3Critical Key Questions & Results

Publication Details

Key Question 1. Does screening for GDM lead to a reduction in perinatal morbidity and mortality for mother and/or infant? A) after 24 weeks gestation? B) during the first trimester and up to 24 weeks gestation?

For this overarching question regarding the benefit of screening and treatment for GDM outcomes, the threshold for evaluating evidence must be higher. Therefore, we required RCT evidence for inclusion for this key question. The ideal study to address the question of whether screening for GDM reduces maternal and/or neonatal morbidity and mortality would be an RCT in which a group of women is not screened and another is screened and, if diagnosed with GDM, treated. No such study for GDM screening was identified. We believe it is unlikely that such a study will ever be conducted in the future in the U.S. given the relatively common current clinical practice of GDM screening and institutionalized ethical constraints for research in human subjects.

Key Question 2. What are the sensitivities, specificities, reliabilities, and yields of current screening tests for GDM: A) after 24 weeks gestation? B) during the first trimester and up to 24 weeks gestation?

Summary. No studies were identified that reported the sensitivity or specificity of GDM screening for the primary maternal and neonatal health outcomes outlined in the analytic framework (Figure 1). Therefore, no articles met inclusion criteria for this key question.

Evaluating screening test performance in GDM is complicated by multiple different accepted standards for screening tests (one-step vs. two-step approach), diagnostic tests (75 g 2 hour OGTT vs. 100 g 3 hour OGTT), and diagnostic criteria (NDDG vs. C&C, see Table 1 for specific cutoff values). Test performance can be evaluated only in the context of how well it accurately identifies people with disease (sensitivity) and excludes those without disease (specificity). With GDM, the “disease” is actually many potential outcomes — and for two different people (mother and baby). Additionally, the primary outcomes against which we were designated to measure test performance (e.g., stillbirth, neonatal death, brachial plexus injury, see Analytic Framework (Figure 1), are rare events, which makes estimates unstable except in a very large study, such as the ongoing Hyperglycemia and Adverse Pregnancy Outcome (HAPO)trial. We found no available evidence that reported sensitivity and specificity for our primary health outcomes — only for the more prevalent macrosomia, which was not a primary outcome. Although we found no studies that met inclusion criteria, we will briefly discuss the limited available data that did not meet inclusion criteria.

Screening at 24 weeks gestation or more.

Sensitivity and specificity of screening tests. There is no universally agreed upon reference test for the diagnosis of GDM. To evaluate the sensitivity and specificity of the 50 g GCT, 75 g OGTT, and 100 g OGTT, we required studies that used perinatal morbidity/mortality measures (primary outcomes) as reference standards and limited our search to current screening and diagnostic tests recommended by the ADA, ACOG, or WHO (Table 1).

Of the studies we considered for inclusion, two cohort studies, one retrospective31 and one prospective32 provided data from which sensitivity and specificity of screening for GDM at ≥24 weeks could be calculated for at least one of the primary outcomes, although these results were not reported in the articles themselves. A third study was identified that used macrosomia as the reference standard for assessing the sensitivity and specificity of screening for GDM at ≥24 weeks with the 50 g GCT, 75 g OGTT, and 100 g OGTT.33 A fourth study conducted in a racially and ethnically diverse population provided only the sensitivity and specificity of the 50 g GCT to detect GDM based on the diagnosis made at 24 or more weeks by 100 g OGTT (C&C criteria). These studies were summarized but ultimately excluded because the authors did not provide the sensitivity and specificity of the screening tests for the primary outcomes and/or because macrosomia, an intermediate outcome, was used as the reference standard.

De Sereday and colleagues used macrosomia, defined as ≥ 4000 g, as the reference standard for evaluating the sensitivity and specificity of the 50 g GCT, 75 g OGTT, and 100 g OGTT.33 In this study of 99 primarily Caucasian women at high-risk of GDM with a mean BMI was 30.8 (±5.6) kg/m2, the sensitivity of the 50 g GCT using a cutpoint of 140 mg/dl was 58.3 percent. Using a cutpoint of 137 mg/dl, the sensitivity was marginally higher, 66.7 percent. For the 75 g OGTT, sensitivities were 41.7 percent and 66.7 percent using cutpoints of 140 mg/dl and 119 mg/dl, respectively. The sensitivity for the 100 g OGTT (GDM using NDDG criteria) was 27.3 percent. The specificities of the tests were 67.8 percent (140 mg/dl cutpoint 50 g GCT), 63.2 percent (137 mg/dl cutpoint 50 g GCT), 90.8 percent (140 mg/dl cutpoint 75 g OGTT), 64.4 percent (119 mg/dl cutpoint 75 g OGTT), and 96.5 percent (2 or more abnormal values of 100 g OGTT). The sensitivities of the 75 g or 100 g OGTT diagnostic tests for detecting macrosomia (≥4,000g) were less than the 50 g GCT by current accepted cutoff values (≥130 mg/dl or ≥140 mg/dl), but the sensitivity for the 50 g GCT was still only 58 percent with the 140mg/dl cutoff.33 In contrast, the specificity was better for either OGTT test (both ≥90 percent specific) than for the 50 g GCT, which had a specificity of 67.8 percent with the 140 mg/dl cutpoint. The OGTT is very specific but not very sensitive, and preceding it by a 50 g GCT increases the sensitivity to a moderate level (but with many more false positive tests after the first test).

Sensitivity calculations for macrosomia (≥ 4000g) based on the prospective study by Deerochanawong and colleagues were 21.4 percent for the 100 g 3 hour OGTT using NDDG criteria compared to 42.9 percent for the 75 g 2-hour OGTT using WHO criteria.32 Sensitivity for stillbirth, a very rare event, was 0 percent for both tests. For neonatal hypoglycemia, we calculated that the sensitivity of the 100 g OGTT was 40 percent and the sensitivity of the 75 g OGTT was 60 percent. For hyperbilirubinemia, we calculated that the 100 g OGTT had a 3.3 percent sensitivity compared to 15 percent for the 75 g OGTT. Calculated specificities for these outcomes (macrosomia, hypoglycemia, hyperbilirubinemia and still birth) ranged from 84.2 to 99.9 percent with NDDG testing criteria yielding specificities >10 percent higher than WHO for all outcomes.

In a retrospective medical record review of a community-based population, Schwartz compared rates of macrosomia (defined in two ways: ≥ 4000 g and ≥ 4500 g), cesarean delivery and stillbirth for screening threshold of 140 mg/dl for 50 g GCT and NDDG and C&C criteria.31 Sensitivity was < 30 percent for all outcomes regardless of screening test used. Specificity was > 80 percent for all outcomes and all tests. Women in this study were primarily Caucasian and were screened at approximately 28 weeks gestation. A total of 18.7 percent had 50 g GCT > 140mg/dl.

