Beruflich Dokumente
Kultur Dokumente
Summary
Background Maternal overweight and obesity are associated with well recognised pregnancy complications. Antenatal Lancet Diabetes Endocrinol 2018
dietary and lifestyle interventions have a modest effect on gestational weight gain without affecting pregnancy Published Online
outcomes. We aimed to assess the effects on maternal and infant outcomes of antenatal metformin given in addition December 4, 2018
http://dx.doi.org/10.1016/
to dietary and lifestyle advice among overweight and obese pregnant women.
S2213-8587(18)30310-3
See Online/Comment
Methods GRoW was a multicentre, randomised, double-blind, placebo-controlled trial in which pregnant women at http://dx.doi.org/10.1016/
10–20 weeks’ gestation with a BMI of 25 kg/m² or higher were recruited from three public maternity units in Adelaide, S2213-8587(18)30337-1
SA, Australia. Women were randomly assigned (1:1) via a computer-generated schedule to receive either metformin Discipline of Obstetrics &
(to a maximum dose of 2000 mg per day) or matching placebo. Participants, their antenatal care providers, and Gynaecology and Robinson
Research Institute
research staff (including outcome assessors) were masked to treatment allocation. All women received an antenatal
(Prof J M Dodd PhD, J Louise PhD,
dietary and lifestyle intervention. The primary outcome was the proportion of infants with birthweight greater than A R Deussen BHSc,
4000 g. Secondary outcomes included measures of maternal weight gain, maternal diet and physical activity, maternal Prof G Dekker MD,
pregnancy and birth outcomes, maternal quality of life and emotional wellbeing, and infant birth outcomes. Outcomes Prof W Hague MD) and School
of Public Health (J Louise),
were analysed on an intention-to-treat basis (including all randomly assigned women who did not withdraw consent
University of Adelaide,
to use their data, and who did not have a miscarriage or termination of pregnancy before 20 weeks’ gestation, or a Adelaide, SA, Australia;
stillbirth). The trial is registered with the Australian New Zealand Clinical Trials Registry, number Department of Perinatal
ACTRN12612001277831. Medicine (J M Dodd, W Hague)
and Department of Neonatal
Medicine (A J McPhee MD),
Findings Of 524 women who were randomly assigned between May, 28 2013 and April 26, 2016, 514 were included in Women’s and Children’s
outcome analyses (256 in the metformin group and 258 in the placebo group). Median gestational age at trial entry Hospital, North Adelaide, SA,
was 16·29 weeks (IQR 14·43–18·00) and median BMI was 32·32 kg/m² (28·90–37·10); 167 (32%) participants were Australia; Department of
Obstetrics and Gynaecology,
overweight and 347 (68%) were obese. There was no significant difference in the proportion of infants with birthweight
Flinders University, Bedford
greater than 4000 g (40 [16%] with metformin vs 37 [14%] with placebo; adjusted risk ratio [aRR] 0·97, 95% CI Park, SA, Australia
0·65 to 1·47; p=0·899). Women receiving metformin had lower average weekly gestational weight gain (adjusted (R M Grivell PhD); and Lyell
mean difference –0·08 kg, 95% CI –0·14 to –0·02; p=0·007) and were more likely to have gestational weight gain McEwin Hospital, Elizabeth
Vale, SA, Australia (G Dekker)
below recommendations (aRR 1·46, 95% CI 1·10 to 1·94; p=0·008). Total gestational weight gain, pregnancy and
Correspondence to:
birth outcomes, maternal diet and physical activity, and maternal quality of life and emotional wellbeing did not differ
Prof Jodie M Dodd, Discipline of
significantly between groups. Similar numbers of women in both treatment groups (76% [159/208] in the metformin Obstetrics & Gynaecology and
group and 73% [144/196] in the placebo group) reported side-effects including nausea, diarrhoea, and vomiting. Two Robinson Institute, University of
stillbirths (placebo group) and one neonatal death (metformin group) occurred; none of the perinatal deaths were Adelaide, and Women’s &
Children’s Hospital, North
determined to be attributable to participation in the trial.
