METFORMIN AND GESTATIONAL DIABETES
Charles J. Glueck MD, Naila Goldenberg MD, Patricia Streicher RD, Ping Wang PhD
From the Cholesterol Center, Jewish Hospital of Cincinnati
Current Diabetes Reports 2003;3:303-312
ABSTRACT
Pregnancy increases requirements for insulin secretion, increasing insulin resistance and demands on pancreatic β-cells,
promoting development of gestational diabetes (GD), particularly in women with pre-existing insulin resistance, commonly in women with polycystic ovary
syndrome (PCOS). Preliminary studies suggest that metformin may have unique potential to prevent development of GD. We postulate that interventions which
reduce insulin resistance and lower requirements for endogenous insulin secretion can preserve beta cell function and prevent development of type 2 DM.
INTRODUCTION:
Pregnancy increases requirements for insulin secretion while increasing insulin resistance, increasing demands on
pancreatic β-cells, promoting development of gestational diabetes (GD), particularly in women with pre-existing insulin resistance, commonly in women
with polycystic ovary syndrome (PCOS). If there is impaired pancreatic β-cell compensation for insulin resistance during pregnancy, then GD develops.
GD appears to have maternal and neonatal ramifications for pregnancy outcomes and for the later development of type 2 diabetes mellitus (DM). Of women
with GD, 14% to 60% will develop type 2 DM later in life, and 30-50% will have GD with consecutive pregnancies. Currently, there is debate about the
existence of GD, how, when, and how broadly it should be diagnosed, how aggressively it should be treated, and the extent of adverse pregnancy and
perinatal outcomes, as well as adverse outcomes later in maternal and offspring's life. The current consensus is that women with unequivocal GD have a
significant risk of adverse perinatal outcomes, and increased risk of later type 2 DM. Fetuses from pregnancies with GD have a higher risk of macrosomia
(associated with higher rate of birth injuries), asphyxia, neonatal hypoglycemia, and neonatal hyperinsulinemia. Uncontrolled GD predisposes fetuses to
accelerated and excessive fat accumulation, insulin resistance, pancreatic exhaustion secondary to prenatal hyperglycemia, and, possibly higher risk of
child and adult obesity and type 2 DM later in adult life. There is no uniform agreement, however, about the degree of linkage of adverse pregnancy,
maternal, fetal, and perinatal outcomes to maternal glucose intolerance less severe than unequivocal GD, and there is major controversy about any
continuous relationship between maternal glycemia less than currently defined GD levels and adverse perinatal outcomes. If dietary intervention is not
sufficient to provide optimal control of GD, then, historically, insulin has been added. Preliminary studies suggest that metformin may have the unique
potential to prevent the development of GD. We postulate that interventions which reduce insulin resistance and lower requirements for endogeneous insulin
secretion can preserve beta cell function and prevent the development of type 2 DM.
Gestational Diabetes: Etiology, Epidemiology, Diagnosis, Maternal and Fetal Ramifications:
Pregnancy increases requirements for insulin secretion while increasing insulin resistance [1], increasing demands on
pancreatic β-cells [2,3], and in the presence of impaired pancreatic β-cell compensation for insulin resistance [4**], leads to gestational
diabetes or even to diabetes mellitus (DM) which persists post-partum [5], Figure 1. In late pregnancy, when insulin sensitivity decreases by 80%, there
is a need for five-fold greater insulin response to achieve an identical glucose disposition index [4**]. By convention, GD excludes women who had
pre-conception type 2 DM, but does not exclude women found to have GD who, soon after the post-partum period, have glucose intolerance, suggesting that
their GD was a manifestation of previously undiagnosed type 2 DM.
Recognized risk factors for GD [3,6-8] include body mass index >25 kg/m 2,
first degree family history of type 2 DM, age ³ 25 years, multiparity, previous GD, and previous macrosomic infants (>9 lbs. or 4000 grams).
Additional risk factors for GD [3,6-8] include pre-conception impaired fasting glucose levels (110-125 mg/dL), pre-conception impaired glucose tolerance
(2 hour post oral glucose load glucose levels 140-199 mg/dL), polycystic ovary syndrome (PCOS), and ethnic group (American Indian or Alaska Native;
African American; Asian; Hispanic; Pacific Islander).
By current diagnostic criteria, GD is common in the US, detected in 3.52% of pregnancies (women ages 15-49) in a
cross-sectional sample of live births [8], and, in another study, in 14% of the pregnant population [7].
One of the most important features of GD lies in the progression to type 2 DM; 14 to 60% of women with GD will develop type
2 DM later in life [7-9] and 30-50% will have GD with consecutive pregnancies [9*]. Depending on women's first trimester weight [10], the risk of GD
recurring in subsequent pregnancies may range from as high as 60% to 90%.
