Research

JAMA | Original Investigation

Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes

in Early Pregnancy
The MOMPOD Randomized Clinical Trial
Kim A. Boggess, MD; Arielle Valint, MS; Jerrie S. Refuerzo, MD; Noelia Zork, MD; Ashley N. Battarbee, MD, MSCR; Kacey Eichelberger, MD;
Gladys A. Ramos, MD; Gayle Olson, MD; Celeste Durnwald, MD; Mark B. Landon, MD; Kjersti M. Aagaard, MD, PhD; Kedra Wallace, PhD;
Christina Scifres, MD; Todd Rosen, MD; Wadia Mulla, MD; Amy Valent, DO; Sherri Longo, MD; Laura Young, MD, PhD; M. Alison Marquis, MStat;
Sonia Thomas, DrPH; Ashley Britt, MS; Diane Berry, PhD

Visual Abstract
IMPORTANCE Insulin is recommended for pregnant persons with preexisting type 2 diabetes Editorial page 2167
or diabetes diagnosed early in pregnancy. The addition of metformin to insulin may improve
neonatal outcomes. Supplemental content

OBJECTIVE To estimate the effect of metformin added to insulin for preexisting type 2 or
diabetes diagnosed early in pregnancy on a composite adverse neonatal outcome.

DESIGN, SETTING, AND PARTICIPANTS This randomized clinical trial in 17 US centers enrolled
pregnant adults aged 18 to 45 years with preexisting type 2 diabetes or diabetes diagnosed
prior to 23 weeks’ gestation between April 2019 and November 2021. Each participant was
treated with insulin and was assigned to add either metformin or placebo. Follow-up was
completed in May 2022.

INTERVENTION Metformin 1000 mg or placebo orally twice per day from enrollment
(11 weeks -<23 weeks) through delivery.

MAIN OUTCOME AND MEASURES The primary outcome was a composite of neonatal
complications including perinatal death, preterm birth, large or small for gestational age, and
hyperbilirubinemia requiring phototherapy. Prespecified secondary outcomes included
maternal hypoglycemia and neonatal fat mass at birth, and prespecified subgroup analyses
by maternal body mass index less than 30 vs 30 or greater and those with preexisting vs
diabetes early in pregnancy.

RESULTS Of the 831 participants randomized, 794 took at least 1 dose of the study agent and
were included in the primary analysis (397 in the placebo group and 397 in the metformin
group). Participants’ mean (SD) age was 32.9 (5.6) years; 234 (29%) were Black, and 412
(52%) were Hispanic. The composite adverse neonatal outcome occurred in 280 (71%) of the
metformin group and in 292 (74%) of the placebo group (adjusted odds ratio, 0.86 [95% CI
0.63-1.19]). The most commonly occurring events in the primary outcome in both groups
were preterm birth, neonatal hypoglycemia, and delivery of a large-for-gestational-age infant.
The study was halted at 75% accrual for futility in detecting a significant difference in the
primary outcome. Prespecified secondary outcomes and subgroup analyses were similar
between groups. Of individual components of the composite adverse neonatal outcome,
metformin-exposed neonates had lower odds to be large for gestational age (adjusted odds
ratio, 0.63 [95% CI, 0.46-0.86]) when compared with the placebo group.

CONCLUSIONS AND RELEVANCE Using metformin plus insulin to treat preexisting type 2 or
gestational diabetes diagnosed early in pregnancy did not reduce a composite neonatal
adverse outcome. The effect of reduction in odds of a large-for-gestational-age infant
observed after adding metformin to insulin warrants further investigation.

TRIAL REGISTRATION [Link] Identifier: NCT02932475

Author Affiliations: Author
affiliations are listed at the end of this
article.
Corresponding Author: Kim A.
Boggess, MD, University of
North Carolina at Chapel Hill, 3010
Old Clinic Bldg/CB 7570, Chapel Hill,
NC 27599-7570 (kboggess@med.
JAMA. 2023;330(22):2182-2190. doi:10.1001/jama.2023.22949 [Link]).

2182 (Reprinted) [Link]

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy Original Investigation Research