Adelaide, SA 5006, Australia
jodie.dodd@adelaide.edu.au
Interpretation For pregnant women who are overweight or obese, metformin given in addition to dietary and lifestyle
advice initiated at 10–20 weeks’ gestation does not improve pregnancy and birth outcomes.
Research in context
Evidence before this study gestational weight gain or pre-eclampsia. In the MOP trial,
We searched the Cochrane Pregnancy and Childbirth Group’s investigators reported no difference in neonatal birthweight
Trials Register by contacting the trials search coordinator (April but decreased gestational weight gain and reduced incidence of
13, 2016). Briefly, the Cochrane Pregnancy and Childbirth pre-eclampsia in the group assigned to metformin.
Group’s Trials Register is maintained by the trials search Importantly, women participating in these previous
coordinator and contains trials identified from: monthly randomised trials were not provided with a dietary and lifestyle
searches of the Cochrane Central Register of Controlled Trials intervention. Additionally, these previous trials have only
(CENTRAL); weekly searches of MEDLINE (Ovid); weekly included women who were obese, whereas there might be
searches of Embase (Ovid); monthly searches of CINAHL clinical benefit in providing treatment for women who are
(EBSCO); manual searches of 30 journals and the proceedings of overweight as well as for those who are obese.
major conferences; weekly current awareness alerts for an
Added value of this study
additional 44 journals and monthly BioMed Central email
To our knowledge, this is the first trial to assess the effect of
alerts. Search terms from this comprehensive register include
metformin given in addition to an antenatal dietary and
“metformin”, “randomiz(s)ed trial”, and “pregnancy”. Antenatal
lifestyle intervention, and to include women who are
dietary and lifestyle interventions for pregnant women,
overweight as well as those who are obese. The use of
particularly those who are overweight or obese, have been
metformin in addition to an antenatal dietary and lifestyle
investigated as a strategy to limit gestational weight gain, but
intervention in women who are overweight or obese did not
have shown only a modest effect on weight gain in pregnancy
affect the proportion of infants with a birthweight greater than
and very little effect on clinical outcomes. Metformin has been
4000 g. There was no effect of metformin on total gestational
proposed as a possible drug treatment for use among obese
weight gain or on clinical pregnancy and birth outcomes.
pregnant women, although recent studies have reported
conflicting findings with regards to the effect on gestational Implications of all the available evidence
weight gain and some pregnancy outcomes. In the EMPOWaR The use of metformin in this clinical setting should not be
trial, the investigators reported no difference between advocated. Future research strategies should focus on
metformin and placebo groups in birthweight centile and no improving women’s health and diet to encourage weight loss
differences in any of the secondary outcomes including before conception.
who are overweight or obese, as a strategy to limit since the risk of adverse pregnancy outcomes increases
gestational weight gain and thereby improve preg with increasing maternal BMI,3 there might be clinical
nancy, birth, and infant outcomes. In an individual
benefit in providing treatment for women who are
participant data meta-analysis incorporating data from overweight as well as for those who are obese.
36 randomised trials and more than 12 500 pregnant The aim of the GRoW (metformin and dietary advice to
women who received an antenatal dietary or lifestyle improve insulin sensitivity and promote Gestational
intervention,2 a modest effect on gestational weight gain Restriction of Weight in pregnant women who are
(mean difference –0·7 kg) was identified, but with very overweight or obese) randomised trial was to assess the
little effect on clinical pregnancy outcomes, consistent effects on maternal and infant outcomes of antenatal
with other reports.5 These findings suggest the need for metformin in addition to dietary and lifestyle advice
assessment of additional strategies in this population. among overweight and obese pregnant women.
