Comparative effectiveness trial of metformin versus insulin for the treatment of gestational diabetes in the USA: clinical trial protocol for the multicentre DECIDE study

Abstract

Introduction

Gestational diabetes mellitus (GDM) is one of the most common medical complications of pregnancy. Glycaemic control decreases the risk of adverse pregnancy outcomes for the affected pregnant individual and the infant exposed in utero. One in four individuals with GDM will require pharmacotherapy to achieve glycaemic control. Injectable insulin has been the mainstay of pharmacotherapy. Oral metformin is an alternative option increasingly used in clinical practice. Both insulin and metformin reduce the risk of adverse pregnancy outcomes, but comparative effectiveness data from a well-characterised, adequately powered study of a diverse US population remain lacking. Because metformin crosses the placenta, long-term safety data, in particular, the risk of childhood obesity, from exposed children are also needed. In addition, the patient-reported experiences of individuals with GDM requiring pharmacotherapy remain to be characterised, including barriers to and facilitators of metformin versus insulin use.

Methods and analysis

In a two-arm open-label, pragmatic comparative effectiveness randomised controlled trial, we will determine if metformin is not inferior to insulin in reducing adverse pregnancy outcomes, is comparably safe for exposed individuals and children, and if patient-reported factors, including facilitators of and barriers to use, differ between metformin and insulin. We plan to recruit 1572 pregnant individuals with GDM who need pharmacotherapy at 20 US sites using consistent diagnostic and treatment criteria for oral metformin versus injectable insulin and follow them and their children through delivery to 2 years post partum. More information is available at www.decidestudy.org.

Ethics and dissemination

The Institutional Review Board at The Ohio State University approved this study (IRB: 2024H0193; date: 7 December 2024). We plan to submit manuscripts describing the results of each study aim, including the pregnancy outcomes, the 2-year follow-up outcomes, and mixed-methods assessment of patient experiences for publication in peer-reviewed journals and presentations at international scientific meetings.

Trial registration number

NCT06445946.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• DECIDE (www.decidestudy.org) is a patient-centred and pragmatic comparative-effectiveness randomised control trial that will compare oral metformin versus injectable insulin for the prevention of adverse pregnancy outcomes and the safety of postpartum outcomes among pregnant individuals with gestational diabetes mellitus who require pharmacotherapy and for their exposed children.

• Strengths of the DECIDE trial include a non-inferiority clinical trial design, assessment of postpartum outcomes to confirm safety, integration of patient-reported outcomes and inclusion of a racially, ethnically and geographically diverse population.

• Limitations of the DECIDE trial include no follow-up beyond 2 years post partum and assessment of participant and infant anthropometry and adiposity by physical exam instead of imaging.

• Challenges of this trial will include recruitment across 20 US sites and postpartum retention.

Introduction

Background

Gestational diabetes mellitus (GDM) is one of the most common medical complications of pregnancy and affects nearly 400 000 or ~1 in 10 pregnant individuals in the USA every year.1,3 The incidence of GDM has more than doubled in the past decade in an environment of rising prevalence of advanced reproductive age and obesity.^{4 5} Moreover, GDM has risen inequitably among racially and ethnically minoritised and lower-income individuals.^{2 6} More than one in four infants born to individuals with GDM will experience an adverse neonatal outcome, such as large-for-gestational age (LGA) birth weight, hypoglycaemia or hyperbilirubinaemia.7,9 After delivery, individuals with prior GDM are at >10 fold increased risk of diabetes, and infants exposed to GDM are at 2-fold increased risk of obesity.^{10 11}

The goal of GDM treatment is to achieve optimal glycaemic control and prevent adverse pregnancy outcomes.^{9 12} The initial therapeutic approach is dietary modification and regular exercise,^{12 13} but >1 in 4 individuals will not achieve glucose control with these interventions.13,15 When pharmacological treatment is needed, guidelines from the American Diabetes Association (ADA) and American College of Obstetricians and Gynecologists (ACOG) recommend insulin as the first-line medication,^{14 16} while the Society for Maternal-Fetal Medicine states both insulin and metformin are reasonable medication options.¹⁷

In the past, insulin has been the first-line option because it provides glycaemic control, improves pregnancy outcomes and does not cross the placenta.^{7 18 19} An alternative to insulin is metformin, which also provides glycaemic control and improves pregnancy outcomes (table 1).18,20

Table 1. Advantages versus disadvantages of metformin versus insulin.