T
ype 2 diabetes is the most prevalent form of diabetes
in the US1 and accounts for most of the preexisting Key Points
diabetes in pregnancy. Pregnancies with preexist-
Question Does metformin added to insulin for the treatment
ing type 2 diabetes or gestational diabetes identified early of preexisting type 2 diabetes or diabetes identified early
in pregnancy are at risk for adverse pregnancy outcomes, in pregnancy reduce the risk of adverse neonatal
intrapartum complications, and neonatal hypoglycemia outcomes?
at birth.2-7 Maintaining maternal euglycemia is the mainstay
Findings In a randomized clinical trial of 794 pregnant adults
of treatment. (18-45 years), compared with placebo, metformin added to
Management of diabetes in pregnancy to optimize peri- insulin for treatment of preexisting diabetes or diabetes
natal outcome includes maternal blood glucose monitoring, identified in early pregnancy did not reduce a composite
dietary management, and insulin therapy to achieve mater- adverse neonatal outcome (71% vs 74%) but resulted in fewer
nal euglycemia. 3,8 Metformin use has been reported for large-for-gestational-age infants.
pregnancies complicated by preexisting type 2 diabetes;9-11 Meaning Using metformin plus insulin to treat preexisting
however, these studies are limited by their study design, type 2 or gestational diabetes diagnosed early in pregnancy
smaller sample size, and common need for adjuvant insulin. did not reduce a composite neonatal adverse outcome.
The American College of Obstetricians and Gynecologists12 The reduction in odds of a large-for-gestational-age infant
observed after adding metformin to insulin warrants further
and the American Diabetes Association13 both recommend
investigation.
insulin as first-line pharmacotherapy for preexisting type 2
diabetes in pregnancy, with metformin reserved for those
who cannot use insulin or decline to do so. The primary
objective of the Medical Optimization of Management of tified prior to 23 weeks requiring insulin, were eligible to
Overt Type 2 Diabetes in Pregnancy (MOMPOD) trial was to participate.14 Because of the relationship between maternal
test the hypothesis that among US pregnant adults with dia- race, ethnicity, and pregnancy outcomes, participant race
betes (preexisting type 2 diabetes or diabetes identified and ethnicity were collected by chart abstraction and if miss-
early in gestation) in pregnancy, the addition of metformin ing from the chart, by participant self-report.
to insulin lowers the odds of composite adverse neonatal The diagnosis of preexisting type 2 diabetes prior to preg-
outcome of perinatal death or severe neonatal complica- nancy was confirmed by medical record review. Diabetes
tions compared with insulin alone, without increasing risk diagnosed early in pregnancy was defined by either 1-step
for adverse perinatal events. testing (75-g oral glucose tolerance testing with ≥1 abnormal
value)15 or 2-step testing (positive 50-g oral glucose tolerance
testing followed by 100-g glucose tolerance testing with ≥2
abnormal values by Carpenter-Coustan criteria16); or by a gly-
Methods cated hemoglobin result of at least 6.5%, a fasting capillary
Study Design blood glucose level greater than 126 mg/dL; or a random cap-
The MOMPOD Study was a multicenter, double-blinded illary blood glucose level of at least 200 mg/dL17,18; or any
randomized clinical trial of metformin vs placebo added to combination of testing that the treating clinician felt required
insulin for the treatment of preexisting type 2 diabetes insulin treatment.
or gestational diabetes diagnosed in early pregnancy. The
trial design and protocol were previously published (Sup- Trial Procedures
plement 1),14 and the statistical analysis plan is available in All enrolled participants were treated with insulin. Study
Supplement 2. sites used weight-based insulin dosing, split as 2 or 3 injec-
The trial was conducted at 17 US clinical sites with US tions using a combination of rapid/short and intermediate
Food and Drug Administration oversight (investigational or long-acting insulin. Participants performed daily capil-
new drug No. 125594) and institutional review board ap- lary blood glucose monitoring as fasting and either at 1 or
proval at each site; all participants provided written in- 2 hours postprandial. Study sites were instructed to increas-
formed consent. An independent data monitoring and safety ingly titrate insulin dosing to achieve glycemic goals of
board provided oversight and an independent medical moni- fasting 70 to 95 mg/dL; 1-hour postprandial level below
tor was available for acute serious adverse events that might 140 mg/dL, or 2-hour postprandial level below 120 mg/dL.
require unmasking. After enrollment, participants were randomly assigned in
a 1:1 ratio to either metformin or an identical appearing pla-
Trial Population cebo to be taken orally twice a day. The data coordinating
Participants were enrolled from June 2017 to November 2021, center prepared the randomization sequence using permuted
and the last follow up study visit was completed in May blocks,19 with stratification by clinical site, gestational age at
2022. Enrollment was suspended from April 2020 to Sep- randomization (<18 weeks vs ≥18 weeks) and timing of diabe-
tember 2020 due to the COVID-19 pandemic. Pregnant adults tes diagnosis (preexisting or during early pregnancy) using
aged 18 to 45 years with a singleton gestation between 10 the web-based data management system Carolina Data
weeks 0 days and 22 weeks 6 days, confirmed by ultrasound Acquisition and Reporting Tool, developed by the data coor-
with either preexisting type 2 diabetes or with diabetes iden- dinating center.

[Link] (Reprinted) JAMA December 12, 2023 Volume 330, Number 22 2183

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Research Original Investigation Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy

Participants and study staff were masked to group allo- the primary outcome as 25% and 10% loss to follow-up. An
cation. Pylant Village Pharmacy (Village Compounding ad hoc masked review of 102 completed participants found a
Pharmacy/Acclaim Allergy Solutions) provided metformin higher than anticipated rate for the primary event (70%) and
500 mg or placebo in identical-appearing capsules and for loss to follow-up (20%); thus, the data and safety moni-
bottles. Participants were instructed to ingest 1 capsule toring board approved a final enrollment sample size of 950
twice a day for 1 week, then if tolerated, to increase to 2 cap- (428 per group completed) to provide 80% power to detect
sules twice a day.14 a 40% reduction in the primary outcome, assuming a base-
Study staff made monthly visits in person or by tele- line event rate of 50% and a 2-sided α = .044, accounting for
phone to query adverse effects, assess adherence, and dis- planned interim analyses.
pense the study agent either in person or by mail. Chart ab- The primary analysis included randomized participants
straction included maternal and delivery data from enrollment who took at least 1 dose of the study drug. Odds ratios (ORs)
until hospital discharge and neonatal data until 30 days of age and 95% CIs for the proportions of participants meeting the
or hospital discharge, whichever occurred first. If neonates primary outcome were calculated using logistic regression,
were discharged prior to 30 days, then follow-up was ob- adjusting for study site, timing of diabetes diagnosis (preex-
tained by telephone at 30 days of age. Adherence with the study isting vs early in pregnancy), gestational age at randomiza-
agent was assessed at each study visit via self-report. At con- tion stratified at 18 weeks, and baseline maternal BMI. An
clusion of participation, participants were asked to return any unadjusted logistic model was generated by removing all
remaining empty bottles and any unused study agent. Study baseline covariates to evaluate whether stratification at ran-
staff collected neonate anthropometric measurements within domization adequately mitigated treatment differences due
72 hours of birth using previously published methods20; neo- to differences in covariates at baseline. Where applicable,
nates born before 28 weeks’ gestation or who were medically values for participants with missing maternal baseline BMI
unstable were not measured. were imputed. Imputation was performed using PROC MI in
SAS 9.4 (SAS Institute), with the fully conditional specifica-
Outcomes tion option and predictive mean matching with 20 imputed
The primary outcome was a composite of 1 or more of the data sets and 50 burn in iterations.22 The statistical analysis
following: fetal or neonatal death (fetal loss between 10 and plan (Supplement 2) describes planned interim and safety
20 weeks, stillbirth ≥20 weeks, or neonatal death within 28 analyses, subgroup analyses, and sensitivity analyses evalu-
days of birth); neonatal hypoglycemia (capillary blood glu- ating the effect of missing data on the effect of metformin
cose <40 mg/dL or any hypoglycemia that required intrave- on the primary outcome (eTable 1 in Supplement 3).
nous glucose replacement); umbilical artery pH less than Using similar procedures, groups were compared for sec-
7.05; shoulder dystocia with brachial plexus injury, clavicu- ondary outcomes using logistic regression and analysis of
lar or humeral fracture, or 3 or more maneuvers to relieve; covariance. Treatment groups were also compared for pro-
hyperbilirubinemia requiring phototherapy within the first portions of large- and small-for-gestational-age infants, mean
72 hours after birth; delivery before 37 weeks’ gestation; birthweight, proportion of participants experiencing exces-
large-for-gestational-age infant; small-for-gestational-age sive weight gain as defined by the Institute of Medicine,23
infant; and/or low birth weight of less than 2500 g. The and other birth outcomes using the χ2 test or a 2-sample
Fenton growth chart was used to determine birthweight t test, as appropriate. Continuous data were tested for skew-
percentile21 and defined large for gestational age as birth- ness, and if skewed, geometric means were compared.
weight above the 90th percentile and small for gestational SAS version 9.3 or later software was used for statistical
age as birthweight below the 10th percentile. Reviewers analyses. All programming and data analysis was conducted
masked to group assignment adjudicated primary outcomes by the MOMPOD data coordinating center, which is housed in
in a random subset of participants (n = 60). the Collaborative Studies Coordinating Center at the Univer-
Prespecified secondary outcomes were the occurrence sity of North Carolina at Chapel Hill.
of clinically relevant maternal hypoglycemia, defined as
capillary blood glucose of less than 60 mg/dL, and neonatal
fat mass. Prespecified subgroup analyses for the primary
outcome included maternal body mass index (BMI, calcu-
Results
lated as weight in kilograms divided by height in meters Participant Characteristics
squared; <30 vs ≥30 [obese]), those with preexisting type 2 Screening for eligibility among 2667 pregnant adults took
diabetes vs those with diabetes identified early in preg- place from June 2017 to December 2021. After 695 partici-
nancy, and gestational age at randomization (<18 weeks vs pants had given birth and completed follow-up, the data
18-<23 weeks). and safety monitoring board recommended cessation of the
study due to futility, and enrollment was halted in Novem-
Statistical Analysis ber 2021. Of the 2003 eligible participants, 835 (42%) pro-
The target sample size was 1200 participants to support vided written informed consent and 831 were randomized.
adequate power to detect a reduction in odds of the compos- Overall, 794 participants who reported taking at least 1 dose
ite adverse neonatal outcome by 40% for those randomized of the study agent were included in the final analysis
to metformin vs placebo, assuming an overall incidence of (Figure). The primary analysis included 397 participants in