Recommendations for universal or selective screening for GD [11-14] are based on the premise that identification and
treatment of GD will avert some adverse outcomes including pregnancy induced hypertension, macrosomia (associated with fetal birth trauma), perinatal
mortality, and neonatal metabolic disorders. However, there is ongoing controversy surrounding the diagnosis of GD, with many authors [15-20]
raising the question that there is insufficient evidence to suggest that identification and treatment of gestational diabetes will result in better
maternal and neonatal outcomes. At one extreme of this controversy, for example, are Hunter and Keirse [19]
who concluded that ".except for research purposes all forms of glucose tolerance testing should be stopped."
It is beyond the limits of this review to examine the strength of the scientific evidence relative to the value of
diagnosing and treating GD in regards to maternal and neonatal outcomes. However, as an example of the controversy a recent study has emphasized the very
limited independent contribution of fasting blood glucose or glucose after a 50 gram load to birth weight. Ouzilleau et al [20ww]
retrospectively assessed 300 consecutive high risk women whose plasma glucose after a 50 g glucose load was 144 mg/dL or more compared to a randomly
selected group of 300 control with plasma glucose < 144 mg/dL after a 50 g glucose load. If glucose intolerance was found, no systematic treatment for
glucose intolerance was given. Plasma glucose levels (either fasting or after glucose load) were independently but poorly correlated with birth weight; no
more than 3% to 5% of birth weight variability could be explained by changes in glucose tolerance [20**].
Pregnancy weight, weight gain during pregnancy, and parity had a significant positive association with birth weight. Smoking had a significant negative
correlation with birth weight. Each of these variables had a closer association with birth weight than either fasting or post-glucose load glucose levels
[20**].
The controversy about GD will end only after a robust, randomized, double-blind trial is conducted to demonstrate
whether identification and management of GD at differing levels of therapeutic intensity is associated with significant improvement in neonatal or
maternal outcomes. Many of the contentious issues regarding diagnosis and therapy of GD may, hopefully, be resolved by the HAPO (Hyperglycemia and Adverse
Pregnancy Outcome Study) [21**]. As recently noted by the HAPO study group [21**]
and by the American College of Obstetrics and Gynecology Practice Bulletin [12**] ".a lack of data from well designed studies has contributed to
the controversies surrounding the diagnosis and management of this condition". The American College of Obstetrics and Gynecology Practice Bulletin
[12**] has summarized several controversies about current screening practices and treatment benefits for GD, as follows:
1. "Although screening for GD has become almost universal, there is as yet no uniform data to demonstrate a benefit
to the population as a whole."
2. "If the general population is going to benefit from the diagnosis of GD, there should be an effective treatment
for the condition and '.there is little information regarding the effectiveness of treatment versus no treatment'."
3. "The goal of treatment is to lower the likelihood of macrosomia and its consequences; neonatal hypoglycemia may
also be reduced. Although the quality of the information varies, evidence is available to confirm these benefits. However, there has been no
demonstrated treatment benefit on long term outcomes for the offspring such as obesity and the development of diabetes."
The HAPO study [21**] was designed to ".to clarify unanswered questions on associations of maternal glycemia, less
severe than overt diabetes, with risks of adverse pregnancy outcome." This five-year, prospective, observational study will assess pregnancies in
25,000 ethnically diverse pregnant women in 10 countries [21**]. Caregivers will be blinded to the status of glucose tolerance in the third trimester of
gestation (75 gram 2 hour oral glucose tolerance test at 24 and 32 weeks gestation). The results will be unblinded to women and caregivers if fasting
plasma glucose is > 104 mg/dL, 2 hour plasma glucose is > 200 mg/dl, or if any plasma glucose is < than 45 mg/dl. Random plasma glucose
measurement will be performed at 34 and 37 weeks, or if symptoms suggest hyperglycemia; these results will be unblinded for values 
160 mg/dl. Maternal blood will be obtained for measurement of serum C peptide and hemoglobin A1C; cord blood will be obtained for serum C peptide and
plasma glucose. One to two hours after delivery, a capillary specimen will be obtained for measurement of neonatal plasma glucose. Neonatal
anthropometrics will be obtained, and follow up data will be collected at 4 to 6 weeks after delivery. Maternal glycemia will be related to the following
primary outcomes: cesarean delivery, increased fetal size, neonatal morbidity, and fetal hyperinsulinism. The HAPO investigators [21**] noted "from
an obstetrical perspective, the significance of GDM is related to the frequency and severity of risk for adverse pregnancy outcome, rather than the risk
for future DM in the mother." Moreover, although the diagnosis of GD identifies women at high risk for future DM, there is controversy about the
long-term risk of the fetus developing DM in adulthood, and controversy about whether and to what degree GD is associated with preeclampsia, neonatal
macrosomia, and cesarean section [21**].