	Metformin	Insulin
Advantages	Convenience of an oral pill Lower cost/less resources No participant hypoglycaemia Less weight gain Improved adherence	Historically, first-line agent Highly effective Does not cross the placenta Confirmed fetal safety
Disadvantages	Crosses the placenta Possible low birth weight Possible risk of child obesity Gastrointestinal distress Risk of supplemental insulin	Injections/inconvenient Higher cost/high resources Participant hypoglycaemia Impractical for short-term use

Patients and providers may prefer metformin to insulin because it is convenient to take as an oral pill, well-tolerated, cheaper and practical when medication is needed for a brief time period. Additionally, metformin does not cause hypoglycaemia and reduces gestational weight gain. Metformin use is increasing in clinical practice—while insulin remains the most common medication for GDM, one in three individuals with GDM in the USA were prescribed metformin by late 2018.^{21 22} Yet metformin has limitations, including that more than one in four individuals will ultimately need supplemental insulin to achieve glucose control, and there is known placental transfer. Historically, another oral agent, glyburide, had been used, but guidelines have since advised against its use following trials that showed it did not appear to be efficacious.^{18 23}

Follow-up data on metformin from individuals with prior GDM and their exposed children are limited.²⁴ Extant data suggest that children exposed to metformin had similar body fat composition but slightly higher body mass index (BMI) compared with those exposed to insulin^{25 26}; but recent population-based data show no difference in BMI.²⁷ However, these studies were limited due to inadequate randomised controlled trial (RCT) follow-up and observed heterogeneity in the effect across different study sites. Also, whether participant metabolic health postpartum varies based on prior metformin versus insulin exposure in pregnancy requires further study.²⁸

Deciding between metformin and insulin can be challenging for patients and providers given variation in treatment guidelines, provider recommendations and lack of conclusive comparative efficacy and safety data.²⁹ Understanding whether patients take medications as directed, how satisfied they are with their medication decision, and how their medication decision impacts their pregnancy experience may help to explain observed heterogeneity of treatment effects (HTE).³⁰ Patient perspectives on barriers to and facilitators of metformin versus insulin use may identify opportunities to improve outcomes.³¹

DECIDE: A Comparative Effectiveness Trial of Oral Metformin vs Injectable Insulin for the Treatment of Gestational Diabetes is a randomised, patient-centred, open-label and pragmatic comparative effectiveness trial in pregnancy with postpartum follow-up. This protocol is written in accordance with the Standard Protocol Items: Recommendations for Interventional Trials 2013 statement.³²

Aims and hypotheses

Primary aims

Aim 1: To evaluate whether pregnant individuals randomised to metformin are not inferior to pregnant individuals randomised to insulin for the composite adverse neonatal outcome (LGA birth weight, hypoglycaemia, hyperbilirubinemia or death).

Aim 2: To evaluate whether mean BMI at 2 years of age is higher in the offspring of pregnant individuals randomised to metformin.

Aim 3: To understand facilitators and barriers associated with metformin versus insulin use and HTE to facilitate evidence-based pharmacotherapy.

Primary hypotheses

Aim 1: We hypothesise that metformin is not inferior or worse than insulin by an absolute margin or difference of more than 8% in the composite adverse neonatal outcome.

Aim 2: We hypothesise that metformin does not result in increased child BMI at 2 years (not inferior by an absolute margin of 0.31 kg/m²) compared with insulin.

Aim 3: We hypothesise that patient-reported factors associated with metformin compared with insulin use will be different, which is important to identify to enable clinical implementation of study findings.

Secondary aims

We will compare outcomes at delivery between pregnant individuals randomised to metformin versus insulin (hypertensive disorder of pregnancy, gestational weight gain, mode of delivery and obstetric anal sphincter injuries) and their infants (preterm birth, mechanical ventilation, neonatal intensive care unit (NICU) admission, oxygen support, respiratory distress syndrome and small-for-gestational-age at birth); as well as the frequency of treatment supplementation with insulin among pregnant individuals randomised to metformin.