2184 JAMA December 12, 2023 Volume 330, Number 22 (Reprinted) [Link]

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy Original Investigation Research

Figure. MOMPOD Study Flow Diagram

2667 Pregnant patients with a singleton; gestation and

pregestational type 2 diabetes or diabetes identified
early in pregnancy assessed for eligibility

1836 Excluded
664 Did not meet inclusion criteria
1172 Declined to participate
375 Declined metformin
264 No reason given
141 Declined research
117 Wanted metformin
47 Declined insulin
35 Clinician declined
199 Other reasons

831 Randomizeda

415 Randomized to receive 416 Randomized to receive

metformin plus insulin insulin plus placebo

18 Excluded from analysis, 19 Excluded from analysis,

no study drug taken no study drug taken

a
113 Discontinued study drugb 119 Discontinued study drugb Randomization was stratified by
51 Patient decision or extended nonuse 61 Patient decision or extended nonuse clinical site, gestational age at
34 Adverse event 28 Adverse event randomization (<18 weeks vs
11 Developed medical contraindication 17 Developed medical contraindication ⱖ18 weeks) and timing of diabetes
or met exclusion criteria or met exclusion criteria
diagnosis (pregestational or during
17 Other reasons 13 Other reasons
7 Lost to follow-upb 8 Lost to follow-upb pregnancy).
b
Patients who discontinued the
study drug or who were lost to
397 Included in the primary analysis 397 Included in the primary analysis
follow-up were included in the
primary analysis.

each group, and the primary outcome was ascertained from groups (280 [71%] of 397 in the metformin group and 292
768 participants (386 of those randomized to metformin [74%] of 397 in the placebo group; adjusted OR, 0.86
and 382 randomized to placebo). Of the participants who [95% CI, 0.63-1.19]). Results from an unadjusted model and
were randomized and took at least 1 dose of the study agent the per-protocol population were similar. Of individual
in the primary analysis, 768 (97%) had follow-up through components of the composite primary outcome, the metfor-
delivery, and 656 (83%) neonates had follow-up through 30 min group had a lower proportion of large-for-gestational-
days of life. Anthropometric measurements were completed age infants compared with placebo (100 [26%] of 386 live
on 437 (55%) neonates, which was lower than anticipated births vs 137 [36%] of 384 live births; OR, 0.63 [95% CI,
due to restrictions during the early COVID-19 pandemic. 0.46-0.86]). Proportions of small-for-gestational age infants
Table 1 shows participant baseline characteristics. The met- were similar. Analysis of prespecified secondary outcomes
formin group had more participants with obesity, medically showed that clinically relevant maternal hypoglycemia was
treated chronic hypertension, and previous metformin use, also similar between the groups, occurring in 87 (22%) of
which is likely due to chance. 397 assigned to the metformin group and 85 (21%) of 397
Reported adherence was comparable between groups. Of assigned to placebo (adjusted OR, 1.02 [95% CI, 0.72–1.46]).
those randomized to metformin, 244 (61%) reported adher- Among the 437 neonates with anthropometric measure-
ence all or almost all the time (90%-100%) and 13 (3%) less than ments, mean (SD) neonatal body fat mass was comparable
half of the time, compared with 236 (59%) of those random- between those exposed to metformin vs placebo (0.46
ized to placebo reporting adherence all or almost all the time [0.30] kg vs 0.50 [0.24] kg).
and 15 (4%) reporting adherence less than half of the time. At Birth outcomes are shown in Table 3. The proportion of live
study conclusion, less than 10% of participants remembered births, gestational age at delivery, proportion of male neo-
to bring in empty study bottles, so this data were not used to nates, and cesarean deliveries (primary vs repeat and follow-
report adherence. ing labor or in absence of labor), were comparable between
groups. However, compared with the placebo group, mean (SD)
Primary and Secondary Outcomes neonatal birthweight (3089 [74] vs 3244 [78] g) and mean (SD)
The occurrence of the primary outcome (Table 2) was higher birthweight z score (0.49 [1.12] vs 0.81 [1.18]) were lower among
than expected and not significantly different between patients in the metformin group.