There is general consensus that women with unequivocal GD by current diagnostic criteria [1,2,7-12] have an increased
risk of adverse perinatal outcomes, and increased risk of later type 2 DM. Fetuses from pregnancies with GD have a higher risk of macrosomia (associated
with higher rate of birth injuries), asphyxia, neonatal hypoglycemia, and neonatal hyperinsulinemia [11,12]. In a study of 9,471 Chinese women, those with
GD had an increased risk for premature rupture of membranes, preterm birth, breech presentation, and top 10th
percentile birth weight after adjusting for confounding factors [22]. Lao et al [23] have reported that despite dietary treatment, maternal GD is
associated with increased perinatal morbidity independent of its effect on fetal size. Neonatal malformations are related to GD, in a continuum from
normal to pre-conception type 2 DM. In a study of 145,196 women, infant malformations occurred in 1.5% of non-diabetic women, 1.2% of women with normal
fasting glucose and GD, 4.8% in women with GD and fasting hyperglycemia, and 6.1% with pregestational DM [24*]. GD is not a homogeneous diagnosis with
regard to risk of major congenital abnormalities [25]. In women who evidenced GD and later were found to have type 2 DM on early postpartum testing, who
most likely had undiagnosed type 2 DM antedating pregnancy, the rate of major congenital abnormalities was the same as for women with established type I
or type 2 DM [25]. Uncontrolled GD predisposes fetuses to accelerated and excessive fat accumulation, insulin resistance, and, possibly, higher risk of
child and adult obesity and adult type 2 DM [26].
Women with GD have higher risk of spontaneous abortions, pre-term delivery, stillbirths, cesarean section, and postpartum
hemorrhage [27,28].
Although unsettled and somewhat controversial, fetuses from pregnancies where GD has occurred may have a higher risk of
congenital defects [27,28,29**,30], and commonly have macrosomia, which is associated with higher rate of birth injuries, asphyxia [21**-31], and neonatal
hypoglycemia.
Polycystic ovary syndrome (PCOS), Gestational diabetes, and type 2 diabetes:
The insulin resistance characteristic of PCOS is directly and causally related to GD when it is augmented by
pregnancy-associated insulin resistance, Figure 1 [2,32**-37]. Related, in part to insulin resistance characteristic of PCOS, the prevalence of impaired
glucose tolerance and type 2 DM in PCOS is much higher than in age- and weight-matched women without PCOS [38]. As many as 54% of women with PCOS [36] may
have impaired glucose tolerance, and the very common obesity of PCOS amplifies insulin resistance, further promoting glucose intolerance. Thirty-one
percent of obese women with PCOS have impaired glucose tolerance, 7.5% DM, and, in non-obese women with PCOS, 10.3% have impaired glucose tolerance and
1.5% DM [35**]. Korhonen et al [39] have shown that women with PCOS form a ". distinct subgroup of a much wider problem, metabolic syndrome."
Polycystic ovaries are common in women first identified by GD and/or type 2 DM [33]. Kousta et al [33] reported that 50% of
women with GD had polycystic ovarian morphology vs 27% of controls. In women with type 2 DM, 82% have polycystic ovaries on ultrasound [40*]. Also, 27% of
women with type 2 DM can be shown to have PCOS [41]. Long or highly irregular menstrual cycles, associated with insulin resistance, are a marker for risk
of developing type 2 DM not entirely explained by obesity [42].
Obesity which is concurrent with oligomenorrhea confers much greater risk for DM than oligomenorrhea alone, since obesity
is a major modifiable risk factor for type 2 DM [43] Insulin resistance and associated β-cell dysfunction, common in obesity and in PCOS, appear to
predispose to type 2 DM, and high insulin levels have predicted progression to type 2 DM in high-risk populations [44], Figure 1.
Hyperandrogenemia, a hallmark of PCOS, may have a direct adverse effect on insulin resistance [45]
Before pregnancy, women with PCOS commonly have insulin resistance [2,46**-49] and obesity, major risk factors for GD
[2,3], Figure 1. Because of infertility [48,49], women with PCOS are older than the general population at the time of conception, another risk factor for
GD [2**].
Screening tests and criteria for gestational diabetes:
Widely used screening criteria for GD have been established by the American Diabetes Association [11**], the American
College of Obstetricians and Gynecologists [12**], and the National Diabetes Data Group's conversion of O'Sullivan et al's diagnostic criteria for the 100
g glucose challenge [50].