Reliability of current screening tests. No articles that evaluated the reliability or reproducibility of GDM screening tests met inclusion criteria. Two articles that tested the reproducibility of the 50 g GCT and the 100 g OGTT34,35 were excluded due to small sample size, samples not representative of the US population, and sparse distribution of outcomes leading to unreliable statistics.

Yields of current screening tests. Using the 75 g oral GTT, de Sereday and colleagues reported a GDM prevalence of 14 percent at a mean gestation of 27.4 (±5.9) weeks.33 This is comparable to the prevalence of 15.7 percent reported by Deerochanawong for screening between 24 and 28 weeks.32

Of those studies that tested for GDM using the 100 g 3 hour OGTT, GDM prevalence ranged from 1.4 to 3.2 percent using the NDDG criteria.31,32 Whereas, in studies that used the less conservative C&C criteria, the prevalence ranged from 4.9 to 6.3 percent.20,31,33

The studies by de Sereday and Deerochanawong compared the yields from both the 50 g GCT followed by the 100 g OGTT using ADA criteria or NDDG criteria to the 75 g OGTT using WHO criteria.32,33 Yields of GDM diagnoses based on WHO criteria (14 to16 percent) were substantially higher than those based on NDDG (1.4 percent) or ADA (6 percent) criteria. A Brazilian cohort study of 4,977 women diagnosed with GDM between 20–28 weeks gestation by the one-step 75 g OGTT found a prevalence of 2.4 percent GDM (95 percent CI 2.0–2.9) by ADA criteria with the 75 g OGTT and 7.2 percent by WHO criteria (95 percent CI 6.5–7.9).36

In a study by Esakoff and colleagues conducted in a diverse population, the prevalence of GDM based on the 50 g GCT and 100 g OGTT (C&C criteria) was 6.3 percent. Stratified by ethnicity, the prevalence of GDM was 4.1 percent in Caucasian, 4.3 percent in African American, 7.0 percent in Latina, and 9.7 percent in Asian.

Screening prior to 24 weeks gestation.

Sensitivity and specificity of screening tests. No articles were identified that reported the sensitivity and specificity of the included GDM screening tests at <24 weeks gestation for our specified health outcomes.

Reliability of current screening tests. No articles were identified that evaluated the reliability or reproducibility of any GDM screening test administered prior to 24 weeks gestation.

Yields of current screening tests. One study that evaluated the ability of the 75 g OGTT measured at ≤16 weeks gestation to predict GDM diagnosis at 24–28 weeks or at 32–34 weeks based on the same test was excluded because it was conducted in a very high-risk Hungarian population that was not representative of primary care practice in the United States.37 This study by Bito and colleagues consisted of 155 women who were considered to be at increased risk for GDM and who were referred to the Diabetic Pregnant Outpatient department in Szeged, Hungary.37 A 2 hour 75 g oral GTT was conducted at ≤16 weeks gestation and again at 24 to 28 weeks and 32 to 34 weeks gestation. Women who tested positive for GDM based on WHO criteria in an early test were not subsequently tested at later gestations. Testing was performed after a 3-day carbohydrate load followed by a 10 to 12 hour fast. The prevalence of GDM was 4.9 percent at ≤16 weeks, 19.6 percent at 24 to 28 weeks, and 29.4 percent at 32 to 34 weeks.

The upcoming results of the HAPO trial may provide new evidence to inform this question.

Key Question 3. Does treatment for GDM lead to reduction in perinatal morbidity and mortality for mother and/or infant? A) after 24 weeks gestation? B) during the first trimester and up to 24 weeks gestation?

Summary. Nine articles were included for this question: eight RCTs3844 for treatment after 24 weeks gestation and one prospective cohort45 of treatment outcomes in women diagnosed at the first prenatal visit compared to 24 weeks gestation or later. A summary of the study population characteristics and primary outcomes of these studies are available in Tables 2 and 3. Further details are available in the Evidence Tables (Appendix C Table 1).

Table 3. Health outcomes of treatment trials (Key Question 3).

Table 3

Health outcomes of treatment trials (Key Question 3).

We found two RCTs that tested treatment versus no treatment of GDM in screen-detected populations and met inclusion and quality-rating criteria—one recent (the Australian Carbohydrate Intolerance Study in Pregnant Women [ACHOIS]) and the sentinel O'Sullivan from over 4 decades ago that laid the groundwork for evidence in this field. 39,44 Both of these trials randomized subjects to treatment versus no treatment of GDM on the basis of a universal screening program approach.

The ACHOIS trial reported that dietary management, glucose monitoring, and insulin treatment as needed in 1000 women with mild GDM diagnosed after 24 weeks gestation improved composite, and individual, neonatal and maternal outcomes compared to no treatment.39 Perinatal mortality, although rare, did not occur in any (0/490) mothers treated, compared to five total stillbirths/neonatal deaths in non-treated (5/510). As glucose control was not part of data collection46 and was not reported, we cannot estimate the relative impact of glycemic control (vs. weight control) on improving outcomes with treatment -—only that treatment improved outcomes.

The fair-quality RCT by O'Sullivan and colleagues44 found that treatment in a screened population of women at high risk for GDM (gestational age at screening unspecified) reduced the intermediate outcome of macrosomia, but without differences in perinatal mortality rate with treatment. Treatment was a small daily dose of insulin (10 units) initially, with irregular glucose monitoring of urine and blood (as this was not available 40 years ago). In contrast, the ACHOIS study participants used insulin only if other therapies failed to achieve tight glycemic control based on study glucose targets, and only 17 percent of the treatment group required insulin.

In addition to the ACHOIS results, we found five fair or good quality GDM treatment trials of various therapies including oral hypoglycemic therapy42,47 and insulin analogues. These trials were reported in six articles that were either newly located or taken from the previous 2003 USPSTF review. These six included articles were heterogeneous in the treatments used and study populations, so synthesizing results in a meta-analysis was not possible. The trials that showed improved glycemic control also found improvements in some (but not all) neonatal and maternal outcomes.

We identified no RCTs for screening and treatment prior to 24 weeks gestation in high-risk women. Therefore, we searched for articles of the next best level of evidence, prospective cohort studies. One fair-quality prospective cohort study of early screening and treatment for GDM was identified in a consecutive population of 3,986 women in Spain screened at the first prenatal visit, and then again at 24 to 28 weeks gestation in those women normal at the initial screen. Its results suggest that an early diagnosis of GDM may represent pre-gestational diabetes as women diagnosed early were more likely to require insulin and had a higher proportion of hypertension, perinatal deaths, and neonatal hypoglycemia than those diagnosed late.