Metformin has been considered for use in pregnant
women who are overweight or obese, in view of the Methods
associations between high maternal BMI and gestational Study design and participants
diabetes, and a similar intrauterine milieu of insulin resis GRoW was a multicentre, randomised, double-blind,
tance, hyperglycaemia, hyperlipidaemia, and chro nic placebo-controlled trial in which women were recruited
inflammation.6 Metformin has insulin-sensitising from the three major public maternity hospitals in
properties, reducing hepatic glucose production and metropolitan Adelaide, SA, Australia (Women’s and
increasing peripheral glucose utilisation,7 and is used Children’s Hospital, Lyell McEwin Hospital, and Flinders
increasingly in the treatment of women with gestational Medical Centre). Eligible women had a live singleton
diabetes.8 pregnancy between 10 and 20 weeks’ gestation and
Previous trials9,10 assessing the use of metformin in were overweight (BMI 25·0–29·9 kg/m²) or obese
obese pregnant women have had conflicting findings with (BMI ≥30·0 kg/m²) at their first antenatal visit. Women
respect to effects on gestational weight gain and some with a multiple pregnancy, type 1 or 2 diabetes diagnosed
pregnancy outcomes. Importantly, women participating prior to pregnancy, or with significant renal or hepatic
in these previous randomised trials were not provided impairment such that metformin therapy was
with a dietary and lifestyle intervention. Furthermore, contraindicated, were excluded.
The study protocol (appendix)11 was approved by the women were provided with written information, an See Online for appendix
Women’s and Children’s Health Network Human individual diet and physical activity plan, recipe book,
Research Ethics Committee, with local institutional and example menu plans. Participants were encouraged
approval at each site. Participating women provided to set achievable goals for dietary and exercise change
written informed consent. and were supported to make these lifestyle changes and
to self-monitor their progress using a SMART (Specific,
Randomisation and masking Measurable, Achievable, Relevant, and Timely) goals
All women presenting for antenatal care at the partici approach. These principles were reinforced at
pating centres had their height and weight measured subsequent face-to-face visits with the dietitian and
and BMI calculated at their first antenatal appointment. research assistant, and during the telephone contacts.
Eligible women were then counselled by a research All participants completed a food frequency question
assistant. Those women who consented to participate naire, exercise diary, and quality-of-life and emotional
were randomly assigned (1:1) to metformin or placebo, wellbeing assessments at trial entry and at 28 and
by use of a central online randomisation service. The 36 weeks’ gestation. Consistent with state-wide clinical
computer-generated randomisation schedule util ised practice guidelines for South Australia, all women were
blocks of four and was prepared by an investigator not screened for gestational diabetes at about 28 weeks’
involved with recruitment or clinical care. Stratification gestation.15 During the course of the trial, diagnostic
occurred for parity (0 vs 1 or more), BMI at antenatal criteria for gestational diabetes changed across the state
booking visit (25·0–29·9 kg/m² vs ≥30·0 kg/m²), and from a positive 75 g oral glucose tolerance test with a
collaborating centre. fasting blood glucose concentration of 5·5 mmol/L or
Metformin tablets (500 mg) and placebo tablets above, or a 2 h blood glucose concentration of
identical in appearance were packaged by an independent 7·8 mmol/L above, to a fasting blood glucose of
pharmaceutical packaging company (Pharmaceutical 5·1 mmol/L or above, a 1 h blood glucose concentration
Packaging Professionals, Port Melbourne, VIC, of 10·0 mmol/L or above, or a 2 h blood glucose con
Australia) and coded in accordance with the centration of 8·5 mmol/L or above.15 Women diagnosed
randomisation schedule. Participants, their antenatal with gestational diabetes remained in the study and
care providers, and research staff (including outcome were offered treatment with further dietary modification
assessors) were masked to treatment allocation. and metformin or insulin added as required to maintain
appropriate glycaemic control.15 Women who required
Procedures metformin for the treatment of gestational diabetes
All participants received a 16-week supply of tablets were prescribed this medication in addition to their
(metformin 500 mg or identical placebo) and a further study medication. All other care during pregnancy and
12-week supply at 28 weeks’ gestation. Women com birth was done in accordance with local hospital
menced tablets from randomisation, starting with one practices.15
tablet per day for 1 week, and increasing to a maximum
of two tablets twice per day (maximum dose metformin Outcomes
2000 mg per day) over 4 weeks as tolerated, and then The primary outcome was the proportion of infants with
continuing until birth. At 36 weeks’ gestation, women birthweight greater than 4000 g.