Currently, it is suggested [11,12] that all women should be tested between 24 and 28 weeks of gestation with a 50 g glucose
challenge test. A one-hour blood glucose >140 mg/dL is considered positive for GD and >120 mg/dL is suspicious. The definitive 100 g oral glucose
tolerance test should be carried out after 8-10 hours overnight fast. The diagnosis of GD requires 2 abnormal values (fasting blood glucose >105 mg/dL,
1 hour >190 mg/dL, 2 hour >165 mg/dL, 3 hour >145 mg/dL); if one value is abnormal, then self-monitored blood glucose should be recorded for 7
days. If self-monitored average fasting blood glucose is >95 mg/dL or the 2-hour post-meal average is >120 mg/dL during this 7-day monitoring
period, re-evaluation for GD is warranted. It has been suggested that women at high risk should be screened as soon as pregnancy is confirmed [11**].
Glucose targets for the treatment of Gestational Diabetes:
When considering glucose targets for the treatment of GD, we agree with the caveats of Buchanan and Kjos [17**], as
follows: "Only a minority of women with GDM are at risk for a perinatal complication. More importantly, perinatal risks increase very slowly and
quite continuously with increasing maternal glucose levels, regardless of the timing of glucose measurements. There is no threshold of glucose below which
risks are low and above which risks increase rapidly. Clinical glycemic targets can be no more than arbitrary cut-points across a shallow continuum of
risk to the fetus. Elimination of all excess risk using glucose levels alone requires treatment of many individuals at no risk whatsoever."
Current recommendations [11,12] include the following targets for treatment:
Self-Monitored Blood Glucose:
All values within target range, as follows: pre-meal and bedtime, <105 mg/dL, 1-hour post-prandial <140 mg/dL, 2
hours post-prandial < 120 mg/dl.
Hemoglobin A1c(HbA1c):
May be used to evaluate prior hyperglycemia, but is not used in GD management; should be within normal range (<6.5%,
differs by laboratories).
Urine Ketones (Fasting):
Should be negative.
Diet, Insulin, Sulfonylurea:
Because we have recently reviewed use of diet, insulins, and sulfonylureas in GD [51], they will not be covered in this
review.
Metformin:
Unless metformin can be shown to be safe, effective, and non-teratogenic during pregnancy, there can be no serious
consideration of its use in prevention or treatment of GD. On the basis of non-randomized, non-blinded studies, to date, metformin, a pregnancy category "B" class drug, appears to be safe, effective, and non-teratogenic during pregnancy [2,48,49,51-62]. No randomized, double blind,
placebo-controlled clinical trial of metformin and GD, pregnancy loss, pre-eclampsia, and potential teratogenicity has been reported to date. In studies
that compare women with prior pregnancy loss without metformin and then infer treatment benefit in subsequent pregnancies on metformin, there is an
unknown "background" success rate, which cannot be quantitated [2,48,49,51,53,55,56]. The phenomenon of "regression towards the mean" [63] also needs to be considered in comparing paired pregnancy outcomes in women initially without metformin and subsequently on metformin
[2,48,49,51,53,55,56]. Assuming adverse pregnancy outcomes without metformin, the subsequent pregnancy on metformin might potentially be more successful,
with its outcome nearer to the mean [63] of normal pregnancy [2,48,49,53,55,56].
Metformin during pregnancy in women with PCOS has been shown to safely reduce first trimester SAB from 73% to 10%, p<.
0001 [48**], and from 62% to 26% (p<.0001) [49**], is not apparently teratogenic [2,48,49,51-62], and does not adversely affect infants' birth weight
or length [2,48, 49,53,55,56], or growth and motor and social development in the first year of life [48,49]. Moreover, in whole embryo culture, metformin
produces no alterations in embryonic growth and no major malformations [64]. A partial placental barrier to metformin transfer is said to exist in animal
models (Physician's Desk Reference,2003), but no scientific citations are available for review.
The first pilot study of the safety and efficacy of metformin during pregnancy appeared in 2001 [48**]. In an open label
study, Glueck et al [48**] gave metformin (1.5-2.55 g/day) throughout pregnancy to 19 previously oligoamenorrheic, non-diabetic women with PCOS.
Previously, without metformin, 10 of these 19 women had 22 pregnancies with 16 first trimester SABs (73%) which was reduced to 10% (current pregnancies on
metformin), p< .0001. There were no birth defects and no maternal or neonatal hypoglycemia. On metformin, reductions in fasting serum insulin and
hypofibrinolytic plasminogen activator inhibitor activity (PAI-Fx), an independent determinant for spontaneous abortion [48**], were correlated.