Study Details.

Diagnosis and Treatment at 24 weeks gestation or more.

RCTs of Treatment versus No Treatment of GDM in Screen-Detected Populations. We found one good-quality study from the recent ACHOIS results reported by Crowther and colleagues, a multi-center blinded randomized controlled trial, conducted at 14 sites in Australia, and four sites in the United Kingdom (UK), that compared treatment versus no treatment of mild GDM.39 ACHOIS was designed to determine whether the treatment of mild gestational diabetes would reduce perinatal complications and to assess the effects of treatment on maternal outcomes, mood, and quality-of-life. Women with chronic disease (except essential hypertension) were not eligible to participate. Inclusion criteria were a singleton or twin pregnancy at 16–30 weeks gestation and positive screening for GDM, which was done in two steps. Step 1: Positive risk factors for GDM or a positive 50 g GCT (≥ 7.8 mmol/l [140 mg/dl] 1 hour post-challenge; 93 percent of women had a positive GCT). Step 2: a 75 g OGTT was given after an overnight fast, with inclusion criteria (a) fasting plasma glucose of < 7.8mmol/l (140 mg/dl) and (b) 2 hour post-OGTT glucose 7.8–11.0 mmol/liter (140–198 mg/dl). At the time of the study, the WHO classified these glucose criteria as glucose intolerance of pregnancy (i.e., intermediate between normal and GDM), and thus it was ethical to randomize and evaluate treatment compared to a blinded untreated group. Subsequently, the WHO changed the classification of GDM so that a 2 hour value 7.8–11.0mmol/l is now defined as GDM, and so the results of the ACHOIS trial can provide direct evidence for treatment of this mild GDM by current practice standards.

The 1000 women who met inclusion criteria were randomized by computer-generated numbers, 490 to treatment (who were informed in writing that they had GDM and an intervention plan), and 510 to no treatment (who received a slip indicating they did NOT have GDM, and no follow-up treatment was provided by the study [only as clinically indicated by their provider]). The full numerical results of the OGTT were not released to the women or their providers until after birth.

Women in the intervention group received both individualized dietary advice and instructions to self-monitor glucose four times daily until it was within the specified range for two weeks, and insulin was initiated and titrated as needed. Glucose goals were as follows: fasting of at least ≤3.5 mmol/l (63 mg/dl) and no more than 5.5 mmol/l (99 mg/dl), pre-prandial levels <5.5 mmol/l, and 2 hour postprandial <7.0 mmol/l (126 mg/dl). The care of the women in the intervention group replicated clinical care in which universal screening and treatment for GDM are available. In contrast, the routine-care (non-treated) group replicated clinical care in which screening for GDM is not available.

After randomization, the treated and non-treated groups were similar in age, BMI, race/ethnicity, gestational age at screening (mean 29 weeks), primiparity, history of GDM, and by screening test results on both the 50 g GCT and the 75 g OGTT (Appendix C Table 1). One population characteristic of note is that the women in this study were, on average, slightly overweight (mean BMI approximately 26 kg/m2). However, recent weight estimates for US women of child-bearing age are similar to ACHOIS (current mean BMI 26.8 kg/m2 and 27.9 kg/m2 for US women age 20–29 years and 30–39 years, respectively).11 In the ACHOIS trial about 75 percent of the women were Caucasians (also similar to the US),48 with Asians comprising the next largest race/ethnicity group.

In ACHOIS, the treated group gained significantly less weight (8.1kg) during pregnancy (measured as difference between first and last prenatal visit weight) compared to the non-treated group (9.8 kg, multivariate adjusted p =0.01). No information is available on glucose values during pregnancy in the treated or not-treated group, as this was not part of the study's data collection.46 However, 100 women (17 percent of the intervention group) required insulin therapy; there were 17 (3 percent) in the non-treated group whose physicians began insulin for clinical indications. Results are presented by intention-to-treat in the article tables (including all women randomized).

The treatment group had one-third the overall risk of the composite outcome of any serious perinatal complication, and this remained significant after adjustment for maternal age, race, and parity (RR 0.33 [95 percent CI 0.14–0.75]). Serious perinatal complications were defined as any of the following: death, shoulder dystocia, bone fracture, and nerve palsy. The absolute rates of these individual perinatal outcomes were also reported in the paper, but could not be compared between groups due to no events for death, bone fracture, or nerve palsy in the treatment group. Overall, there were seven infants with serious perinatal complications in the treatment group (all shoulder dystocia), compared to 23 infants with perinatal serious complications in the non-treated group (five deaths, one fractured humerus, three nerve palsies, and 16 shoulder dystocia [25 total events, but calculated as 23 infants in the analysis as one had both a fractured humerus and a radial-nerve palsy and another infant had both shoulder dystocia and Erb's palsy in the non-treated group]). Shoulder dystocia was not a specified health outcome for this evidence review. The remaining components in the composite outcome (neonatal death, fracture, nerve palsy) were final health outcomes specified by the Task Force for this review. The causes of the five deaths in the untreated group were: two stillbirths (unexplained intrauterine deaths at term of appropriately grown infants), one stillbirth at 35 weeks gestation associated with pre-eclampsia and intrauterine growth restriction, one infant died from asphyxia during labor without antepartum hemorrhage, and one had a lethal congenital anomaly.

The majority of infants in both groups were admitted to the neonatal nursery and differed by treatment group: 357/506 (71 percent) in the treated group, and 321/524 (61 percent) in the non-treated group (adjusted RR 1.13 (1.03–1.23). The length of stay in the neonatal nursery among the infants admitted did not differ significantly between groups (median of 1 day for both groups; interquartile range, 1 to 2 days in the intervention group and 1to 3 days in the routine-care group; adjusted p=0.81).

The rate of admission to the neonatal ICU was not specifically reported, but treatment for the relevant specified health outcomes for this evidence report (hypoglycemia, hyperbilirubinemia, respiratory distress syndrome) were reported individually. There was no significant difference in infants requiring intravenous therapy for hypoglycemia after birth based on mother's treatment group: 35/506 (7 percent) in the treated group, and 27/524 (5 percent) in the routine-care group (adjusted RR 1.42 [95 percent CI 0.87–2.32]). There was no significant difference in risk of needing phototherapy for jaundice among infants whose mothers were treated or not treated for GDM: 44/506 (9 percent) in the treated group, and 48/524 (9 percent) in the routine-care group (Adjusted RR 0.93 [95 percent CI 0.63–1.37]). Similarly, risk of respiratory distress syndrome in the neonate (needing supplemental oxygen > 4 hours after birth) was not significantly different based on mother's GDM treatment 27/506 (5 percent) in the treated group, and 19/524 (4 percent) in the routine-care group (Adjusted RR 1.52 [95 percent CI 0.86–2.71]).