completed a questionnaire to ascertain compliance with Several categories of secondary outcomes were also
medication and the occurrence of any side-effects. assessed. First, maternal weight gain outcomes assessed
All participating women received dietary and lifestyle were total gestational weight gain (defined as the
advice delivered over three face-to-face sessions (two difference between pregnancy weight obtained at
with a dietitian, shortly after trial entry and at 28 weeks, 36 weeks’ gestation or nearest to birth and early
and one with a research assistant at 36 weeks) and pregnancy weight); average weekly gestational weight
three telephone calls from the research assistant at 20, gain (calculated as total gestational weight gain divided
24, and 32 weeks. Dietary advice was consistent with by actual time in weeks between measurements); and
Australian standards.12,13 Women were advised to main gestational weight gain below, within, or above the US
tain a balance of carbohydrates, fat, and protein, while Institute of Medicine (IoM) recommendations according
specifically encouraging a reduction in the intake of to early pregnancy BMI (defined as 7·0 to 11·5 kg for
energy-dense and non-core foods high in refined women who were overweight and 5·0 to 9·0 kg for
carbohydrates and saturated fats, increased intake of women who were obese).16 Second, maternal diet and
fibre, and consumption of two servings of fruit, physical activity, as measured by questionnaires
five servings of vegetables, and three servings of dairy completed at trial entry and at 28 and 36 weeks’ gestation
each day.12,13 (using the Harvard Semi-Quantitative Food Frequency
Tailoring of the intervention was informed by stage Questionnaire17 and the Short Questionnaire to Assess
theories of health decision making.14 Initially, there was Health-Enhancing Physical Activity18). Third, maternal
a planning session with a research dietitian, in which pregnancy and birth outcomes assessed were
by combining the estimates from each imputation using onset fetal growth restriction and pre-eclampsia). There
Rubin’s rules.25 As there were no missing values for the was one neonatal death in the metformin group following
primary outcome, no missing-not-at-random sensitivity extremely preterm birth (before 22 weeks). None of the
analyses were done. perinatal deaths were determined to be attributable to
Secondary analyses for the primary outcome and participation in the trial.
secondary infant anthropometric outcomes were The baseline characteristics of women at trial entry
prespecified to test for evidence of modification of the were similar in the two treatment groups (table 1). Across
effect of treatment by maternal BMI category (overweight the whole study cohort, median gestational age at trial
vs obese). entry was 16·29 weeks (IQR 14·43–18·00) and median
Statistical analyses were done with SAS version 9.4 or BMI was 32·32 kg/m² (28·90–37·10); 167 (32%)
Stata version 14. participants were overweight and 347 (68%) were obese.
We also did a post-hoc meta-analysis incorporating the Adequate data were available for all 514 infants for the
findings of the GRoW trial into our previously reported primary outcome of birthweight greater than 4000 g.
Cochrane systematic review,26 which was done in 404 (79%) of 514 participants provided data for
accordance with standard Cochrane methods, to assess experience of adverse effect, with no differences
the impact of metformin on gestational weight gain, identified between treatment groups (table 2). Of 354
birthweight, and risk of pre-eclampsia. Briefly, (69%) participants who provided data for medication
randomised and quasi-randomised trials assessing compliance, 296 (84%) took the maximum of four tablets
metformin use (compared with placebo or no metformin) per day (appendix).
in pregnant women who were overweight or obese, There was no significant difference in the proportion
defined as an early pregnancy or pre-pregnancy BMI of of infants with birthweight greater than 4000 g between
25·0 kg/m² or higher, were identified from a search of the metformin and placebo groups (40 [16%] of 256 in the
the Cochrane Pregnancy and Childbirth’s Trials Register.
Two review authors independently assessed risk of bias 3546 eligible women approached to participate
for each study, with statistical analyses done with Review
Manager 5. A fixed-effect meta-analysis was used to
combine data where trials were judged to be sufficiently 3022 declined to participate
259 did not want to
similar, and a random-effects meta-analysis was used to participate in research
produce an overall summary if substantial statistical 481 did not want to take
study medication
heterogeneity was detected. For the present report, we 547 too busy or unable
incorporated the data from the present trial into the to attend for study
visits
existing systematic review and meta-analysis using 531 other reasons*
Review Manager 5. For dichotomous data (risk of pre- 1204 were not interested,
eclampsia), results are presented as summary risk ratio did not give a reason,
or could not be
with 95% CIs; for continuous data (gestational weight contacted
gain and birthweight), results are presented as mean
difference with 95% CIs.
