Abstract
After many years of uncertainty regarding the true pathological nature of mild gestational diabetes and the possible benefits of treatment, the situation appears to have been resolved by the publication of the Australian Carbohydrate Intolerance Study in Pregnant Women (ACHOIS). It is now appropriate for obstetric units to review their treatment and screening programs for gestational diabetes mellitus. Furthermore, with the publication of the Metformin in Gestational Diabetes (MiG) trial, consideration should be given as to whether metformin should be the first choice when diet fails to maintain glycemic control.
Keywords
As recently as 2003 in a Cochrane systematic review, it was possible to conclude as follows, ‘it is uncertain as to whether intensive treatment can influence and reduce perinatal morbidity, i.e., is there any benefit to treating women with gestational diabetes mellitus (GDM) or impaired glucose tolerance (IGT) in pregnancy?’ [1]. In other words, as recently as 2003 there was sufficient uncertainty regarding the pathological status of mild GDM and there was no high-quality evidence of benefit from treatment. In the same year, the UK National Institute for Clinical Excellence (NICE) in its antenatal care guidelines stated that, ‘there is an absence of evidence to support routine screening for GDM and therefore it is not recommended’ [2].
If there was no evidence that abnormal glucose tolerance in pregnancy (i.e., GDM) had any adverse effects on the fetus, or if there was no evidence the treatment was effective, there would be no justification for screening.
Matters have been overtaken by the publication in mid-2005 of the so-called Australian Carbohydrate Intolerance Study in Pregnant Women (ACHOIS), by Crowther and colleagues [3]. This was a randomized, double-blind, controlled trial of treatment versus no treatment for mild IGT in women who met the former WHO criteria [4] for the diagnosis of IGT, in other words, a fasting plasma glucose of below 7.8 mmol/l and a 2 h level, following a 75 g glucose load, of 7.8 mmol/l or above. Excluded from this study were women with frank GDM, that is a 2 h value of 11.1 mmol/l or more. This is a group of women for whom there has not really been any dispute regarding the potential benefit of treatment.
The Australian Carbohydrate Intolerance Study in Pregnant Women randomized 1000 women and in the untreated arm of the study, there were 510 women who were not aware that their glucose tolerance test was in the abnormal range nor were their obstetric attendants.
The headline observation in the study was that a composite measure of serious adverse perinatal complications was present in 23 (4%) of the routine care group but in only seven (1%) of the intervention group who had attended for joint diabetes/antenatal care. This was a highly significant reduction with a relative risk of 0.33. Induction of labor was more common in the intervention group; however, the proportion of large-for-gestational-age babies was halved. There was an equivalent rate of cesarean delivery in the two arms. In a follow-up study of a smaller subgroup observed for any psychological adverse outcomes as a result of being labelled as having GDM, it was found that, in fact, the rate of postnatal depression was halved in the intervention group compared with the routine care group. Therefore, the conclusion of this study was that mild GDM is a disease capable of causing a small but significant rate of serious perinatal morbidity and that treatment is effective. Therefore, if treatment is effective, screening would be justified. Support for this view comes from a US study of similar design published only as an abstract at present [5]. Treatment in this study was associated with reductions in shoulder dystocia, cesarean section and macrosomia.
How should screening be performed
The most recent report from the NICE Antenatal Care Guideline Group suggests screening for risk factors alone, a very traditional approach to screening. One study from the USA using this approach suggested screening on the grounds of obesity, a family history of diabetes, previous pregnancy failure or macrosomia affecting an earlier infant [6]. They calculated that this approach had a sensitivity of 69%, a specificity of 68% and a positive predictive value of 5%. It has also been suggested that membership of a high prevalence ethnic group for GDM should automatically enter someone into a screening program. An alternative approach, particularly where the disease is thought to be more common, might be to use the 50 g glucose challenge test first described by O'Sullivan in 1973 [7]. Here, a 50 g glucose load in water is given to a subject who is unprepared from the dietary point of view between 24 and 28 week' gestation. A positive test would be a plasma glucose 1 h after the glucose load greater than or equal to 7.8 mmol/l and such a result is associated with a sensitivity and specificity of 79 and 87%, respectively, with a positive predictive value of 15% for subsequent abnormal glucose tolerance testing.
A more modified and perhaps logistically sensible approach has been described more recently by Åberg and colleagues in southern Sweden [8]. In this study, a community-based population screen was conducted in which women at approximately 28 week' gestation were advised to fast overnight and then take a 75 g load of glucose at a fixed time. They were then visited 2 h later by a community midwife who took a blood glucose sample that was then interpreted by the WHO criteria (see next section).