In addition to having significantly less weight gain during pregnancy, women in the treatment group had a 30 percent lower risk of pre-eclampsia (defined as blood pressure > 140/90 mm Hg more on two occasions more than four hours apart) compared to women who were not treated for GDM: (58/490 [12 percent] in the treated group and 93/510 [18 percent] in the untreated group; Adjusted RR 0.70 [95 percent CI 0.51–0.95]).

Other outcomes assessed with ACHOIS that were not part of our key question are summarized here. Infants of the treated mothers had a modest reduction of a mean 145g in birth weight (3335 g vs. 3482 g, p< 0.001) compared to those whose mothers were not treated. The proportion of large babies was also significantly reduced based on mothers' treatment for GDM when measured either as LGA (defined as >90th percentile) or macrosomia (≥ 4000 g). The infants of mothers treated for GDM had about half the rate of macrosomia compared to those whose mothers were not treated (Adjusted RR 0.47 [95 percent CI 0.34–0.64]). Shoulder dystocia (as reported by the primary caregiver) occurred in seven babies whose mothers were treated, compared to 16 babies whose mothers were not treated (Adjusted RR 0.46 [95 percent CI 0.19–1.10]). There were no significant differences in other infant outcomes based on mother's treatment group for GDM (i.e., small for gestational age, 5-minute Apgar score < 7, neonatal convulsions).

Women treated for GDM were more likely to be induced for labor (189/506 [39 percent]) compared to women not treated (150/524 [29 percent]), Adjusted RR 1.36 [95 percent CI 1.15–1.62]. Women treated for GDM also had a slightly earlier gestational age at birth, which was statistically significant (mean 39.0 vs. 39.3 weeks, p=0.01). Overall rates of cesarean, however, did not differ in the treated group (152/506 [31 percent]) compared to the untreated group (164/524 [32 percent]); Adjusted RR 0.97 (0.81–1.16). The lack of difference by treatment group remained when c-sections were stratified by indication (emergency or elective). There were also no differences in other maternal outcomes by treatment group (i.e., any perineal trauma, puerperal pyrexia (≥ 38 degrees Celsius), length of postnatal stay, or proportion breast-feeding at discharge).

The fair-quality rated RCT of screening for GDM in women at high risk for diabetes mellitus was reported in 1966 by O'Sullivan and colleagues.44 The authors screened 943 women with a 1 hour 50 g GCT, followed by a 3-hour 100 g OGTT. If the women had a GCT value of ≥ 130mg/dl or one or more risk factors for GDM, they underwent a 3 hour 100 g OGTT. Women were ineligible for the study if they had previously been diagnosed with diabetes, had blood sugars exceeding 300 mg/dl, had classic diabetic symptoms, or registered for prenatal care at ≥37 weeks gestation: 615 women tested positive for GDM, and were randomly allocated to treatment (n=307) and no treatment (termed ‘positive controls’; n=308) A third group of women (termed ‘negative controls’, n=238) was selected randomly at regular intervals and had to have completely normal glucose tolerance. We will discuss only the results of the women with GDM randomized to treatment versus no treatment in the text as this normal group was not randomized and reported results for all three groups were unadjusted (see Appendix C Table 1 for further details). The gestational age at screening was not specifically reported, but early screening or screening upon entry was implied as the authors stated: “women who had normal glucose tolerance in one trimester received repeat tests in subsequent trimesters.”44

Women who were treated for GDM received an individualized diet (40 percent carbohydrates, 30 calories/kg ideal body weight, and 1.5–2 g protein/kg ideal body weight) and 10 units of NPH insulin once each morning. The insulin dose increased if glycosuria was noted by tests performed daily at home or during a clinic visit. At the time of this study, home capillary blood glucose monitoring was not yet available. The untreated GDM group received routine prenatal care. The treated GDM patients did not differ from untreated GDM in mean postprandial blood sugars, except for between 1 and 2 hours post-prandial (88.8 vs. 92.6 mg/dl, p<0.01), but they did have significantly lower fasting blood sugars (69.1 vs.74.3 mg/dl, p<0.05).

The perinatal loss rate was 4.3 percent for treated GDM patients and 4.9 percent for untreated GDM patients (p=non-significant). The number of infants weighing nine pounds or more at birth (macrosomia) was three times higher in the untreated group (13/305 [4.3 percent] viable deliveries in treated versus 40/306 [13.1 percent] in the untreated group, p=not reported). There were no significant differences between treated and untreated GDM in the number of infants diagnosed with congenital anomalies (13.6 percent vs. 16.0 percent) or delivered preterm (8.5 percent vs. 7.8 percent). When stratifying by weight (either normal or underweight vs. 10 percent overweight) and comparing treated and untreated GDM patients, the authors found those who were treated were less likely to have large babies and that the relative reduction was greatest in lean women (2.3 percent vs. 10.0 percent among normal or underweight women and 7.6 percent vs. 16.4 percent for overweight women, p=not reported). For both treated and untreated GDM, women who were overweight were more likely to have large babies.

This fair-quality RCT must be considered in the historical context of clinical care 4 decades ago.44 At that time, home blood glucose meters (and thus regular monitoring of glycemic control) were not available, so it was not possible to tightly control glucose levels—only to look for hyperglycemia severe enough to cause “overflow” of glucose into the urine that can then be detected by a urine test strip. Also, significant (life-threatening) maternal hypoglycemia was a greater risk in an era when subjects were not able to accurately monitor blood glucose levels regularly or accurately. The treatment options with shorter-acting insulins that clinicians have today were also not available in the 1960s.

The rate of macrosomia observed by O'Sullivan in the GDM group treated with once daily insulin was significantly lower in treated versus untreated GDM patients (4.3 percent vs. 13.1 percent). In contrast, there was no significant difference observed in rate of fetal or neonatal death. There are several possibilities for this discordance in treatment effect including power, since rare events require a very large sample to detect a difference. One possibility is that treatment does not have an effect on mortality risk. Another is that the physiologic changes of normalizing fasting glucose reduces macrosomia, but normalization of post-prandial levels is also required to affect mortality risk. One limitation to note is that mean glucose values were calculated from 2701 measured blood sugars (an average of six blood draws per woman during pregnancy). These were likely not as representative of mean glucose values that can currently be assessed by home glucose monitoring.