524 consented and randomly assigned
The GRoW trial is registered with the Australian New
Zealand Clinical Trials Registry, number
ACTRN12612001277831.
261 assigned to metformin 263 assigned to placebo
Role of the funding source
The funder had no role in the study design, data
0 withdrew consent 1 withdrew consent
collection, analysis, interpretation, or writing of the 5 had miscarriage before 2 had miscarriage before
report. The corresponding author had full access to the 20 weeks or termination of 20 weeks or termination of
pregnancy pregnancy
data and final responsibility for the decision to submit 0 had stillbirth after 20 weeks 2 had stillbirth after 20 weeks
for publication. 256 liveborn infants 258 liveborn infants
1 neonatal death 0 neonatal death
0 maternal death 0 maternal death
Results
Between May 28, 2013 and April 26, 2016, 3546 eligible
women were approached to participate; 524 women were 256 woman and infant pairs 258 woman and infant pairs
included in the analysis included in the analysis
randomly assigned, 261 to metformin and 263 to placebo
(figure). 514 women and infant pairs were included in
the analyses (256 in the metformin group and 258 in the Figure: Trial profile
*Other reasons includes any reason not covered by the other categories,
placebo group). No maternal deaths occurred. There including partner or family against participation, did not think they were
were two stillborn infants, both in the placebo group (one overweight, were already participating in another research project, and did not
secondary to acute chorioamnionitis and one due to early think that they needed lifestyle intervention.
Women’s and Children’s Hospital 110 (43%) 112 (43%) Nausea 74 (36%) 65 (33%)
Lyell McEwin Hospital 117 (46%) 118 (46%) Vomiting 49 (24%) 46 (23%)
Gestational age at entry (weeks) 16·29 (14·43–18·00) 16·29 (14·57–18·14) Diarrhoea 62 (30%) 44 (22%)
BMI (kg/m )2
32·50 (28·71–37·54) 32·05 (29·10–36·80) Fatigue 92 (44%) 98 (50%)
25·0–29·9 kg/m 2
83 (32%) 84 (33%) Skin rash 14 (7%) 19 (10%)
≥30·0 kg/m 2
173 (68%) 174 (67%) Loss of appetite 47 (23%) 40 (20%)
Public patient 252 (98%) 254 (98%) Data are the number (%) of women who reported at least one specified adverse
event during the study period up to completion of the questionnaire at 36 weeks.
Weight (kg) 92·89 (19·76) 91·80 (19·79)
Data are for all randomly assigned women who did not withdraw consent to use
Height (cm) 165·27 (6·76) 164·87 (6·79) their data, and who did not have a miscarriage or termination of pregnancy
Race before 20 weeks’ gestation, or stillbirth. *Excludes 48 (19%) of 256 participants
White 210 (82%) 221 (86%) who did not complete the questionnaire assessing compliance with medication
and experience of adverse effects. †Excludes 62 (24%) of 258 participants who did
East Asian 5 (2%) 7 (3%) not complete the questionnaire assessing compliance with medication and
Indian 12 (5%) 6 (2%) experience of adverse effects.
Other 24 (9%) 19 (7%)
Table 2: Adverse effects
Unknown 5 (2%) 5 (2%)
Smoker 24 (9%) 43 (17%)
Nulliparous 88 (34%) 92 (36%) (table 4), including diagnosis of gestational diabetes,
Previous preterm birth 13 (5%) 13 (5%) irrespective of the diagnostic criteria used (appendix).
Previous pre-eclampsia 13 (5%) 10 (4%) Furthermore, following diagnosis of gestational diabetes,
Previous stillbirth or neonatal death 1 (<1%) 6 (2%)
there were no significant differences between the
Previous caesarean section 48 (19%) 67 (26%)
treatment groups with respect to use of either metformin
Family history of diabetes 83 (32%) 78 (30%)
or insulin therapy (appendix).