Further research is needed into appropriate screening for different populations; however, as a rough guide, the expected prevalence of the disorder in a whole hospital population or in ethnic subgroups within the hospital population might be used. In my opinion, the risk factor screening method is probably suitable for populations with a prevalence of GDM less than 2.5%; two stage screening with a glucose challenge test and a glucose tolerance test is suitable for those with an expected prevalence between 2.5 and 4.5%; and either a full or a modified glucose tolerance test can be applied to the whole population where the prevalence is expected to exceed 4.5%.
For those confronted by the difficult logistics of organizing these screening programs the American Diabetes Association [9] has suggested some low-risk criteria that it was felt could be used to exclude safely women from the screening program. These criteria are an age less than 25 years, normal prepregnancy bodyweight, with no history of abnormal glucose metabolism, no family history of diabetes or poor obstetric outcome and belonging to a low prevalence ethnic group. In the UK, these criteria would exclude approximately 25% of women from the screening program and the total proportion of cases of GDM that are likely to be missed as a result of implementing these criteria has been calculated as being approximately 3%.
Diagnostic criteria for GDM
There has been a transatlantic split in the preferred diagnostic test for GDM. In the USA, it is traditionally favored to use the 100 g glucose tolerance with 3 h criteria of abnormality or venous plasma glucose, set at fasting greater than or equal to 5.3 mmol/l; 1 h greater than or equal to 10.0 mmol/l; 2 h greater than or equal to 8.6 mmol/l; 3 h greater than or equal to 7.8 mmol/l; of which two or more had to be abnormal for the diagnosis [10]. In Europe, most centers have followed the WHO criteria using the 75 g glucose load and the 2 h sample measuring only the fasting and the 2 h levels and diagnosing diabetes with a fasting glucose above 7.0 and/or a 2 h level above 11.0 mmol/l. The group with a normal fasting level and a 2 h level between 7.8 and 11.0 mmol/l were labelled IGT. However, more recently, it has been suggested that the 2-h cutoff for GDM should be reduced to 7.8 mmol/l, thus abandoning the diagnostic configuration IGT althogether [4].
Into this unsatisfactory state has been placed further confusion by the publication of the Hyperglycaemia and Adverse Pregnancy Outcome (HAPO) trial in mid-2008 [11]. In this study, 23,000 women recruited internationally at 24–32 week' gestation had a 75 g glucose tolerance test. The results were blinded unless the fasting plasma glucose exceeded 5.8 mmol/l or the 2 h level exceeded 11.1 mmol/l or if the woman developed diabetic symptoms and was found to have a random plasma glucose above 8.9 mmol/l. The four primary outcomes of the study were primary cesarean delivery, fetal macrosomia, neonatal hypoglycemia and fetal hyperinsulinism detected by levels of cord blood c-peptide. The majority of women were in the non-GDM range but the results were expressed as septiles of the glucose ranges at fasting, 1 h and 2 h. In three of the categories there was a progressive increase in frequency of the outcome in the fifth, sixth and seventh septiles. This was most obvious in relation to birthweight above the 90th percentile and cord serum c-peptide above the 90th percentile. A different phenomenon was observed with primary cesarean section in which there was an increase in the rate of cesarean section by septiles throughout the range and in the diagnosis of clinical neonatal hypoglycemia there was less variation with advancing glucose septiles. Interestingly, each of the primary outcomes was increased independently when each of the three blood glucose categories; fasting, 1 h and 2 h, was considered separately.
At the time of writing, the HAPO study research group are considering whether the recommended cutoff values for the diagnosis of GDM should be revised in the light of their findings.
Fetal hyperinsulinism in both pre-existing & GDM
Fetal hyperinsulinism is the trigger to macrosomia and, is in most cases, induced by excess transplacental transfer of glucose and other nutrient substrates from the mother with diabetes. Taken across the board, birthweights in infants of mothers with diabetes are approximately 20% higher than in nondiabetic women of similar stature and BMI and therefore measures of fetal insulinization will reflect the extent of metabolic perturbation in a particular pregnancy. If measures are made at the end of pregnancy, such as cord blood insulin or c-peptide, or as in our own laboratory, measurements of amniotic fluid insulin (
Amniotic fluid insulin levels (mU/l) at delivery in women with Type 1 or 2 diabetes and gestational diabetes mellitus, compared with a nondiabetic group.