Although it is implied from O'Sullivan that early screening occurred (as testing was repeated in each trimester if negative on the initial screening), we found no RCTs that directly compared screening at <24 weeks with screening at ≥24 weeks gestation. One fair-quality prospective cohort study was identified that reported results for women who were serially screened for GDM, and is detailed below under early screening.45

RCT of Treatment Comparisons for GDM. The six included articles from five randomized controlled trials for KQ3 (one good-quality, five fair-quality) compared different treatment strategies for GDM. Given the treatments involved, it was not feasible to blind the subjects to type of treatment (e.g., insulin before or after a meal). It was also not possible to synthesize the results as treatments were heterogeneous.

The best comparative evidence came from one good-quality RCT reported by Langer and colleagues that compared perinatal outcomes with treatment of GDM with the oral hypoglycemic agent glyburide versus the standard treatment of insulin (note: Glyburide is not currently approved by the Food & Drug Administration (FDA) for use in GDM).47 Women with GDM and singleton pregnancies who attended maternal health clinics in San Antonio, Texas (83 percent Hispanic, 12 percent White, 5 percent Black), and required medical treatment for their GDM were randomized (n=404) to treatment with either glyburide or insulin. Outcomes evaluated were glycemic control and maternal and neonatal complications.

Gestational diabetes was diagnosed by the two-step method among women with singleton pregnancies at 11–33 weeks gestation. Step 1: A 50 g GCT was performed, and those with a 1-hour plasma glucose > 130 mg/dl had a 100 g OGTT. Step 2: Two or more abnormal values on the 3-hour OGTT by C&C criteria were diagnostic of GDM. Women with fasting plasma glucose (FPG) < 95 mg/dl were initially treated with diet alone, but were later eligible for randomization to medical treatment if their FPG became > 95mg/dl or they had postprandial plasma glucose levels ≥120 mg/dl. The majority of women were defined as obese (BMI>27.3 kg/m2), 70 percent and 65 percent in the glyburide and insulin-treated groups, respectively. The two randomized groups were also similar in age, nulliparity, gestational age at screening (mean 24 and 25 weeks for glyburide and insulin-treated), history of prior GDM, and screening test results.

Both randomized groups received nutritional instructions for three meals and four snacks a day and instructions in glucose monitoring, with glycemic goals for titration of medication. The glyburide group was initiated on a 2.5 mg dose of glyburide in the morning, and increased as needed by 2.5 mg up to a 20 mg daily dose. The average dose of glyburide a day was 9 mg (± 6 mg). Eight women (4 percent) on glyburide did not achieve good glycemic control and were switched to insulin. The insulin group received an average daily dose of 85 units/day (±48 units). The mean glycosylated hemoglobin was 5.7 for the glyburide group, and 5.6 for the insulin-treated group (p=0.42). Glucose control during pregnancy also did not differ between the two groups with glucose monitoring (measured as fasting, pre-prandial, postprandial, or mean glucose). However, women in the glyburide group were significantly less likely to have hypoglycemia (<40 mg/dl) during pregnancy. Specifically, only four women in the glyburide group (versus 41 women in the insulin group) experienced hypoglycemia (p=0.03). None of the women in either group reported severe symptoms with hypoglycemia. Weight gain during pregnancy (week prior to delivery minus pre-pregnancy weight) also did not differ with glyburide versus insulin treatment (mean 21 kg for both groups).

There were no differences in any of the neonatal outcomes based on maternal treatment with glyburide or insulin. Specifically, perinatal mortality rates (stillbirth or neonatal death), metabolic outcomes (NICU admission, need for IV therapy for hypoglycemia, hyperbilirubinemia, polycythemia, hypocalcemia, lung complications, need for respiratory support, congenital anomalies), and anthromorphometric features (birth weight, proportion with macrosomia or LGA) did not differ by treatment group.

A secondary analysis by Langer and colleagues of the above RCT was recently published and was rated as fair quality. Their analysis compared outcomes (both for glyburide and insulin-treated groups) stratified by whether the fasting glucose on the diagnostic OGTT was ≤ 95 mg/dl vs. > 95 mg/dl.47 Consistent with the results of O'Sullivan in 1966,44 a normal fasting glucose was associated with a significant reduction in LGA babies in both glyburide and insulin treatment groups (18 percent LGA if diagnostic OGTT fasting glucose was > 95 mg/dl in both treatment groups, and 7–8 percent LGA if OGTT fasting glucose was ≤ 95 mg/dl), but level of maternal fasting glucose did not show any difference for either treatment group in neonatal complications including metabolic complications or a composite neonatal outcome. The composite neonatal outcome was defined as any of the following: metabolic complications (neonatal hypoglycemia, hyperbilirubinemia, polycythemia); LGA/macrosomia; neonatal intensive care unit admission >24 hours; the need for respiratory support). Appendix C, Table 1 provides a more detailed explanation.

Bancroft and colleagues reported a fair-quality small randomized controlled pilot study in the UK to evaluate neonatal outcomes in 68 women with mild GDM treated with diet and home glucose monitoring up to four times daily compared to diet without home glucose monitoring.38 Both groups received dietary counseling and monthly glycosylated hemoglobin testing (though glycosylated hemoglobin results were not made available for the standard care group during the study). The glucose monitored group had significantly lower 2 hour OGTT levels at study entry, and achieved significantly lower glycosylated hemoglobin levels only at the 32 weeks measurement point, compared to standard treatment (glycosylated hemoglobin was generally lower but not significantly different at 28 weeks, 36, 38 weeks, or at term). The rates of admission to the special care baby unit (the primary outcome) were 2/32 [6 percent] in the glucose monitored group, and 6/36 [17 percent] in the standard care group, and did not reach statistical significance. One shoulder dystocia in the unmonitored group resulted in admission to a special baby care unit but no long-term consequences. The frequency of hypoglycemia was 2/32 [6 percent] in the glucose monitored group, and 6/36 [17 percent] in the standard care group, which did not reach statistical significance. There were no stillbirths or neonatal deaths in either group. Other neonatal outcomes were not notably or significantly different (gestational age at delivery, birthweight, LGA [> 90th percentile]). The authors acknowledged the lack of power to assess the significant differences in outcomes between the two treatments with the small sample size, and concluded that they had demonstrated the feasibility of a larger study, which was then commencing with ACHOIS.

Jovanovic and colleagues randomized 42 women with GDM (95 percent Hispanic) who required medical treatment into two groups comparing treatment with NPH+lispro insulin versus NPH+regular women.41 While the trial was small and designed to assess differences in insulin antibodies, and primarily provides information regarding lack of harm with treatment (KQ5), there were none of the following complications in either of the treatment groups: neonatal hypoglycemia or hypocalcemia, fetal abnormality, or macrosomia (> 90th percentile). There were no statistically significant differences in rate of cesarean delivery, gestational age at delivery, or newborn 1- and 5-minute Apgar scores.