[Link] (Reprinted) JAMA December 12, 2023 Volume 330, Number 22 2185

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Research Original Investigation Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy

Table 1. Study Participant Baseline Characteristics Table 1. Study Participant Baseline Characteristics (continued)

No. (%)a No. (%)a

Metformin Insulin Metformin Insulin
plus insulin plus placebo plus insulin plus placebo
(n = 397) (n = 397) (n = 397) (n = 397)
Demographics Gestational age
at enrollment, wkf
Age, mean (SD), y 32.8 (5.5) 33.1 (5.7)
18 to <23 162 (41) 165 (42)
Raceb
<18 235 (59) 232 (58)
American Indian 3 (1) 1 (0)
or Alaska Native HbA1c % at enrollmentd n = 361 n = 361
Asian 10 (3) 13 (3) Mean (SD) 7.7 (2.02) 7.7 (1.97)
Black or African American 119 (30) 115 (29) Median (IQR) 7.2 (6.0-9.1) 7.3 (6.1-9.1)
Native Hawaiian 2 (1) 2 (1) Gestational age 11.1 (4.4) 10.6 (5.4)
or Other Pacific Islander at HbA1c measurement,
White 55 (14) 57 (14) mean (SD), wk
Gestational age 10.3 10.3
≥ 2 Races 8 (2) 2 (1) at HbA1c measurement, (7.9-14.1) (7.4-13.7)
Not reported or declined 95 (24) 97 (24) median (IQR), wk
to report
Abbreviations: HbA1c, hemoglobin A1c; OGTT, oral glucose tolerance test.
Hispanic ethnicityb
SI conversion: for glucose to mmol/L, multiply by 0.0555.
No 194 (49) 188 (47) a
Data are presented as No. (%) unless otherwise indicated.
Yes 203 (51) 209 (53) b
Race and ethnicity were determined by medical record abstraction or if not in
Type of insurance n = 396 n = 395 the record, by participant self-report.
Medicaid, federal, 290 (73) 292 (74) c
Subcategories do not sum because data were missing for 2 participants.
or state insurance d
Reported as %, to convert mmol/mol use (10.93 × HbA1c) − 23.50.
Private 77 (19) 79 (20) e
Calculated as weight in kilograms divided by height in meters squared.
Military 0 3 (1) f
Gestational age determined by menstrual date confirmed by first or second
None 29 (7) 21 (5) trimester ultrasound, and if menstrual date was missing, by ultrasound alone.
First pregnancy 66 (17) 55 (14)
Prior birth ≥20 wk 302 (76) 296 (75)
Diabetes characteristics Subgroup Analyses
Preexisting type 2 diabetes 311 (78) 313 (79) Prespecified subgroup analyses of the primary outcome
requiring medical therapy showed no significant differences in composite adverse neo-
prior to pregnancy
Diagnosis of diabetes 86 (22) 84 (21)
natal outcome in subgroup analyses by maternal BMI of 30 or
early in pregnancy above, timing of diagnosis of diabetes, or gestational age at ran-
Use of diabetes medication 174 (44) 181 (46) domization (eTable 2 in Supplement 3).
prior to pregnancy
Metformin used earlier 162 (41) 175 (44)
in pregnancy Exploratory Outcomes
Criteria to diagnose diabetes n = 86 n = 84 Participants randomized to metformin vs placebo had com-
early in pregnancyc parable daily insulin dosing (geometric mean [SD]: 68.9 [2.1]
HbA1c ≥6.5%d 25 (29) 22 (26)
vs 73.7 [2.04] units) and dosing by maternal weight (mean [SD]:
Fasting blood glucose 5 (6) 5 (6) 0.90 U/kg [0.61] vs 1.11 U/kg [3.10]). There were also compa-
≥126 mg/dL
1-h OGTT≥200 mg/dLd 18 (21) 24 (29) rable between-group proportions with excessive maternal
2-Step method positive 25 (29) 20 (24)
weight gain (metformin group, 185 [52%] of 357 vs 195 [55%]
of 353 in the placebo group). Of 794 participants, 308 (39%)
1-Step method positive 12 (14) 12 (14)
had the third-trimester hemoglobin A1c (HbA1c) level mea-
Other clinical characteristics
and laboratory results sured (155 in the metformin group; mean [SD], 31.5 [3.5] weeks
Chronic hypertension 103 (26) 86 (22) and 153 in the placebo group; mean [SD], 31.7 [3.3] weeks). The
requiring medication
geometric mean [SD] level of HbA1c was lower in the metfor-
Smoking during pregnancy 28 (7) 32 (8)
min group compared with the placebo group (5.97 [0.08] vs
Body mass index n = 383 n = 379
at enrollmente 6.22 [0.08]; geometric mean ratio, 0.96 [95% CI, 0.93-1.00]).
Mean (SD) 36.4 (8.0) 36.3 (8.9) The prevalence of fetal intolerance to labor (32 [8%] vs
Median (IQR) 35.2 34.4 31 [8%]) and failed induction of labor (38 [10%] vs 37 [10%])
(31.2-41.1) (29.6-41.5) was comparable between metformin and placebo groups.
Distribution The prevalence of preeclampsia (metformin group, 122
<30 76 (19) 105 (26) [35%] vs 109 [30%] in the placebo group; adjusted OR, 1.31
30 to <40 198 (50) 164 (41) [95% CI, 0.96-1.78]) was comparable between groups; how-
≥40 109 (27) 110 (28) ever, there were more patients with chronic hypertension
(continued)
requiring medication in the metformin group (103 [26%] of
397) vs the placebo group (86 [22%] of 397). After adjusting