Treatment of GDM
It has been traditional in the USA to approach treatment through a combination of diet therapy and both short- and long-acting insulin regimes administered more or less universally. In a randomized study addressed to the necessity for insulin treatment for all of these women, Persson and colleagues in Sweden demonstrated that there were no significant differences in birthweight, number of large newborns, skin-fold thicknesses or typical neonatal morbidity in the infants of a group of women randomized to diet compared with a group of women treated with diet and insulin [14]. Approximately one in six of the women in the diet only group developed nonresponsive hyperglycemia at some point in the pregnancy and required supplementary insulin. The importance of this study was that it demonstrated that outcomes would be similar with only a minority of women requiring insulin treatment. Dietary approaches have varied but the most important studies have been completed in France by Romon and colleagues in which they showed that a high carbohydrate, low glycemic index (GI) diet was associated with a reduction in the mean birthweight of babies born to mothers with GDM [15]. They found that none of the babies who were large-for-gestational-age were born to women in the upper two quintiles of average carbohydrate intake during the pregnancy. This was achieved without any significant increase in the rates of small-for-gestational-age babies. Therefore, it is recommended that, as a starting point in diet therapy, the traditional Diabetes UK criteria are followed with 45–60% of energy coming from carbohydrate sources, favoring those with a low GI [16].
A newly published randomized study of low GI diet in pregnancy versus conventional high-fiber diet reduced the number of women requiring insulin from 59 to 29% (p = 0.023) with no significant differences in obstetric and fetal outcomes [17].
There has been considerable debate on the subject of insulin therapy and at least one important randomized, controlled trial [18] comparing preprandial glucose monitoring with postprandial glucose monitoring. Women with GDM requiring insulin were randomized to preprandial monitoring, aiming for a blood glucose of 3.3–5.9 mmol/l, or postprandial monitoring, aiming for a 1 h postprandial blood glucose of less than 7.8 mmol/l. The average insulin dose was 25 units per day higher in the postprandial monitoring group than the preprandial monitoring group and late pregnancy HbA1C was 8.1% in the pre-prandial group but 6.5% in the postprandial group, a highly significant difference. There was no excess of pre-eclampsia and the gestational age at delivery was the same in both groups. The cesarean section rate and the third degree tear rate were both lower in the postprandial monitoring group although these trends were not significant. This postprandial regime was associated with a highly significant reduction in the birthweight of approximately 400 g and a reduction in the proportion of large-for-gestational-age babies from 42 to 12%.
Two US studies have suggested that refinement of insulin treatment might also be modified by the selection of women for insulin treatment not based on glycemic targets alone, but glycemic targets and ultrasound measurements of fetal abdominal circumference as an indicator of incipient macrosomia. Again, both studies demonstrated that under some circumstances insulin treatment could be avoided [15,16].
Oral hypoglycemic drugs in the management of GDM
It has been traditional to advise against the use of oral hypoglycemic drugs both in pre-existing diabetes and GDM complicating pregnancy; however, the picture has changed as a result of important recent publications. Whilst the older generation of sulphonylureas crossed the placenta and were capable of causing significant neonatal hypoglycemia, more rapidly acting oral hypoglycemics, such as glibenclamide, have been shown to be safe and effective in pregnancy in at least one large scale randomized, controlled trial [21]. In this study, 404 women in the USA who required hypoglycemic therapy after a trial of diet were randomized to either glyburide (known as glibenclamide in some countries) or insulin. A total of 4% of women in the glyburide group required supplementary insulin; however, the remainder were well controlled on glyburide alone and comparisons demonstrated that third trimester HbA1C levels were identical, as were the proportion of large-for-gestational-age babies: 12% on glyburide versus 13% on insulin. Mean cord serum insulin levels at 15 μU/ml were identical in both groups. There were no significant differences in neonatal morbidity.
More recently, in the summer of 2008 the Metformin in Gestational Diabetes (MiG) study was published [22]. This important and long awaited study aimed to test the hypothesis that metformin treatment compared with insulin in GDM will result in similar perinatal outcomes, improve the insulin sensitivity in the mother and the baby and be associated with improved treatment acceptability. A total of 750 women with GDM requiring hypoglycemic therapy based on the finding of a persistent fasting capillary glucose more than 5.4 mmol/l or a 2 h postprandial glucose more than 6.7 mmol/l, were randomized to insulin or metformin. Metformin dose was increased in 500 mg increments up to a maximum of 2500 mg/day. If control was not obtained at this level, supplementary insulin was added and the metformin was continued. In the 363 women of the metformin group, 195 went through pregnancy on metformin alone but 168 required supplementary insulin. The outcomes demonstrated that there was a slightly higher rate of neonatal hypoglycemia in the insulin group but no difference in respiratory distress or shoulder dystocia rates or rates of admission to a neonatal unit. Perhaps the most important outcome of this study concerns net weight loss from enrolment to 6 weeks postpartum – this was a mean of −8.1 kg in the metformin group but a mean of only −6.9 in the insulin group. This suggests, as theory would also have suggested, that metformin may be particularly suitable and have some advantages over both insulin and the sulphonylureas in women with obesity complicating their pregnancy as well as GDM.