In a recent (2002) open label study, Glueck et al [49**] assessed 72 oligoamenorrheic women with PCOS who conceived
on metformin (2.55 g/day) and maintained it throughout their pregnancies. Of the 84 fetuses, to date [49**], there have been 63 normal live births without
congenital defects (75%), 14 first trimester spontaneous abortions (17%), and 7 ongoing pregnancies > 13 weeks with normal sonograms without
congenital defects (8%). Without metformin, 40 of the 72 women had 100 previous pregnancies (100 fetuses) with 34 (34%) live births and 62 (62%) 1st
trimester SABs. In current pregnancies on metformin in these 40 women (46 pregnancies, 47 fetuses), there have
been 33 live births (70%), 2 pregnancies ongoing > 13 weeks (4%), and 12 SABs (26%), p <. 0001 [49**]. There was no maternal lactic acidosis,
and no maternal or neonatal hypoglycemia. At 6 month follow-up, height was greater (p =. 008) and weight did not differ (p =. 26) from normal pediatric
populations; motor and social development were normal [49**].
Metformin during pregnancy in non-diabetic women with polycystic ovary syndrome is not associated with increased
pre-eclampsia [52]. Among 574 women referred for a study of efficacy and safety of metformin in PCOS, 110 women conceived on metformin (1.5-2.55 g/day)
[52]. Pre-eclampsia, gestational diabetes, and other pregnancy outcomes were compared in those 86 women with ³ 1 live birth (91 pregnancies, 94 live
births) versus 252 healthy non-PCOS women with ³1 live birth consecutively delivered in a community obstetrics practice [52]. Pre-eclampsia in PCOS (5/91
pregnancies, 5.5%), did not differ (p = 0.4) from controls (9/252, 3.6%), nor did it differ (p =1.0) in PCOS vs control primigravidas (2/43 [4.7%] vs 4/91
[4.4%]) [52]. By stepwise logistic regression, group (PCOS, controls) was not a significant correlate of development of pre-eclampsia, p> 0.15 [52].
Jakubowicz et al [54**] retrospectively studied 65 women with PCOS who received metformin during pregnancy and 31
who did not. The early pregnancy loss rate in the metformin group was 6 of 68 pregnancies (8.8%) compared with 13 of 31 (41.9%) in the untreated control
group (p<.001) [54**]. In a subset of women with prior history of miscarriage, the early pregnancy loss rate was 4 of 36 pregnancies (11.1%) in the
metformin group vs 7 of 12 (58.3%) in the controls (p=. 002) [54**].
In 49 anovulatory patients with PCOS, Heard et al [62**] gave metformin 500 mg twice per day, increased to 500 mg three
times per day if no ovulation occurred, and then added Clomiphene citrate (50 mg/day) if no ovulatory response occurred after 6 weeks. Nineteen women
(40%) resumed spontaneous menses with metformin alone; 15 (31%) required Clomiphene in conjunction with metformin and 10 of these (67%) had evidence for
ovulation. Of 48 women, 20 (42%) conceived, and, with the metformin being stopped at 12 weeks gestation, 7 of the 20 (35%) had spontaneous abortions.
Heard et al [62**] concluded that ".metformin alone in patients with PCOS results in a substantial number of pregnancies, with 69% (20/29) of those
who ovulated conceiving in less than 6 months."
Although promising [2,48,49,52-56] the safety and efficacy of metformin in the reduction of SAB in women with PCOS, along
with its apparent lack of teratogenicity must be confirmed by randomized, double-blind, controlled clinical trials.
Glucophage (metformin) has been used in Europe for over 30 years for the treatment of type 2 DM, and has been available in
the United States since 1995. How metformin increases insulin action is not completely understood. Post-receptor effects include suppression of hepatic
glucose output, increased insulin-mediated glucose utilization in peripheral tissues (such as muscle and liver), particularly after meals, and an
antilipolytic effect that lowers serum free fatty acid concentrations, thereby reducing substrate availability for gluconeogenesis.
Metformin and gestational diabetes:
Three new prospective, case-control studies on the safety and efficacy of metformin in the prevention of GD have recently
been completed [2, 49, 52]. Although promising [2,49,52], the safety and efficacy of metformin in the prevention of GD must be confirmed by randomized,
double-blind, controlled clinical trials. We postulate that interventions which reduce insulin resistance and lower requirements for endogeneous insulin
secretion (metformin, troglitazone, pioglitazone, rosiglitazone] can preserve beta cell function and prevent the development of type 2 diabetes.