Family history of hypertension 105 (41%) 104 (40%)
Overall the risks of maternal labour complications did
not differ significantly between the treatment groups
Family history of heart disease 58 (23%) 61 (24%)
(table 5). Although there was a significant reduction in
Index of socioeconomic disadvantage*
risk of caesarean birth among women in the metformin
Quintile 1 (most disadvantaged) 76 (30%) 95 (37%)
group compared with those in the placebo group, this
Quintile 2 78 (30%) 74 (29%)
finding was partly the result of a difference in the number
Quintile 3 31 (12%) 30 (12%)
of women undergoing an elective repeat caesarean birth
Quintile 4 52 (20%) 43 (17%)
(30 women in the metformin group vs 44 women in the
Quintile 5 (least disadvantaged) 19 (7%) 16 (6%)
placebo group). Self-reported maternal quality of life and
Data are mean (SD), median (IQR), or n (%). Data are for all randomly assigned women who did not withdraw consent emotional wellbeing did not differ between groups
to use their data, and who did not have a miscarriage or termination of pregnancy before 20 weeks’ gestation, or a
stillbirth. *Socioeconomic index of relative social disadvantage (as defined by the Australian Bureau of Statistics’
(appendix).
2011 Socio-Economic Index for Areas—Index of Relative Socio-Economic Disadvantage quintile24). Mean gestational age at birth was similar in the
two treatment groups, as was mean infant birthweight
Table 1: Baseline characteristics
(table 3). There were no significant differences between
the two groups with respect to other infant outcomes
metformin group vs 37 (14%) of 258 in the placebo group; (table 3) or newborn anthropometric measures
aRR 0·97, 95% CI 0·65 to 1·47; p=0·899; table 3). (appendix) apart from infant abdominal circumference at
Compared with the placebo group, women in the birth, which was 0·5 cm smaller in the metformin group
metformin group had lower average weekly gestational (p=0·049; appendix), although this could be a chance
weight gain and were more likely to have weight gain finding and the small difference is of questionable
below the IoM recommendations; total gestational clinical significance.
weight gain was not significantly different between the A prespecified secondary analysis identified some
treatment groups (table 4). Women in both groups were evidence of effect modification by maternal BMI category,
successful in improving the quality of their dietary with the use of metformin associated with a significant
intake, but dietary patterns and physical activity did not reduction in the proportion of infants with birthweight
differ between the two treatment groups at trial entry or greater 4000 g and reductions in total gestational weight
over the course of pregnancy (appendix). There were no gain and weight gain above IoM recommendations
significant differences between the two treatment groups among overweight women but not among obese women
with respect to maternal risk of pregnancy complications (appendix).
Metformin Placebo (n=258) Unadjusted risk Unadjusted Adjusted risk ratio Adjusted
(n=256) ratio (95% CI) p value (95% CI) p value
Chorioamnionitis* 3 (2·22%) 5 (3·40%) ·· ·· ·· 0·725
Induction of labour 104 (40·63%) 87 (33·72%) 1·20 (0·96 to 1·51) 0·107 1·19 (0·95 to 1·49) 0·139
Antibiotic use during delivery† 135 (52·73%) 147 (56·98%) 0·93 (0·79 to 1·08) 0·334 0·95 (0·81 to 1·11) 0·493
Caesarean section (all)† 87 (33·98%) 111 (43·02%) 0·79 (0·63 to 0·99) 0·037 0·80 (0·64 to 1·00) 0·049
Elective caesarean section‡ 44 (17·19%) 54 (20·93%) 0·82 (0·57 to 1·18) 0·281 0·72 (0·51 to 1·00) 0·053
Emergency caesarean section 43 (16·80%) 57 (22·09%) 0·76 (0·53 to 1·09) 0·131 0·79 (0·56 to 1·11) 0·168
Postpartum haemorrhage 66 (25·79%) 58 (22·55%) 1·14 (0·84 to 1·56) 0·393 1·15 (0·85 to 1·57) 0·367
>600 mL
3rd or 4th degree perineal 6 (2·34%) 3 (1·15%) ·· ·· ·· 0·336
trauma*
Wound infection* 3 (1·17%) 3 (1·16%) ·· ·· ·· >0·999
Endometritis* 3 (1·17%) 3 (1·16%) ·· ·· ·· >0·999
Thromboembolic disease* 1 (0·39%) 0 ·· ·· ·· 0·497
Results presented are based on imputed data, unless otherwise specified. Data are n (%; with numbers calculated from proportions averaged across imputations), unless
otherwise indicated. Data are for all randomly assigned women who did not withdraw consent to use their data, and who did not have a miscarriage or termination of
pregnancy before 20 weeks’ gestation, or stillbirth. Adjusted analyses included stratification variables (centre, parity, BMI category), maternal smoking status, age at consent,
and socioeconomic status quintile as covariates. *Insufficient events for analysis or imputation; p value from Fisher’s exact test on un-imputed data. †Poisson model with
robust variance estimation used for adjusted analysis because binomial model did not converge. ‡Includes 30 women in the metformin group and 44 women in the placebo
group who had an elective repeat caesrean section.