Nachum and colleagues randomized 274 women in Israel with gestational diabetes who required insulin treatment, diagnosed at a mean 26 weeks gestation, to insulin four times daily (regular insulin before three meals and an intermediate duration insulin before bedtime) versus insulin twice daily (mixed dose of intermediate and regular insulin morning and evening).43 Baseline characteristics, including BMI (mean 28 kg/m2), were similar in both treatment groups after randomization. With intensified treatment, the four-times-daily insulin treatment group had significantly better glycemic control (mean daily glucose, HbA1c, and fructosamine) than the twice-daily insulin treated group. The HbA1c values were 5.5 percent and 5.8 percent in the four-times-daily and twice-daily insulin treatment groups, respectively. Moreover, 91 percent of the four-times-daily insulin group reached target mean glucose values (< 5.8 mmol/l) versus only 74 percent of the twice-daily insulin group. Of note, this excellent glycemic control did not increase severe maternal hypoglycemia (requiring help from another person); 1/138 and 1/136 women in the four-times-daily and twice-daily insulin groups experienced severe hypoglycemia. Neonatal hypoglycemia and hyperbilirubinemia were both significantly reduced in the intensified four-times-daily insulin versus twice-daily insulin maternal GDM treatment groups (RR 0.12[95 percent CI 0.02–0.97] and RR 0.51[95 percent CI 0.29–0.91], respectively, for hypoglycemia and hyperbilirubinemia). Neonatal hypoglycemia was defined as plasma glucose <1.9 mmol/l in term infants or <1.4 mmol/l in preterm infants ≥ 2 occasions in first 48 hours of life. Hyperbilirubinemia was defined as >205 mmol/l at >=34 weeks gestation or >137 mmol/l at <34 weeks gestation. There was only one perinatal death, and it occurred with a mother who was in the twice-daily insulin (less intensive) treatment group. Overall neonatal morbidity rates were reduced by half in the four-times-daily versus twice-daily insulin treatment groups (RR 0.51[0.29–0.91]). The authors did not specify which elements were combined in this composite overall morbidity.

Finally, the last fair-quality RCT was a small trial of 66 women who required insulin treatment for GDM, randomized to postprandial versus preprandial glucose monitoring to guide insulin dose titration.40 Baseline characteristics were similar after randomization, including BMI, gestatational age at diagnosis and treatment, 1 hour 50 g GCT and fasting plasma glucose on the 3 hour diagnostic 100 g OGTT. Weight gain during pregnancy and percent achieving glycemic control goals were the same in both treatment groups. In this context, while there were no differences in glycosylated hemoglobin at baseline (8.9 percent vs. 8.6 percent, p=0.55), the final glycosylated hemoglobin was both significantly improved and different in the group with postprandial monitoring compared to preprandial monitoring (6.5 percent vs. 8.1 percent, p=0.006). The postprandial group also received more daily insulin than the preprandial group (1.1u/kg vs. 0.9 u/kg, p=0.0001). The rate of neonatal hypoglycemia (defined as <= 30 mg/dl [1/7 mmol/l]) was 1/33 [3 percent] vs. 7/33 [21 percent] in the post versus preprandial treatment groups (p=0.05). Macrosomia (>4 ,000 g) was also dramatically reduced in post versus preprandial groups (9 percent vs. 26 percent of babies in each treatment group, p=0.01). There was only one stillbirth, which occurred in the group with less intensive glycemic control (in this case the preprandial group). Cesarean for cephalopelvic disproportion was reduced in the postprandial versus the preprandial treatment groups (12 percent vs. 36 percent of women, p=0.04). There were no differences in rates of pre-eclampsia in the two groups. In summary, while this was a small RCT, it found significant improvements in glycemic control and reduction in neonatal hypoglycemia and macrosomia, and there was no apparent increased harm associated with this improved glycemic control. Given that the initial HbA1c values were high in both groups (more severe GDM), and that the postprandial group had both greater improvement in hyperglycemia and higher doses of insulin used for treatment, it is not clear if it was improvement in glycemic control or timing of the treatment (post vs. preprandial) that resulted in improved health outcomes.

Diagnosis and Treatment prior to 24 weeks gestation.

Prospective cohort study early vs. late screening. Bartha and colleagues administered a 50 g GCT (cutoff 140 mg/dl) to 3986 consecutive pregnant Spanish women at their first antenatal visit (early-onset).45 Abnormal results were followed by administration of the 100 g 3-hour OGTT (NDDG criteria). Women with negative testing at the first visit were retested again at 24–28 weeks (late-onset). Women diagnosed with GDM were hospitalized and capillary glucose values were assessed, and those with pre-prandial glucose levels of <105mg/dl and 2-hour postprandial glucose concentrations of <120mg/dl were given only dietary recommendations. Insulin therapy was initiated for women who did not meet these criteria. The mean gestational age at hospitalization was 18.1±6.5 weeks for those diagnosed early and 33.1±3.9 weeks for those diagnosed late (p<0.000001). Of 3986 women, 65 (1.6 percent) were diagnosed early with GDM and 170 (4.3 percent) were diagnosed later with GDM.

Women with early-onset gestational diabetes were more likely to have hypertension (18.5 percent vs. 5.9 percent, p=0.006), largely due to a high rate of pre-existing chronic hypertension (10.8 percent vs. 2.4 percent, p=0.01). With all cases of pre-eclampsia analyzed together (pre-eclampsia plus superimposed pre-eclampsia), the rate was significantly higher in the early-onset group (6.2 percent vs. 0.6 percent, p=0.02). The authors did not specify the definitions used for hypertension, pre-existing hypertension, pre-eclampsia, or superimposed pre-eclampsia.

There were no significant differences in most pregnancy outcomes (cesarean delivery, preterm birth, 5-minute Apgar < 7, mean neonatal weight, fetal weight > 4000g or < 2500g, meconium passage, and admission to special care baby unit) between those diagnosed early and late. The neonates of women diagnosed early were more likely to have hypoglycemia (8 percent vs. 0 percent, p=0.005) and perinatal death (6 percent vs. 0 percent, p=0.02). The definition of neonatal hypoglycemia used was also not specified.

Women with early-onset GDM differed significantly (p<0.05) from those with late-onset GDM in all but one measure of glycemic control (glycosylated hemoglobin). Women with early-onset? GDM had higher mean fasting glucose levels, higher mean 2 hour postprandial glucose levels (after breakfast, lunch, and dinner), and higher mean pre-dinner glucose levels. In addition, 33.9 percent of women who were diagnosed early required insulin compared to 7.1 percent of those diagnosed late (p<0.00001).