We will compare outcomes at 2 years post partum between individuals randomised to metformin versus insulin (obesity, anthropometry, adiposity, diabetes, cholesterol and hypertension) and their children (obesity, anthropometry and adiposity).

We will compare patient-reported outcomes (PROs) at randomisation (mental and physical health; Diabetes Knowledge Questionnaire (DKQ), Diabetes Distress Scale (DDS) and Diabetes Management Self-Efficacy Scale (DMSES); lifestyle; health behaviours and diet), and at 6 weeks and 2 years post partum for the individual (pregnancy and childbirth experiences; treatment adherence and satisfaction; Maternal-Infant Bonding Scale (MIBS); lactation practices; lifestyle; health behaviours and diet) and child (lifestyle; health behaviours and diet).

Methods and analysis

Design

DECIDE is a randomised, controlled, open-label, patient-centred and pragmatic multicentre comparative effectiveness trial that is designed to determine whether metformin is not inferior to insulin in reducing adverse pregnancy outcomes and is comparably safe for exposed pregnant individuals and children and to identify patient-reported factors associated with metformin versus insulin that facilitate and enable implementation of study findings (online supplemental file 1).

The DECIDE Study Consortium includes 20 clinical sites under a clinical coordinating centre (CCC) and an independent data coordinating centre (DCC). The consortium is governed by a steering committee and guided by a patient advisory board and stakeholder engagement group. Data management, coordination and analysis will be completed by the DCC, led by the study statisticians (CM and RGC). Participant data will be collected, stored and maintained in OpenClinica, a browser-independent electronic data capture system. Enrolled individuals will be randomised in a 1:1 ratio of metformin to insulin within the web-based data management system according to a computer-generated permuted block design with variable block sizes. Randomisation will be stratified by study site.

Population

Individuals will be recruited across 20 US clinical sites with diabetes and prenatal care programmes (figure 1). These sites have been selected with the goal of achieving racial and ethnic, urban and rural, and geographical diversity at both academic and community-based medical centres. Individuals who continue to receive routine prenatal care in their local community, and then receive high-risk prenatal and diabetes care from the clinical site will also be eligible for study participation. After delivery, individuals and their infants will be followed up with data ascertainment at 6 weeks and 2 years post partum.

Figure 1. Geographic distribution of DECIDE sites across the USA. CCC, clinical coordinating centre; DCC, data coordinating centre.

Inclusion criteria

Inclusion criteria are age >18 years, singleton pregnancy, gestational age between 20^0/7 and 31^6/7 weeks, GDM diagnosis between 20^0/7 and 31^6/7 weeks, requiring medication for glycaemic control, and willingness and ability to attend 2-year follow-up visit (table 2). The decision to initiate medication will be consistent with current US recommendations, defined as ≥30% elevation of either fasting or 1-hour or 2-hour postprandial glucose values in the prior week.

Table 2. Inclusion and exclusion criteria.

Inclusion	Exclusion
Age >18 years	Renal disease
Singleton pregnancy	Chromosomal, genetic or major malformation.
Gestational age 200/7–316/7 weeks	Known fetal aneuploidy based on invasive testing or positive for aneuploidy on cell-free fetal DNA screening.
GDM diagnosis 200/7–316/7 weeks	Contraindication to metformin or insulin
Needs medication (≥30% elevated fasting or postprandial glucose values in the prior week)	Diabetes or GDM <20 weeks or A1c >6.5%
Willingness and ability to attend 2-year follow-up visit	Fasting hyperglycaemia >115 mg/dL (≥50% fasting glucose values in past week)
	Enrolled in a trial that influences the primary study outcome
	Planned delivery at a site without access to the medical record
	Language barrier (appropriate translation resources unavailable at the site)
	Participation in this trial in a previous pregnancy

GDMgestational diabetes mellitus

Exclusion criteria

We will exclude individuals who have known underlying chronic kidney disease; a fetus with a chromosomal, genetic or major structural malformation; contraindication to metformin or insulin; pregestational diabetes (either type 1 or 2); early-onset GDM <20 weeks; prior haemoglobin A1c >6.5%; concurrent enrollment in a trial with a primary aim that influences the primary study outcome; planned delivery at an outside clinical site where access to medical records cannot be obtained for outcome data abstraction; language barrier (appropriate translation resources unavailable at the site); participation in this trial in a previous pregnancy and fasting hyperglycaemia defined as >115 mg/dL for ≥50% glucose values in the past week (table 2). We include fasting hyperglycaemia as an exclusion criterion as prior data suggest that individuals with this finding are likely to require insulin to achieve glycaemic control.²⁰