Timing of delivery in GDM
There has only been one randomized, controlled trial of quality that has addressed the question of timing of delivery in GDM – that of Kjos and colleagues [23]. In this study, 200 subjects were randomized, almost all of whom were at 38 week' gestation with insulin-treated GDM, to an active group who were induced within 5 days or an expectant group who were monitored with twice-weekly cardiotocography and weekly ultrasound measurements of amniotic fluid volume until labor supervened. The expectation was that there would be a higher rate of spontaneous labor in the expectantly managed group and possibly a reduction in cesarean section rates with no increase in morbidity. In fact, although the spontaneous labor rate was 44% in the expectant group and only 22% in the active group the cesarean section rate was unexpectedly slightly lower in the induction group 25 versus 31% (not significant). The number of births higher than the 90th percentile was only 10% in the active group but doubled at 23% in the expectant group and all the shoulder dystocia in the series occurred in the expectant group (3%).
The study was not powered to look at perinatal mortality as an end point but of course late pregnancy stillbirth is seen in cases of GDM, as well as pre-existing diabetes, and it can be assumed that a policy of active management by induction of elective delivery before the completion of 38 week' gestation will result in an overall reduction in stillbirth rates.
Conclusion
It is important to search for refinements in both diagnosis and treatment of GDM. In many parts of the world, such as the Middle East, abnormal carbohydrate tolerance can complicate as many as 40% of pregnancies and there is an upward trend worldwide in parallel with the rapid increase in rates of Type 2 diabetes.
It is now clear that treatment of GDM reduces adverse perinatal outcomes even in the more mild categories. Treatment options that have traditionally been limited to diet and insulin must now be reviewed in the light of clear equivalence from short-acting sulphonylureas, which may be a lot more acceptable to the woman with GDM than self-injected insulin. The apparent safety of the use of metformin in late pregnancy demonstrated by the recently published MiG trial suggests that although a relatively large proportion of these women will require supplementary insulin they are likely to benefit in terms of reduced maternal net fat gain at the end of pregnancy. Obviously, this is an important consideration since many of the women identified as having GDM are at risk of developing Type 2 diabetes later in their lives unless they are able to modify both lifestyle and bodyweight following a pregnancy in which GDM was diagnosed.
Executive summary
Gestational diabetes mellitus (GDM) is associated with an increase in adverse perinatal outcomes.
Treatment reduces the number of adverse perinatal outcomes by two-thirds.
In low prevalence populations, screening by risk factor may be logistically acceptable.
In higher prevalence populations provocative testing (either two stage with a glucose challenge test and the oral glucose tolerance test or one stage with the oral glucose tolerance test) is likely to be appropriate.
There is continuing evidence that, despite treatment, excess glucose transfer across the placenta stimulates fetal hyperinsulism in approximately a third of cases.
Enhanced maternal glycemic control may reduce rates of fetal hyperinsulinism and the associated newborn complications.
All women with GDM should be treated in the first instance with a high carbohydrate low gycemic index diet.
Insulin retains an important place but should not be administered universally.
Self monitoring with a 1 h postprandial glucose target of 7.8 mmol/l is associated with improved outcomes of pregnancy.
Short acting sulphonylureas such as glibenclamide have been proven to be safe and effective alternatives to insulin.
The biguanide metformin is also likely to have an important role in management, particularly where GDM is associated with maternal obesity.
On the basis of currently available evidence, delivery at 38 week' gestation by induction of labor is recommended.
Further randomized trials are required on a country by country basis to apply optimal approaches to GDM screening.
The role of short-acting sulphonylureas and metformin in overall management of GDM needs to be established.
Delivery after 38 week' gestation may be possible if trials in progress help to identify the fetus at risk of shoulder dystocia/intrauterine fetal death.
Footnotes
The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. No writing assistance was utilized in the production of this manuscript.