Glueck et al assessed whether metformin safely reduced development of GD in women with polycystic ovary syndrome (PCOS)
[2**]. A prospective study was carried out in an outpatient clinical research center of 33 non-diabetic women with PCOS who conceived on metformin and had
live births, 28 on metformin through delivery. At the same time, a retrospective assessment was made of 39 non-diabetic women with PCOS who had live birth
pregnancies without metformin. In the prospective cohort, all women were instructed in a 1500-calorie, high protein (26% of calories), low carbohydrate
(44%) diet, with 30% of calories as fat. After conception, calorie restrictions were dropped, but the caloric distribution was maintained. Metformin, 2.55
g/day, was given throughout pregnancy. The main outcome measure was development of GD. Pre-metformin, the cohorts of 33 and 39 women had high
fasting insulin, were insulin resistant (IR), and had high insulin secretion. In the 33 women who received metformin, GD developed in 1 of 33 (3%)
pregnancies, vs 8 of 12 (67%) of their previous pregnancies without metformin (p<. 0001), vs 14 of 60 (23%) pregnancies in the 39 women without
metformin (p=. 016). Combining all live births without metformin, GD occurred in 22/72 pregnancies (31%) vs 1/33 pregnancies (3%) on metformin, p =. 0009.
With GD in 93 pregnancies as the response variable, age at delivery, and treatment group (metformin, no metformin) as explanatory variables, the odds
ratio of GD for pregnancies on metformin vs no metformin was 0.115 with 95% CI 0.014 to 0.938, p =. 04. In PCOS, metformin is associated with a ten-fold
reduction in GD (31% to 3%), reducing IR and insulin secretion, thus lowering secretory demands imposed on pancreatic β-cells by IR and by pregnancy
[2**], Figure 1.
Mean (SD) and median gestation for the 33 pregnancies on metformin was 38.2 (3.3) and 39.7 weeks [2**]. The distribution of
height and weight in the 34 neonates (one twin pair, 32 singleton births) closely approximated a normal distribution [2**].
In 21 of the 33 women who conceived on metformin, complete data was available for weight, BMI, insulin, IR, and insulin
secretion, comparing levels pre-metformin to values for the last pre-conception visit on metformin. In these 21 women, on metformin, median weight fell
from 94 to 88 kg (p<. 0001), BMI from 33.6 to 29.6 (p <. 0001), insulin from 23 to 14 uU/ml (p=. 001), IR from 5.33 to 2.87, (p=. 0005), and insulin
secretion from 274 to 226 (p=. 04) [2**].
In 15 women receiving metformin throughout pregnancy, complete data was available at pre-metformin, pre-pregnancy baseline,
at the last pre-conception visit on metformin, at the first visit at 4-6 weeks gestation on metformin, and the mean of subsequent visits during pregnancy
on metformin. From baseline to the last visit on metformin before conception, median weight fell from 102 to 88 kg (p=. 0001), BMI fell from 33.9 to 30.3
(p=. 0001), insulin fell from 23 to 15 uU/ml (p=. 04), and IR fell from 4.82 to 3.18, p=.03. These effects were then maintained without significant change
(p> 0.05) throughout pregnancy [2**].
Of the 33 women taking metformin, none developed lactic acidosis. Intermittent diarrhea and/or gastritis were common in the
first 3 weeks on metformin but resolved spontaneously and were not limiting factors. There were no major fetal malformations or fetal hypoglycemia in the
34 live births [2**].
In a second, larger study by Glueck et al [49**], 72 women (84 fetuses) with PCOS conceived on metformin 2.55 g/day
and remained on it through pregnancy. In 68 pregnancies on metformin (63 live births, 5 ongoing pregnancies ³ 26 weeks), GD developed in 3 pregnancies
(4%). GD was diagnosed in 9/34 (26%) previous pregnancies not on metformin, p=. 025 [49**].
A third study of the effects of metformin on development of GD has recently been reported by Bornovali et al [52].
Development of gestational diabetes in PCOS did not differ from controls (9/89 pregnancies [10%] vs 36/251 [14%], p = 0.31). Of the 94 live births to 86
women with PCOS, there were no major birth defects. Mean ± SD birth weight of the 76 live births ³ 37 weeks gestation in women with PCOS (3423 ± 495 g)
did not differ from controls' 206 live births ³ 37 weeks (3481± 555 g), p = 0.47, nor did the percentage of ³ 37 week gestation neonates ³ 4000 g
(13.2 % vs 17.5%, p =0.4) or ³4500 g (1.3% vs 2.9%, p = 0.7)[52].