2186 JAMA December 12, 2023 Volume 330, Number 22 (Reprinted) [Link]

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy Original Investigation Research

Table 2. Primary Composite Neonatal Outcome and Components

Metformin Insulin
plus insulin plus placebo Unadjusted absolute difference Adjusted odds ratio
(n = 397)a (n = 397)a (95% CI) (95% CI)b
Composite primary outcomec 280 (71) 292 (74) −3.95 (−11.49 to 3.99) 0.86 (0.63 to 1.19)
Live birthsd 386 (97) 384 (97)
Fetal and neonatal death 11 (3) 13 (3) −4.30 (−24.54 to 15.95) 0.83 (0.36 to 1.89)
Miscarriage <20 wk 7 (2) 4 (1)
Stillbirth ≥20 wk 3 (1) 7 (2)
Neonatal death <28 d 1 (<1) 2 (1)
Preterm birth <37 wk 130 (34) 143 (37) −3.89 (−11.27 to 3.49) 0.86 (0.64 to 1.16)
Neonatal hypoglycemia 152 (39) 162 (42) −2.91 (−10.09 to 4.28) 0.89 (0.67 to 1.19)
Birth trauma 16 (4) 16 (4) −0.14 (−17.83 to 17.56) 1.02 (0.50 to 2.07)
Umbilical artery pH <7.05 9 (2) 9 (2)
Shoulder dystocia 7 (2) 7 (2)
Hyperbilirubinemia 87 (23) 92 (24) −1.99 (−10.35 to 6.37) 0.93 (0.66 to 1.30)
requiring phototherapy
Large for gestational age 100 (26) 137 (36) −11.46 (−19.04 to −3.88) 0.63 (0.46 to 0.86)
(>90th percentile)
Small for gestational age 30 (8) 26 (7) 3.71 (−9.86 to 17.28) 1.17 (0.68 to 2.02)
(<10th percentile)
Low birth weight 81 (21) 73 (19) 3.08 (−5.73 to 11.90) 1.14 (0.80 to 1.63)
(<2500 g)
a
Eleven participants in the metformin plus insulin group and 15 in the insulin and low birth weight odds ratios are reported for live births (observed and
plus placebo group had missing data, and as prespecified in the statistical assumed) and are adjusted for all baseline characteristics noted in Table 1,
analysis plan, are assumed to have the composite primary outcome. except study site as some sites had too few events.
b d
Adjusted by study site, timing of diabetes diagnosis (pregestational vs Percent values for the following categories: preterm birth, neonatal
diagnosed early in pregnancy), gestational age at randomizations stratified at hypoglycemia, birth trauma, umbilical artery pH, shoulder dystocia,
18 weeks, and baseline maternal body mass index as a continuous variable. hyperbilirubinemia requiring phototherapy, large for gestational age, small for
c
Preterm birth before 37 weeks’ gestation, neonatal hypoglycemia, birth gestational age, and low birth weight are calculated from live birth
trauma, hyperbilirubinemia, large for gestational age, small for gestational age, denominators.

Table 3. Study Participant Birth Outcomes

% Unadjusted difference
Metformin Insulin median, or mean
plus insulin plus placebo (95% CI)a
Gestational age at birth, median (IQR), wkb 37.4 (36.1-38.6) 37.3 (35.8-38.3) 0.14 (−0.14 to 0.29)
Cesarean delivery, No. (%)c 237/377 (63) 236/376 (63) 0.11 (−7.28 to 7.50)
Primary, no labor 101 (43) 108 (46)
Primary, after labor 11 (5) 19 (8)
Repeat, no labor 51 (22) 33 (14)
Repeat, after labor 74 (31) 76 (32)
Macrosomia as indication 11 (5) 21 (9)
Birthweight, mean (SD), gb 3089 (74) 3244 (78) −155 (−265 to −45)
Birthweight z score, mean (SD)b,d 0.49 (1.12) 0.81 (1.18) −0.32 (−0.48 to −0.15)
Neonatal fat mass, mean (SD), kge 0.46 (0.30) 0.50 (0.24) −0.04 (−0.09 to 0.01)
Infant male sex, No. (%) 185/379 (48) 204/375 (53) −5.59 (−12.72 to 1.54)
a d
Unadjusted difference for the median is reported using the Hodges-Lehmann Birthweight z score is calculated using z score for gestational age by
estimator and Moses confidence limits using PROC NPAR1WAY in SAS 9.4. weeks = (birthweight – mean)/standard deviation.
b e
Gestational age at birth and birthweight is only reported for live births having Neonatal fat mass was only collected on a proportion of neonates, 227 in the
delivery data, 376 participants in the metformin plus insulin group and 370 metformin plus insulin group and 210 in the insulin plus placebo group.
participants in the insulin plus placebo group. Neonatal fat mass determined by anthropometric measurements using the
c
Denominators were 377 for the metformin plus insulin group and 376 for the formula: fat mass = 0.39055(birthweight) +0.0453(flank skinfold) –
insulin plus placebo group. Cesarean delivery and its subcategory percentages 003237(length) +0.54657 and calculated on 422 (55%) neonates.
are calculated from these denominators, and subcategory percentages may
not sum to 100% due to rounding.

for chronic hypertension requiring medication, the adjusted Infants in the metformin and placebo groups had com-
OR for preeclampsia in the metformin group was 1.24 (95% parable rates of neonatal intensive care unit (NICU) ad-
CI, 0.9-1.71). mission (152 [40%] of 376, vs 165 [45%] of 370), as well as