The planned health economic cost analysis was not participants, central randomisation, and appropriate
done because the intervention had no significant impact masking of participants, clinicians, and outcome asses
on clinical outcomes in the overall study population. sors. Furthermore, we prespecified outcomes of clinical
In a post-hoc meta-analysis incorporating the findings relevance, followed a prespecified analysis plan, and had
of the GRoW trial into the previously reported Cochrane complete follow-up data for the primary outcome.
systematic review,26 metformin was identified to have GRoW is the first trial to assess the effect of metformin
only a modest impact on gestational weight gain among given in addition to an antenatal dietary and lifestyle
pregnant women who were overweight or obese (mean intervention. Following dietary intervention over the
difference –2·27 kg, 95% CI –4·45 to –0·08 [ four studies, course of pregnancy, women across both treatment
1278 women]). There was no evidence of an impact on groups were successful in modifying their dietary intake,
birthweight (mean difference 11·22 g, –52·87 to 75·31 specifically reducing their overall energy intake and
[three studies; 1348 infants]) or the risk of pre-eclampsia improving their healthy eating index score as an
(risk ratio 0·82, 95% CI 0·25 to 2·68 [three studies; indicator of diet quality. These findings are consistent
1355 women]). with those reported previously from the LIMIT trial,12,27
substantiating the reproducibility of the intervention
Discussion among overweight and obese pregnant women in
In the GRoW randomised trial, use of metformin in producing dietary change. Together, these findings also
addition to dietary and lifestyle advice did not affect the suggest that many of the complex determinants of
proportion of infants with a birthweight greater than gestational weight gain might not be readily modifiable
4000 g compared with placebo and dietary and lifestyle by changes to dietary intake and physical activity.
advice. Although there were no significant differences in Our trial population was predominantly white and
total gestational weight gain, women who received from areas with high levels of socioeconomic dis
metformin were more likely to have weight gain below advantage. Additionally, 85% of eligible women declined
IoM recommendations, and had lower average weekly to participate—among the most common reasons cited
gestational weight gain. However, these differences in were time constraints and lack of willingness to take
gestational weight gain were not accompanied by medication during pregnancy. These factors might limit
significant effects on maternal and infant pregnancy or the generalisability of our findings to some extent.
birth outcomes. However, such rates of uptake of trial participation (15%)
Our trial has several strengths. The trial is the first of are within the range described by similar trials in the
its kind to include both overweight and obese women, scientific literature, which have varied from 13% to 47%.9,10
which is important in view of the potential for benefit Our measure of compliance suggests that 84% of
from interventions at different levels of BMI above the respondents took the maximum of four tablets per day,
normal range. We used robust methods, including with no differences identified between the treatment
prospectively measured height, weight, and BMI in all groups. These figures are difficult to directly compare
with other studies, which have used varying measures to –4·45 to –0·08 [ four studies; 1278 women]). Although
assess treatment compliance,9,10 although our measure is there are inherent variations in the design of the
similar to the overall compliance of 80% reported by randomised trials done to date,9,10,29 including the BMI of
Syngelaki and colleagues10 and 66% reported by Chiswick women recruited (overweight and obese in the current
and colleagues.9 trial vs obese only9,10,29), the dose of metformin assessed
Although metformin is increasingly used in pregnancy (1000 mg29 up to 3000 mg10 per day), and the inclusion of
for the treatment of gestational diabetes, and there is a dietary and lifestyle intervention for all women in the
evidence to support its safety,8 longer-term childhood current trial, these differences seem insufficient to
follow-up of the offspring of women who participated in explain the overall observations of a limited reduction in
this trial will be important. The primary findings from gestational weight gain and lack of effect on clinical
the MIG trial did not identify any differences in infant outcomes.