In contrast to the uniformly reported safety and efficacy of metformin during pregnancy in predominantly non-diabetic women
with PCOS [2,48,49,51, 52-56] and in diabetic women as well [57-61], Hellmuth et al [65] reported adverse effects of metformin on pregnancy outcomes in a
retrospective study of women with predominantly pre-conception type 2 DM, not reported to have PCOS. Hellmuth et al [65] studied 50 women treated with
metformin, 68 treated with sulphonylurea, and a reference group of 42 diabetic women treated with insulin during pregnancy. Hellmuth et al [65] reported
an increased prevalence of pre-eclampsia on metformin (32%) vs 7% on sulphonylurea, vs 10% on insulin, and also reported higher perinatal mortality (11.6%
on metformin, 1.3% on sulphonylureas or insulin). This retrospective study [65] included diabetic pregnancies over a 25-year period (1966-1991) and
included very different treatment schedules initiated over a very wide range of gestation ".women included during 1966-1984 were treated with oral
anti-diabetic compounds because their metabolic control was insufficient on dietary treatment alone, in an attempt to avoid insulin therapy. All women
included after 1984 have unintentionally taken oral hypoglycaemic agents at the time of conception and in the first part of pregnancy but were changed to
insulin at the first visit to our department." The retrospectively selected, 42 woman control group [65] was highly variegated and included women
whose initial treatment date varied by as much as 32 weeks! Of the 42 female controls, there were 35 with GD diagnosed in gestation week 26 (mean) (range
5-37 weeks) who were treated with insulin from week 28 (mean) (range 5-37 weeks). Of the 42 female controls, there were 7 with pre-gestational
diet-treated type 2 DM diagnosed 1-8 years before pregnancy who received insulin treatment from gestation week 15 (mean) (range 8-25 weeks). The ability
to generalize from the data in this retrospective, 25-year data collection study [65] is very limited because of major differences in treatment type and
onset during pregnancy in both the patient and control groups. Moreover, the study by Hellmuth et al [65] focused on women with predominantly
pre-conception type 2 DM. In contrast, our prospective studies in non-diabetic women with PCOS [2,48,49,51,52,55,56], the retrospective study by
Jakubowicz et al [54**], the prospective study by Heard et al [62**], and reports by Coetzee and Jackson [57-60], and Jackson and Coetzee [61] in diabetic
women and non-diabetics with GD point to metformin's safety and pregnancy-enhancing efficacy. In the study by Bornovali et al (52), pre-eclampsia in PCOS
(5/91 pregnancies, 5.5%), did not differ (p = 0.4) from controls (9/252, 3.6%), nor did it differ (p =1.0) in PCOS vs control primigravidas (2/43 [4.7%]
vs 4/91 [4.4%]). By stepwise logistic regression, group (PCOS, controls) was not a significant correlate of development of pre-eclampsia, p>
0.15. In the 86 age-race-weight-matched PCOS-control pairs, the incidence of pre-eclampsia was the same, 5.8%, 5 of 86 women in both groups. Unlike
Hellmuth et al [65], there were no cases of perinatal mortality in any of the cohort of women with PCOS studied by Bornovali et al [52]. Moreover, Coetzee
and Jackson [60] studied 60 pregnant women with type 2 and gestational DM who were treated with metformin in the second and third trimester after dietary
treatment had failed. Of these 60 patients, 27 did not have adequate control on metformin, and were transferred to other therapy. The remaining 33
patients received metformin up to delivery. The perinatal mortality rate in these 33 patients (61/1000) compared favorably with 103/1000 in the metformin
failure group who had subsequently received insulin, an 105/1000 in a group of insulin-dependent diabetics treated during the same period. Coetzee and
Jackson concluded that infant morbidity in the metformin group was low [60].
Primary Prevention of type 2 diabetes mellitus
Glueck et al [2,49] have speculated that metformin, like troglitazone [5], may protect pancreatic β-cells from failure
(Figure 1), thus lowering the later risk of developing type 2 DM.
If subsequent, placebo-controlled, blinded studies confirm preliminary evidence that metformin and diet during
pregnancy reduce development of GD ³ four - fold (2,49,52), then, given the fact that 14% to 60% of women with GD subsequently develop type 2 DM [7-9],
metformin during pregnancy can be considered as a speculative approach to primary prevention of type 2 DM (Figure 1). Within this frame of reference,
troglitazone has been shown to preserve beta cell function and reduce development of type 2 diabetes in 133 high risk Hispanic women with previous GD
[66**]. During a median follow-up of 30 months of a placebo-blinded protocol, average annual DM incidence rates in the 236 women who returned for at least
1 follow-up visit were 12.1% on placebo and 5.4% on troglitazone (p<.01). The placebo group had increased glucose levels and reduced beta cell
compensation for increased insulin resistance. However, the troglitazone group had stable glucose levels, and adequate beta cell compensation for stable
insulin resistance [66**]. Buchanan et al [66**] concluded that women with the most marked pre-treatment insulin resistance and hyperinsulinemia had the
greatest beta cell rest when insulin resistance was improved by troglitazone. They suggested [66**] that troglitazone ".had fundamentally altered
the underlying metabolic changes that lead to diabetes." In agreement with Buchanan et al [66**], we postulate that interventions which reduce
insulin resistance and lower requirements for endogeneous insulin secretion [metformin, troglitazone, pioglitazone, rosiglitazone] can preserve beta cell
function and help prevent the development of type 2 diabetes.