This single study of early screening suggests an early diagnosis of GDM may represent pre-gestational diabetes as women diagnosed early were more likely to require insulin and had a higher proportion of perinatal deaths and neonatal hypoglycemia than those diagnosed at 24 weeks gestation or later.

Key Question 4. What are the adverse effects associated with screening for GDM?

Summary. The primary adverse effects associated with screening would be the psychological impact of screening to the mother with GDM - and potentially to the mother who does not have GDM but has the added time, cost, and psychological burden of screening. A review of the literature revealed that available evidence is mixed in terms of the initial psychological impact of GDM screening. In the first few weeks after screening, women who screen positive for GDM may report higher anxiety, more psychological distress, and poorer perceptions of their general health than women who screen negative. Available evidence, however, suggests that these differences, even if present shortly after diagnosis, do not persist into the late third trimester or postpartum period.4951

Study Details. Three fair quality articles, two prospective cohort studies, and one cross-sectional study met inclusion criteria (Appendix C Table 4).4951

Table 4. Summary of evidence.

Table 4

Summary of evidence.

Rumbold and Crowther serially assessed 209 Australian women (two-step method: 50 g GCT, 75 g GTT, WHO criteria) at 24–28 weeks and again toward the end of the third trimester at about 36 weeks: 150 women who screened negative on the OGCT, 37 who had a positive GCT screen but normal OGTT, and 25 women diagnosed with GDM (2-hour glucose>11.1 mmol/l after a 75 g OGTT).49 The validated measures used in the questionnaire were the Spielberger State-Trait Anxiety Inventory (STAI), the Edinburgh Postnatal Depression Scale (EPDS), and the Short Form 36 Item Health Survey (SF-36).49,5254

No differences were found for the mean STAI scores after screening, between women screening negative or positive. Similarly, none of the groups differed from those that screened negative in the late third trimester.

No differences in rates of depression (EPDS>12) were found in women after screening or in the late third trimester among the screen negative, false positive GCT, or GDM groups.

For the SF-36 measures, in the first post-screening assessment, women who had negative GCTs had better health perceptions, lower vitality, and were more likely to rate their health as much better than one year before compared to women who screened positive with GCTs. They did not differ in any of the other six SF-36 health status domains. In the late third trimester, women who had a negative GCT reported less vitality than women who had a positive GTT, and greater social functioning than women who had a false positive GCT; these groups did not differ in any other domain or in health rating compared to one year before. After screening, women with negative GCTs were more likely than those with positive GCTs to rate their experience of screening as positive (77 percent vs. 57 percent, p<0.01), but did not differ in the likelihood of requesting screening during a future pregnancy. Later in pregnancy (towards the end of the third trimester), there were no differences in the experience of screening between women with false positive GCTs and women with positive GTTs (GDM).

Daniells and colleagues conducted a prospective cohort study of 50 women with GDM diagnosed at the beginning of the third trimester and 50 women with normal glucose tolerance.51 During the 30th week of gestation, women diagnosed with GDM had higher mean scores on the Mental Health Inventory 5 (13.9 ± 4.8 vs. 11.4 ± 3.8, p<0.004) and higher mean anxiety scores on the Spielberger State-Trait anxiety inventory (40.6 ±13.3 vs. 34.2 ± 9.9, p<0.007) than women with normal glucose tolerance, indicating greater psychological distress and anxiety. There were no statistically significant differences at 36 weeks of gestation or at 6 weeks postpartum. The GDM and control groups did not differ in their attitudes toward testing for GDM at any assessment period.

Spirito and colleagues used the Profile of Mood States-Bipolar From to assess the psychological status of 68 women with GDM, diagnosed at approximately 28 weeks gestation, and 50 non-diabetic pregnant controls at about 35 weeks of gestational age.50 Women with GDM did not differ from controls on any of the Profile of Mood States-Bipolar form subscales, indicating no differences in emotional status. In addition, the 33 women with GDM who were prescribed insulin did not differ in emotional status from the 33 who were not. None of the Profile of Mood States-Bipolar Form subscales was predictive of glycemic control.

Key Question 5. What are the adverse effects associated with treatment of GDM?

Summary. We identified two potential domains of adverse treatment effect in GDM: physical and psychological. For the mother, hypoglycemia is the potentially most serious (that is, life-threatening or requires assistance to treat). In the psychologic domain, maternal adverse effects could potentially arise from diagnosis and treatment. The potential teratogenicity of certain newer treatments for GDM (oral hypoglycemic agents or insulin analogues) is a potential physical harm to the fetus that clearly could relate to GDM treatment, but this would be treatment-specific, for relative benefits and harms of differing treatment modalities compared to insulin and thus is a sub-question. That is, the primary purpose of this Task Force update is to review the evidence regarding potential benefits and harms of screening and treatment for GDM, not to determine which treatment regimen is preferred. Several of the studies of newer agents assess placental crossing of the treatment modality which we will report, but one must put this in the context that most treatments for GDM began in the second trimester (after the period of major organogenesis),55 and thus data is very limited to assess potential teratogenicity of newer agents for treatment.

Several studies included for treatment benefit also provided evidence for potential harm, with the best evidence again arising from the ACHOIS results. Overall for KQ5, we found two good-quality trials39,42 and five fair-quality studies38,40,41,43,56, including six trials from KQ3. The additional study 56 was a fair-quality prospective cohort study evaluating the emotional adjustment to diagnosis and treatment of GDM. (Appendix C Table 3; Table 2).

Not all studies monitored or reported maternal hypoglycemia, but the rates are rare with treatment and no worse with alternate therapies in those that did. For the psychological domain, the evidence suggests no harm from treatment. The best evidence comes from the ACHOIS trial, which found in a subgroup that responded to the questionnaire that treatment was potentially associated with overall improved self-reported health status and reduced post-partum depression at three months post-partum compared to no treatment. Crowther and colleagues reported that the full numerical results of the OGTT for the ACHOIS were not released to the women or their providers in the treatment group until after birth, but the exact timing is not specified. Alternative explanations for the reduced post-partum depression and improved quality-of-life responses in the treated group could include unblinding prior to the three months post-partum before the questionnaire was completed or what is sometimes termed the Hawthorne effect (in which additional attention given to the treatment group rather than the treatment itself could improve perceptions).57 Finally, a prospective study found that mood did not differ with in women treated for GDM compared to controls.56

In summary, we found no evidence for significant harms associated with treatment with GDM, and it is possible that treatment may impart an additional benefit to maternal quality-of-life.

Study Details.

