increase the cost of screening programs. Because of the lack of evidence, the magnitude of other potential harms of aggressive glycemic-lowering therapy, such as increased maternal starvation ketosis and infants who are small for gestational age, is difficult to quantify.18,102
As the effectiveness of screening in improving health outcomes is uncertain, so the cost-effectiveness cannot be calculated with any precision. We do not have good information about the differences in health care costs between screened and unscreened women.
Maternal and neonatal morbidity increase with increasing levels of maternal hyperglycemia. Screening and intensive treatment for GDM aim to reduce this morbidity. Various screening strategies can detect women with different degrees of hyperglycemia, but the threshold at which health outcomes begin to deteriorate to a clinically important degree is uncertain.
The magnitude of any benefit of intensive treatment at the various levels of hyperglycemia associated with GDM is also uncertain, but it is likely to be small among the many women with mild hyperglycemia. For women with GDM who have more severe hyperglycemia, intensive treatment is likely to reduce macrosomia. The extent to which this translates into reductions in birth trauma is uncertain but probably substantially less than reductions in macrosomia.
The evidence about the health outcomes of intensive treatment of women with GDM at various levels of maternal hyperglycemia is indirect. It is also limited by a small number of studies, small number of participants, lack of masking of obstetrical care, lack of control for important confounders, and lack of emphasis on health outcomes rather than intermediate outcomes.
By making various assumptions, we can calculate the number needed to screen (NNS) to prevent various adverse health outcomes. Take, for example, the number of women needed to screen to prevent one case of brachial plexus injury (Table 4). Assume that 4 percent of pregnant women have GDM,2 that 30 percent of them will have a high enough glycemic level to require insulin,79 and that, among these women, the macrosomia rate is reduced to the degree seen in the most positive study.59 The NNS to prevent one brachial plexus injury is about 8,900 (Case 1, Table 4).32,33,103 If we make more generous assumptions, the NNS becomes 3,300 (best case scenario, shown in Case 3 and footnote, Table 4). Assumptions including a lesser reduction in macrosomia, accounting for cesarean delivery rates, or using an outcome of permanent brachial plexus injury, would give much higher NNS estimates.2
One potential benefit of detecting women with GDM is the knowledge that they have a higher risk for developing type 2 diabetes. The extent to which this information can lead to a health benefit for younger women with few cardiovascular risk factors, however, is uncertain.63
The evidence concerning the harms and costs of screening and intensive treatment is even more limited than the evidence about benefits, but several harms are of concern. Many women may suffer anxiety of uncertain duration because of a false-positive screening test. Labeling women with GDM as having an increased risk for future GDM and type 2 diabetes may have psychological implications. Detection of GDM may increase the probability of cesarean delivery; multiple antenatal tests may increase the probability of a false-positive test leading to unnecessary procedures. Costs may be increased with little health benefit for many women, especially those many women with mild hyperglycemia.
It is difficult to see how the issue of screening for GDM can be clarified without large RCTs with an untreated control arm, masked obstetrical care, and measurement of health (not just intermediate) outcomes. These studies should also monitor and report harms and costs associated with screening and intensive treatment. The National Institute for Child Health and Human Development is sponsoring one such study—the Maternal-Fetal Medicine Network multicenter trial of treatment of mild GDM, involving approximately 2,400 women—to be completed in 2004.
Screening for GDM is contentious. The reason for this controversy is largely a lack of high-quality research addressing the central issues. Only good research can end the controversy and inform us how to best serve our patients.
This study was developed by the RTI-UNC Evidence-based Practice Center under contract to the Agency for Healthcare Research and Quality (Contract No. 290-97-0011), Rockville, MD.