[Link] (Reprinted) JAMA December 12, 2023 Volume 330, Number 22 2187

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Research Original Investigation Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy

comparable percentages requiring a NICU stay longer than trial, and Hispanic persons tend to have greater insulin
2 days (124 [33%] of 376 vs 131 [35%] of 370). The median resistance compared with non-Hispanic African American
length of NICU stay was 7.5 days for neonates in the metfor- and non-Hispanic White persons. 26 In the current trial,
min group (median IQR, 3-22 days) and 7.0 days for neo- approximately 20% were enrolled with diabetes diagnosed
nates in the placebo group (median IQR, 3-17 days). early in pregnancy, compared to approximately 10% in the
The most common neonatal complication other than hypo- MiTy trial, which may also reflect differences in insulin
glycemia or hyperbilirubinemia was respiratory distress resistance. Compared with the MiTy trial, this trial had
syndrome, which occurred in 61 (16%) of 376 neonates lower prevalence of small-for-gestational-age infants and
exposed to metformin and 64 (17%) of 370 of those exposed was not powered to determine a difference for this out-
to placebo. come. Other factors, including regional differences in health
care access and type and dosing of insulin, could also con-
Adverse Events and Serious Adverse Events tribute to observed trial differences.
The most common adverse events of metformin use were Although this trial observed a modest reduction in
nausea, vomiting, and diarrhea.24 The proportions of par- large-for-gestational-age infants in the metformin group,
ticipants randomized to metformin vs placebo reporting there were no differences in neonatal outcomes nor delivery
nausea (36% vs 39%) and vomiting (25% vs 23%) were com- by cesarean. The cesarean delivery rate in this trial was
parable, while a higher proportion of participants ran- higher than the national rate and the rate among states
domized to metformin reported diarrhea (28% vs 12%; P < .01). where the study was conducted,1 but it was comparable to
Serious adverse event data are provided in eTables 3 and 4 the rate in the MiTy study,25 which reflects the complexity
in Supplement 3. Unadjusted rate differences in maternal of intrapartum care for patients with type 2 diabetes and
and fetal/neonatal serious adverse events were not statisti- diabetes identified early in pregnancy. Decisions about
cally significant. cesarean delivery are complex and not solely made on esti-
mates of fetal weight, but the proportion of participants
delivered by primary and repeat cesarean and in the pres-
ence and absence of labor were comparable between groups
Discussion in this study.
Among pregnant adults with preexisting type 2 diabetes or
diabetes diagnosed prior to 23 weeks’ gestation, metformin Limitations
added to insulin did not reduce the frequency of the com- Some study limitations exist. First, enrollment was sus-
posite adverse neonatal outcome. Further, no significant pended for as many as 9 months due to the COVID-19 pan-
differences were observed in prespecified secondary out- demic, and the study was halted at 75% for futility, both of
comes, including maternal hypoglycemia and neonatal fat which limit the ability to measure metformin effect on less
mass. Findings were similar regardless of maternal BMI (<30 common outcomes.
or ≥30), timing of diabetes diagnosis, or timing of enroll- Second, while maternal BMI was comparable between
ment (<18 weeks’ gestation or ≥18 weeks’ gestation). How- groups, there were more participants with obesity in the met-
ever, compared with those randomized to placebo, partici- formin group compared with the placebo group, which could
pants randomized to metformin were less likely to deliver skew our findings toward the null because of increased insu-
a large-for-gestational-age infant. lin resistance.
To our knowledge, this is the largest trial of metformin Third, adherence was only measured by self-report;
added to insulin to treat preexisting type 2 diabetes or dia- thus, the negative findings could be explained by poor
betes identified early in pregnancy. Study strengths include adherence that was underreported, although using patient
the randomized, blinded, placebo-controlled design and report for adherence is more generalizable than pill counts
enrollment of a diverse study population with a large pro- because it mimics what is done in clinical practice.
portion of patients of Hispanic ethnicity. These strengths Fourth, HbA1c data were only collected if done as part of
contribute to generalizability of our findings to a US preg- clinical care and only in a small percentage of participants
nant population. (≈39%). Given the proportion of untested participants, defini-
The current trial’s finding of no significant reduction in tive conclusions about effects of metformin on glycemic con-
composite neonatal morbidity aligns with that of the MiTy trol cannot be drawn.
trial, a previous randomized clinical trial of insulin plus
metformin or placebo in pregnant participants with dia-
betes.25 However, in contrast to the current study, those
randomized to metformin in the MiTy trial were found to
Conclusions
have improved maternal glycemic control, less maternal Using metformin plus insulin to treat preexisting type 2 or
weight gain, less neonatal adiposity, fewer cesarean deliver- gestational diabetes diagnosed early in pregnancy did not
ies, and more small-for-gestational-age neonates.25 The dif- reduce a composite neonatal adverse outcome. The effect of
ferences in results may be explained by differences in study reduction in odds of a large for gestational age infant
populations. More than 50% of the current study’s partici- observed after adding metformin to insulin warrants further
pants were Hispanic, compared with 2% to 3% in the MiTy investigation.

2188 JAMA December 12, 2023 Volume 330, Number 22 (Reprinted) [Link]

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy Original Investigation Research