Recruitment

The start date for recruitment is 1 August 2024 and is anticipated to end by 1 May 2026, with final data collection at the 2-year follow-up ending on 1 May 2028. All individuals who present for prenatal care at sites in the DECIDE Study Consortium will be screened for eligibility. Individuals who meet study criteria will be approached for participation by study staff, which will include a study pamphlet with a weblink and QR code (www.decidestudy.org) and a 3 min video about GDM medication management (https://youtu.be/CGmYCmF4vDo). After eligibility is confirmed, individuals will be asked to participate after study information is given. Individuals who agree will complete the written informed consent process (see online supplemental file 2 for sample consent document). Reasons for ineligibility and rates of declining to participate will be collected. The patient advisory board will review recruitment and retention materials to create participant-friendly information and to assist with provider trainings.³³

Pregnancy (aim 1)

Baseline visit

A research team member or healthcare provider will ask patients if they are interested in joining the study either in person or virtually. Those who are interested will be given an orientation to the study by a research team member (figure 2). Written informed consent will be obtained in English or Spanish. Enrolled individuals will be randomised in a 1:1 ratio to metformin or insulin. Study staff will inform the provider about randomisation arm via telephone, email and the electronic medical record. Because this is an open label, non-blinded pragmatic trial comparing two treatments that are standard of care, individuals allocated to either arm will obtain their medication from their preferred pharmacy with a prescription from their provider, which will account for brand of insulin on formularies of their insurance plans.

Figure 2. Flow diagram of DECIDE study events.

Using defined data fields, we will record participant demographics, medical history and obstetric characteristics. Participants will complete standardised surveys at randomisation to assess lifestyle, health behaviours, diet, mental and physical health during pregnancy, DKQ, DDS and DMSES.

Follow-up visits

Consistent with a pragmatic trial, all dosing changes will be performed by the participant’s provider. The frequency of participant clinical encounters will be about every 2 weeks, which is standard clinical practice for GDM management,³⁴ with virtual or in-person clinic visits at the discretion of the provider. Study staff will visit with participants monthly (preferably in person and otherwise virtually), ask them about adherence to assigned treatment, review side effects related to their medication including nausea and symptoms of hypoglycaemia and assess medication adherence. Additionally, study staff will review and abstract the following information from the participant’s medical record: capillary blood glucose log values or continuous glucose monitoring logs for the past 1 week period, current metformin and insulin dosing, type(s) of insulin, and gestational weight gain. Finally, study staff will assess for adverse events (AEs) and serious AEs (SAEs) at each study visit.

Delivery

The assigned treatment (insulin vs metformin) will be discontinued at delivery. The provider will have the responsibility for intrapartum and postpartum management. Data on intrapartum and postpartum glycaemic management will be abstracted by the study team.³⁵ Additionally, participant and infant data will be collected from the EHR until hospital discharge. We will collect comprehensive antepartum and intrapartum data, such as labour and delivery details; glycaemic control and neonatal outcomes. To address concerns about placental transfer of medication and fetal safety, cord blood and placental samples will be collected when possible for further analyses.

Postpartum follow-up through 2 years (aim 2)

~6 weeks post partum

At ~6–8 weeks post partum, participants will complete standardised surveys with the study team in person, by a virtual platform, or by mail, per participant and site preference. This visit will include standardised measures to assess treatment adherence and satisfaction, lactation, lifestyle, health behaviours, diet, pregnancy and childbirth experience, and MIBS. Additional participant and child postpartum data through the ~6 weeks postpartum visit will be collected from the EHR. Testing for diabetes at the postpartum visit is standard of care. We will collect these results and emphasise best practices to increase uptake of diabetes screening.³⁶ Research staff will actively maintain contact with participants every 6 months after delivery by telephone, email or post.