Treatment versus No Treatment of GDM: The one good-quality article (detailed in study design in KQ1) reported on the ACHOIS clinical trial of treatment for mild GDM. Maternal hypoglycemia rates were not reported for the treatment group, so we cannot assess this outcome in this trial. However, detailed analyses of psychological well-being were done six weeks after diagnosis and three months post-partum, among a subset that responded. At six weeks after diagnosis, 332/490 treated women and 350/510 non-treated women completed a questionnaire about quality-of-life (QOL). Multiple QOL components were measured by the SF-36, a well-validated QOL questionnaire that ranges from zero (worst) to 100 (best) on multiple components.5860 The treated and non-treated groups differed significantly on six components—and all of these differences favored a better QOL (higher score) with treatment. Anxiety was also assessed by the Spielberger State-Trait Anxiety Inventory (with scores below 15 considered normal), and no differences in anxiety were detected at six weeks after diagnosis (mean score 11, or normal, for both groups).

At three months post-partum, 278/490 treated and 295/510 non-treated women completed the QOL questionnaire. Three components on the SF-36 bordered on significant difference (physical functioning, general health, and overall physical component), with these differences favoring better self-ratings with treatment. Mean anxiety levels were normal for both groups (again assessed by the Spielberger State-Trait Anxiety Inventory) and did not differ (mean = 11 for both). Post-partum depression was also assessed at three months with the Edinburgh Postnatal Depression Scale (EDPS; a score above 12 is considered abnormal). In the treatment group, 23 women (8 percent) had an EDPS>12, compared to 50 women (17 percent) in the non-treated group. Thus, the risk of post-partum depression was reduced by half in treated women, i.e., the relative risk of post-partum depression was 0.46 (95 percent CI 0.29–0.73) with treatment of GDM. These results should be interpreted with caution in that, unlike the other ACHOIS results, only a subgroup responded to the QOL questionnaire. The data, however, suggest lack of harm and raise the question of potential benefit of decreased post-partum depression and improved QOL in the mother treated for GDM.

Studies of Treatment Comparisons for GDM. Another good-quality RCT reported by Langer and colleagues evaluated potential harms of glyburide versus insulin, and was also detailed in the treatment (KQ3) section (note: Glyburide is not currently approved by the Food & Drug Administration (FDA) for use in GDM).42 Although glycemic control did not differ between the two treatments (fasting, post-prandial or glycosylated hemoglobin percent), women in the glyburide group were significantly less likely to have hypoglycemia (<40 mg/dl) during pregnancy. Specifically, only four women in the glyburide group, versus 41 women in the insulin group, experienced hypoglycemia (p=0.03). None of the women in either group reported severe symptoms with hypoglycemia.

Evaluation of the glyburide group for safety revealed no detectable glyburide in the cord serum of any infant (mean sampling of the cord blood was 8±4 hours after the last dose of maternal glyburide). In 12 women randomly selected from the glyburide group, glyburide was measured simultaneously in the maternal and cord serum. Maternal serum concentrations were easily detectable (range 50–150 mg/ml), but were undetectable in cord serum. Finally, the authors stratified outcomes in this trial based on whether the women entered the study prior to or after 20 weeks (prior to 20 weeks would be during the period of organogenesis where risk of congenital anomalies is greater). They found no differences in any outcomes based on treatment groups (glyburide vs. insulin).

A fair-quality small study of 42 mostly Hispanic women randomized to NPH+lispro versus NPH+regular insulin, reported by Jovanovic and colleagues, also assessed treatment with lispro, an insulin analogue (which has a theoretical concern of teratogenicity because of the modified amino acid structure that might be metabolized differently than the natural hormone).41 Maternal hypoglycemia (glucose <55 mg/dl) was rare in both groups, and tended to be lower with lispro, but this was only statistically significant with the fasting pre-breakfast measurements (percent of all fasting blood glucoses in the hypoglycemic range was 0.93 percent for regular insulin vs. 0.65 percent for insulin lispro, p=0.025). The primary outcome was antibody response to insulin (because placental transfer of insulin occurs when complexed to immunoglobulin). Neither the lispro-insulin nor regular-insulin treated groups showed a statistically significant change in antibody response with treatment compared to the baseline antibody response for individual patients. In a subset of patients who received a continuous infusion of insulin lispro during delivery, there were measurable maternal concentrations of insulin lispro, but no insulin lispro could be detected in the cord blood, suggesting that insulin lispro does not cross the placenta.

Bancroft compared treatment with diet+glucose monitoring versus diet without glucose monitoring.38 Only six monitored women required insulin; rates of maternal hypoglycemia were not reported. Nachum and colleagues randomized 274 women in Israel with GDM who required insulin treatment four times daily versus twice daily.43 With intensified treatment, the four-times-daily group had significantly better glycemic control (mean daily glucose, HbA1c, and fructosamine) than the twice-daily groups as detailed in KQ3. However, this excellent glycemic control did not increase the rate of severe maternal hypoglycemia (requiring help from another person); 1/138 and 1/136 women in the four-times-daily and twice-daily insulin groups, respectively, experienced severe hypoglycemia.

A fair-quality randomized trial reported by deVeciana and colleagues comparing pre-prandial versus post-prandial glucose monitoring to guide insulin treatment in GDM did not report specific rates of maternal hypoglycemia.40 However, in the text where they report no differences in hospitalization to optimize glycemic control during pregnancy between the treatment groups and similar rates of pre-eclampsia in the groups, the authors note that were no other maternal complications.

The one fair-quality paper that was specifically included for this question (but not for KQ1 or KQ3) was the prospective cohort study by Langer and Langer that evaluated emotional adjustment to GDM diagnosis and intensified treatment.56 Diagnosis was based on an OGTT at a mean gestational age of 28 weeks in 69 diet-controlled and 137 insulin-treated women. These 206 women with newly diagnosed GDM and 95 pregnant women with normal OGTT (controls) were administered the Profile of Mood States Bipolar Test. The women with GDM (both diet controlled and insulin treated) had a mean age of 29 years, and those without GDM had a mean age of 24 years. Obesity rates (definition not specified) were 20 percent in diet controlled, 50 percent in insulin treated, and 26 percent in controls. On each of the six poles of the mood scale (composed-anxious, agreeable-hostile, elated-depressed, confident-unsure, energetic-tired, clearheaded-confused), the overall mean values did not differ between the diet-treated or insulin-treated groups compared to the controls. When the two GDM treatment groups were stratified by good versus poor glycemic control, those with better glycemic control had significantly better moods (p<0.05) on four of the six axes tested. These results suggest that treatment does not harm psychological well-being, and that improving glycemic control might be associated with improved well-being.