ARTICLE INFORMATION Nordisk, Eli Lilly, Sanofi, vTv Therapeutics, Dexcom, 2020;222(5):495 e1-495 e8. doi:10.1016/[Link].2019.
Accepted for Publication: October 20, 2023. and Boehringer Ingelheim outside the submitted 12.021.
work. Dr Marquis reports grants from the Eunice 3. Simmons D, Immanuel J, Hague WM, et al;
Author Affiliations: University of North Carolina at Kennedy Shriver National Institue of Child Health
Chapel Hill School of Medicine (Boggess, Young); TOBOGM Research Group. Treatment of gestational
and Human Development (NICHD) during the diabetes mellitus diagnosed early in pregnancy.
University of North Carolina Gillings School of conduct of the study. Dr Thomas reports grants
Global Public Health Chapel Hill (Valint, Marquis, N Engl J Med. 2023;388(23):2132-2144. doi:10.
from NIH during the conduct of the study. Dr Britt 1056/NEJMoa2214956
Britt); University of Texas Health Houston reports grants from NICHD during the conduct of
McGovern Medical School Houston (Refuerzo); the study. 4. Dunne F, Brydon P, Smith K, Gee H. Pregnancy in
Columbia University Irving Medical Center, New women with type 2 diabetes: 12 years outcome
York, New York (Zork); University of Alabama at Funding/Support: Supported by a grant from the data 1990-2002. Diabet Med. 2003;20(9):734-738.
Birmingham Heersink School of Medicine NICHD (HD86139) and by support from the doi:10.1046/j.1464-5491.2003.01017.x
(Battarbee); University of South Carolina School of University of North Carolina Department of
Obstetrics and Gynecology. 5. Balsells M, García-Patterson A, Solà I, Roqué M,
Medicine Greenville/Prisma Health-Upstate Gich I, Corcoy R. Glibenclamide, metformin, and
(Eichelberger); University of California, San Diego Role of the Funder/Sponsor: Neither the Eunice insulin for the treatment of gestational diabetes:
(Ramos); University of Texas Medical Branch Kennedy Shriver National Institute of Child Health a systematic review and meta-analysis. BMJ. 2015;
Galveston (Olson); University of Pennsylvania and Human Development nor the University of 350:h102. doi:10.1136/bmj.h102
Perelman School of Medicine, Philadelphia North Carolina Department of Obstetrics and
(Durnwald); The Ohio State University College of Gynecology had any role in the design and conduct 6. Melamed N, Chen R, Soiberman U, Ben-Haroush
Medicine and Wexner Medical Center, Columbus of the study; collection, management, analysis, and A, Hod M, Yogev Y. Spontaneous and indicated
(Landon); Baylor College of Medicine and Texas interpretation of the data; preparation, review, or preterm delivery in pregestational diabetes
Children’s Hospital, Houston (Aagaard); University approval of the manuscript; and decision to submit mellitus: etiology and risk factors. Arch Gynecol
of Mississippi Medical Center, Jackson (Wallace); the manuscript for publication. Obstet. 2008;278(2):129-134. doi:10.1007/s00404-
Indiana University School of Medicine, Indianapolis 007-0541-z.
Data Sharing Statement: See Supplement 4.
(Scifres); Rutgers Health/Robert Wood Johnson 7. Knight KM, Pressman EK, Hackney DN,
Medical School, New Brunswick, New Jersey Additional Contributions: MOMPOD Data Safety Thornburg LL. Perinatal outcomes in type 2 diabetic
(Rosen); Temple University Lewis Katz School of Monitoring Board (DSMB): Geeta Swamy, MD, Duke patients compared with non-diabetic patients
Medicine, Philadelphia, Pennsylvania (Mulla); University, Durham, North Carolina (chair); Anne matched by body mass index. J Matern Fetal
Oregon Health & Science University, Portland Lyerly, MD, MA, University of North Carolina at Neonatal Med. 2012;25(6):611-615. doi:10.3109/
(Valent); Ochsner Health, New Orleans, Louisiana Chapel Hill; William Grobman MD, MBA, The Ohio 14767058.2011.587059.
(Longo); RTI International, Research Triangle Park, State University, Columbus; Diane Catellier DrPH,
Research Triangle Institute International, Research 8. ElSayed NA, Aleppo G, Aroda VR, et al; on behalf
North Carolina (Thomas); University of North of the American Diabetes Association. 15.
Carolina at Chapel Hill School of Nursing (Berry). Triangle Park, North Carolina; Richard I.G. Holt,
PhD, University of Southampton Southampton, UK; Management of diabetes in pregnancy: standards
Author Contributions: Dr Boggess had full access and Brenda Poindexter, MD, Emory University, of care in diabetes-2023. Diabetes Care. 2023;46
to all of the data in the study and takes Atlanta, Georgia. Members of the MOMPOD DSMB (suppl 1):S254-S266. doi:10.2337/dc23-S015
responsibility for the integrity of the data and the were compensated for their time. For coordination 9. Hickman MA, McBride R, Boggess KA, Strauss R.
accuracy of the data analysis. among clinical research sites: Amber Ivins, MS, and Metformin compared with insulin in the treatment
Concept and design: Boggess, Refuerzo, Durnwald, Karen Dorman, RN, MS, from the University of of pregnant women with overt diabetes:
Landon, Aagaard, Young, Thomas. North Carolina at Chapel Hill. Dwight Rouse, MD, a randomized controlled trial. Am J Perinatol. 2013;
Acquisition, analysis, or interpretation of data: Brown University, Providence, Rhode Island, for 30(6):483-490. doi:10.1055/s-0032-1326994
Boggess, Valint, Zork, Battarbee, Eichelberger, serving as medical monitor. For primary outcome
Ramos, Koutrouvelis, Durnwald, Landon, Aagaard, 10. Refuerzo JS, Gowen R, Pedroza C, Hutchinson
adjudication from the University of North Carolina M, Blackwell SC, Ramin S. A pilot randomized,
Wallace, Scifres, Rosen, Mulla, Valent, Longo, at Chapel Hill: Celeste Durnwald, MD, Jerrie
Marquis, Britt, Berry. controlled trial of metformin versus insulin in
Refuerzo, MD, Ashley Battarbee, MD, MSCR, Todd women with type 2 diabetes mellitus during
Drafting of the manuscript: Boggess, Valint, Rosen, MD, Sandy Ramos, MD, John Thorp Jr, MD,
Durnwald, Aagaard. pregnancy. Am J Perinatol. 2015;30(2):163-170.
Jacqueline Patterson MD, MPH, Kimberly Brownley, doi:10.1055/s-0034-1378144
Critical review of the manuscript for important PhD, Victoria Bae-Jump, MD, PhD, Andrea
intellectual content: Boggess, Valint, Refuerzo, Zork, Trembath, MD, MPH, William Goodnight MD, MSCR; 11. Ainuddin JA, Karim N, Zaheer S, Ali SS,
Battarbee, Eichelberger, Ramos, Koutrouvelis, from the University of Utah, Salt Lake City: Marcela Hasan AA. Metformin treatment in type 2 diabetes
Durnwald, Landon, Aagaard, Wallace, Scifres, Smid, MD; from WakeMed Health, Raleigh, North in pregnancy: an active controlled, parallel-group,
Rosen, Mulla, Valent, Longo, Young, Marquis, Carolina: Carmen Beamon, MD; from Baylor randomized, open label study in patients with type
Thomas, Britt, Berry. University, Houston, Texas: Kenneth J. Moise, MD; 2 diabetes in pregnancy. J Diabetes Res. 2015;2015:
Statistical analysis: Valint, Marquis. and from Wake Forest Baptist Health, Winston 325851. doi:10.1155/2015/325851
Obtained funding: Boggess, Durnwald, Aagaard, Salem, North Carolina: Chad Grotegut, MD, MBA. 12. American College of Obstetricians and
Thomas, Berry. We thank all patients, faculty, and fellows at Gynecologists’ Committee on Practice
Administrative, technical, or material support: participating sites. None of the other individuals Bulletins—Obstetrics. ACOG practice bulletin
Boggess, Valint, Refuerzo, Zork, Eichelberger, listed in this section were compensated for their No. 201: pregestational diabetes mellitus. Obstet
Ramos, Koutrouvelis, Durnwald, Landon, Rosen, contributions to this article. Gynecol. 2018;132(6):e228-e248. doi:10.1097/AOG.
Mulla, Young, Marquis, Britt. 0000000000002960
Supervision: Battarbee, Koutrouvelis, Durnwald, Additional Information: Coauthor Diane Berry,
Landon, Aagaard, Wallace, Rosen, Mulla, Marquis, MD, died March 9, 2022. 13. American Diabetes Association Professional
Thomas, Britt. Practice Committee. 15. Management of diabetes in
Other - participation in the trial: Longo. REFERENCES pregnancy: standards of medical care in
1. Centers for Disease Control and Prevention. diabetes-2022. Diabetes Care. 2022;45(suppl 1):
Conflict of Interest Disclosures: Dr Wallace S232-S243. doi:10.2337/dc22-S015
reports grants from National Institutes of Health National Diabetes Statistics Report website.
(NIH) during the conduct of the study. Dr Scifres Accessed June 1, 2023. [Link] 14. Berry DC, Thomas SD, Dorman KF, et al.
reports grants from NIH during the conduct of the diabetes/data/statistics-report/[Link] Rationale, design, and methods for the Medical
study. Dr Longo reports grants from the University 2. Harper LM, Jauk V, Longo S, Biggio JR, Optimization and Management of Pregnancies with
of North Carolina and NIH during the conduct of the Szychowski JM, Tita AT. Early gestational diabetes Overt Type 2 Diabetes (MOMPOD) study. BMC
study. Dr Young reports grants from NIH during the screening in obese women: a randomized Pregnancy Childbirth. 2018;18(1):488. doi:10.1186/
conduct of the study; and grants from Novo controlled trial. Am J Obstet Gynecol. s12884-018-2108-3