birthweights for women with gestational diabetes treated Such a limited effect of metformin is, however,
with insulin or metformin.8 However, follow-up of child consistent more broadly with the evidence relating to
participants at age 7–9 years suggested that children ex prenatal dietary and lifestyle interventions2 for women
posed in utero to metformin, as compared with insulin, who are overweight or obese, and raises important
had higher bodyweight and BMI, with some suggestion questions concerning their modest at best reduction in
of increased adipose tissue deposition as measured by gestational weight gain, and little evidence of effect on
dual-energy x-ray absorptiometry.28 clinical outcomes. Importantly, the IoM recom men
During the course of the GRoW trial there was a dations16 for gestational weight gain are based on
change in the diagnostic criteria for gestational diabetes extensive observational evidence identifying associations
across South Australia.15 However, the criteria were between gestational weight gain and pregnancy out
applied equally across the two treatment groups, and we comes, with an underlying assumption that the observed
demonstrated no significant differences in diagnosis or associations are causal in nature. Furthermore, it has
requirements for metformin or insulin between the two been assumed that gestational weight gain can be
treatment groups. The change in criteria might have modified either through use of metformin, or via
resulted in an increase in the number of women being changes to diet and lifestyle during pregnancy, and that
diagnosed with gestational diabetes; however, discussion weight gain within the recommended range leads to
of the relative merits and evidence base to support a optimal maternal and infant outcomes.
range of different testing approaches and criteria is These prevailing assumptions need to be questioned
beyond the scope of this report. in view of the mounting available clinical evidence from
Our findings do not suggest that metformin impacts randomised trials. For pregnant women who are
clinical pregnancy and birth outcomes—this conclusion overweight or obese, intervention during pregnancy
is broadly consistent with the existing scientific (whether through metformin, dietary or lifestyle
literature.9,10,29 Syngelaki and colleagues reported a signi modification, or a combination of the two) might be too
ficant reduction in the risk of pre-eclampsia with met little, too late, emphasising the need to target women,
formin use.10 However, this effect was not evident when particularly those who are overweight or obese, before
the current trial data were incorporated into the available conception to improve their diet and lifestyle and to
published meta-analysis,26 raising the possibility that this encourage weight loss.30 Notwithstanding the logistical
was a spurious finding from a secondary outcome in implications of intervention before conception, there is
Syngelaki and colleagues’ trial. currently little evidence to support improved pregnancy
We did not identify a significant reduction in total outcomes following preconception weight loss, and
gestational weight gain, although there was some robust evalu ation of such an approach is urgently
evidence of effect modification, with a possible benefit needed.
from metformin in reducing the risk of infant birth In conclusion, our findings suggest that the use of
weight greater than 4000 g and reducing total gestational metformin in addition to a dietary and lifestyle inter
weight gain among overweight women, but not among vention among pregnant women who are overweight or
obese women. Although this secondary analysis was obese was associated with some evidence of reduced
prespecified, these findings should be interpreted with gestational weight gain measures, but did not affect
caution, and warrant confirmation by others. clinical pregnancy and birth outcomes. Use of metformin
Our overall findings in relation to gestational weight in this clinical setting should not be advocated on the
gain are consistent with those of Chiswick and colleagues basis of these data.
(EMPOWaR trial),9 but not with those of Syngelaki and Contributors
colleagues (MOP trial)10 and Abd El Fattah.29 All authors were involved in the development and design of the trial, the
Incorporation of our data into the Cochrane systematic conduct of the trial, drafting of the report, and revision for intellectual
content. JL and JMD did the statistical analyses.
review and meta-analysis26 suggests that, overall,
metformin has a very modest effect on gestational Declaration of interests
We declare no competing interests.
weight gain (mean difference –2·27 kg, 95% CI