The Diabetes Prevention Program [67**] has recently reported that lifestyle changes and metformin both reduced the
incidence of type 2 DM in high risk subjects, with the lifestyle intervention being more effective than metformin. In the Diabetes Prevention Program,
3234 non-diabetic subjects with elevated fasting and post-load glucose levels were randomized to placebo, metformin (1700 mg/day), or an intensive
weight-loss-exercise program, with average follow-up of 2.8 years [67**]. The lifestyle intervention program incorporated a 16-lesson course directed at ³ 7% reduction in body weight, taught by case managers on a one-to-one basis for 24 weeks after enrollment [67**]. There were also subsequent monthly
individual and group sessions to reinforce lifestyle changes [67**]. The lifestyle and metformin interventions reduced the DM incidence by 58% and 31%,
respectively as compared to placebo. The authors concluded ".to prevent one case of diabetes during a period of three years, 6.9 persons would have
to participate in the lifestyle-intervention program, and 13.9 would have to receive metformin." [67**]
Finnish investigators [68**] have reported that a combination of diet and exercise reduced progression to type 2 DM by 58%
in subjects with impaired glucose tolerance, at high risk for progression to frank type 2 DM.
Freemark and Bursey [69] recently conducted a double-blind, placebo-controlled study of the effects of metformin (1 g/day)
on BMI, serum leptin, glucose tolerance, and serum lipids in obese adolescents with fasting hyperinsulinemia and a family history of type 2 DM. Metformin
facilitated reduction in fasting blood glucose and insulin concentrations and moderated weight gain. The authors concluded ". metformin might
complement the effects of dietary and exercise counseling and reduce the risk of type 2 diabetes in selected patients." [69] Glueck et al's studies
in adolescents with polycystic ovary syndrome using metformin and diet [70], like those of Freemark and Bursey, reduced fasting insulin, glucose, and
weight, and should have promise in prevention of later type 2 DM (Figure 1). Kent and Legro [71], reviewing treatment of polycystic ovary syndrome in
adolescents, discuss the importance of diet and metformin in primary prevention of type 2 DM, noting that "short-term trials have shown promising
effects in reducing insulin secretion, improving insulin sensitivity, restoring normal menstrual cycles, and correcting lipid abnormalities." Primary
prevention of type 2 DM in children and adolescents should be important, given the near epidemic status of type 2 DM in children and youth [72]. As
recently reviewed by Kaufman [73], "In 1992, type 2 DM was a rare occurrence in most pediatric centers. By 1994 it represented up to 16% of new cases
in urban areas, and by 1999, the incidence of new type 2 DM diagnoses ranged between 8% and 45%, depending on geographic location." Nestler (74)
emphasized the capability of metformin in primary prevention of DM in women with PCOS, concluding ".PCOS should be regarded as a general health
issue and the use of insulin-sensitizing drugs such as metformin should be considered for prevention of type 2 diabetes." Nestler [74] also
emphasized the use of insulin-sensitizing drugs in women with PCOS ".Women with PCOS are an insulin-resistant group with high prevalence of
impaired glucose tolerance, who are at a markedly increased risk for type 2 diabetes. Therefore, it seems reasonable to presume that the demonstrated
efficacy of insulin-sensitizing drug-such as metformin-to stave off type 2 diabetes should apply to them as well." Legro (71) noted that
"because PCOS is associated with a 40% prevalence of abnormal glucose tolerance, every adolescent patient should be evaluated regularly for glucose
intolerance with a 2-hour oral glucose tolerance test."
Conclusions:
In aggregate, metformin, and (probably) the glitazone class of drugs have real promise in prevention of GD. Interventions
that reduce insulin resistance and lower requirements for endogenous insulin secretion both during pregnancy (with metformin) and in the non-pregnant
state (with metformin and/or the thiazolidinediones, diet, and exercise), preserve B-cell function, and have promise in the primary prevention of type 2
DM.
Figure Legends:
Figure 1: Insulin resistance, insulin resistance of pregnancy, metformin,
gestational diabetes
References
Papers of particular interest, published recently, have been highlighted as:
* Of importance
** Of major importance
E-mail: glueckch@healthall.com
or cglueck@fuse.net
Fax: 513-585-7950