[Link] (Reprinted) JAMA December 12, 2023 Volume 330, Number 22 2189

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025
 Research Original Investigation Metformin Plus Insulin for Preexisting Diabetes or Gestational Diabetes in Early Pregnancy

15. Metzger BE, Lowe LP, Dyer AR, et al; HAPO 19. Matts JP, Lachin JM. Properties of obstetrician/gynecologists should know. Curr Opin
Study Cooperative Research Group. Hyperglycemia permuted-block randomization in clinical trials. Obstet Gynecol. 2009;21(6):521-526. doi:10.1097/
and adverse pregnancy outcomes. N Engl J Med. Control Clin Trials. 1988;9(4):327-344. doi:10.1016/ GCO.0b013e328332d24e
2008;358(19):1991-2002. doi:10.1056/ 0197-2456(88)90047-5 24. Bonnet F, Scheen A. Understanding and
NEJMoa0707943 20. Catalano PM, Thomas AJ, Avallone DA, overcoming metformin gastrointestinal intolerance.
16. Carpenter MW, Coustan DR. Criteria for Amini SB. Anthropometric estimation of neonatal Diabetes Obes Metab. 2017;19(4):473-481. doi:10.
screening tests for gestational diabetes. Am J body composition. Am J Obstet Gynecol. 1995;173 1111/dom.12854
Obstet Gynecol. 1982;144(7):768-773. doi:10.1016/ (4):1176-1181. doi:10.1016/0002-9378(95)91348-3 25. Feig DS, Donovan LE, Zinman B, et al; MiTy
0002-9378(82)90349-0 21. Fenton TR, Kim JH. A systematic review and Collaborative Group. Metformin in women with
17. American Diabetes Association Professional meta-analysis to revise the Fenton growth chart for type 2 diabetes in pregnancy (MiTy): a multicentre,
Practice Committee. Summary of revisions: preterm infants. BMC Pediatr. 2013;13:59. international, randomised, placebo-controlled trial.
standards of medical care in diabetes-2022. doi:10.1186/1471-2431-13-59 Lancet Diabetes Endocrinol. 2020;8(10):834-844.
Diabetes Care. 2022;45(suppl 1):S4-S7. doi:10.2337/ 22. van Buuren S. Multiple imputation of discrete doi:10.1016/S2213-8587(20)30310-7
dc22-Srev and continuous data by fully conditional 26. Raygor V, Abbasi F, Lazzeroni LC, et al. Impact
18. ElSayed NA, Aleppo G, Aroda VR, et al; specification. Stat Methods Med Res. 2007;16(3): of race/ethnicity on insulin resistance and
on behalf of the American Diabetes Association. 2. 219-242. doi:10.1177/0962280206074463 hypertriglyceridaemia. Diab Vasc Dis Res. 2019;16
Classification and diagnosis of diabetes: standards 23. Rasmussen KM, Catalano PM, Yaktine AL. New (2):153-159. doi:10.1177/1479164118813890
of care in diabetes-2023. Diabetes Care. 2023;46 guidelines for weight gain during pregnancy: what
(suppl 1):S19-S40. doi:10.2337/dc23-S002

2190 JAMA December 12, 2023 Volume 330, Number 22 (Reprinted) [Link]

© 2023 American Medical Association. All rights reserved.

Downloaded from [Link] by Universiti Malaya user on 05/05/2025