2-years post partum

At or after 2 years post partum, participants will be invited to return for an in-person assessment and physical exam of both the participant and child. Participants will complete standardised surveys to assess lifestyle, health behaviours and diet in the mother and child. Physical exam of the child will include growth (weight and height), anthropometry (arm and abdominal circumference) and adiposity (skinfolds) measurements. For the participant, blood pressure, height, weight, anthropometry (waist and hip circumferences) and adiposity (skinfolds) will be obtained. The assessor at the visit will be masked to study arm assignment. A calibrated scale with stadiometer will be used for weight and height, a tape measure for anthropometry and callipers for skinfold measurements. As has been done in prior GDM cohorts (co-I PC), all sites will undergo central training and assessment in these techniques to promote standardisation of measurements and adherence to the study protocol.^{37 38} Ongoing training and strict monitoring of the study team measurement techniques will be performed and regular assessment of interobserver variation will be conducted via training videos every 6 months. For the postpartum individual, blood will be obtained using standard venipuncture techniques, including for haemoglobin A1c, cholesterol panel and 2-hour 75 g oral glucose tolerance test using standardised collection procedures.³⁹

Mixed-methods assessment (aim 3)

In years 2–4 of the study, 150 individuals across all sites will be invited to complete a 30–45 min interview approximately 6 weeks after delivery. Individuals will be purposefully recruited from each study site and across each year of the study to ensure diversity with respect to race and ethnicity, age, insurance status and study arm. Participants will be given the option to be interviewed by phone or via video (eg, zoom).

A semistructured interview guide has been developed and refined based on feedback from patients and experts on qualitative research methods (online supplemental file 2). These interviews will be performed centrally at OSU under the supervision of co-I ASM. The guide will include open-ended questions in the following domains: DECIDE trial participation, GDM and pregnancy, medications to treat GDM, experiences taking medication to treat GDM, and GDM and postpartum health.³¹ During the interviews, participants will be asked to describe their experiences using metformin and/or insulin with question probes to address specific aspects of their experiences, including barriers to and facilitators of metformin and insulin use, and the factors that might improve adherence and pregnancy experience. The draft guide will be pilot tested and finalised prior to use in the study. All interviews will be audio recorded and transcribed verbatim to allow rigorous qualitative analysis.

Patient and public involvement

Patient and stakeholder interactions have helped develop the research question, comparators, study participants characteristics and relevant outcomes.³¹ The DECIDE investigator team will engage key opinion leaders, patients with a history of GDM or living with diabetes and stakeholders, including public and private-sector insurers, advocacy organisations and professional societies, to elicit feedback through the Patient Advisory Board (co-I SC) and Stakeholder Engagement Board (co-I AT), both of which were established for this study and DiabetesSisters (co-I DR), a national patient advocacy organisation dedicated to women living with diabetes. The team will draw on participatory learning approaches, such as adaptive management, rapid assessments, data-driven decision-making and human-centred design to codevelop recommendations to inform project improvement.

Measures

The outcomes in this study include both clinical outcomes and PROs that cumulatively measure obstetric and perinatal morbidity and mortality that impact quality of life, well-being and pregnancy experience. Clinical outcomes relevant to contemporary practice are based on prior GDM pharmacotherapy trials,^{20 23 40} meta-analyses comparing treatment strategies for GDM to prevent adverse outcomes,^{18 19 26} engagement with stakeholder organisations and providers,⁴¹ and core outcome sets for GDM that used a Delphi methodology.^{42 43} PROs⁴⁴ are based on validated and standardised instruments that address birth experiences, living with diabetes, and treatment experiences and adherence; systematic reviews of prior qualitative studies of patient experiences with GDM; and testimonials from affected individuals and their families.2930 45,47

Primary clinical outcomes

The primary pregnancy outcome (aim 1) is a neonatal composite adverse outcome of LGA birth weight, hypoglycaemia, hyperbilirubinaemia and/or death (table 3).⁴⁸ This measure is based on neonatal outcomes causally related to glycaemic control and consistent with that used in recent trials²³ and meta-analyses.¹⁹

Table 3. Primary and secondary outcomes.

Primary/secondary	Name of outcome	Specific measure to be used	Time points
Composite primary*	Large-for-gestational-age (child)	Birth weight ≥90th percentile for gestational ageUS birth certificate reference adjusted for parity and/or fetal sex48	Delivery
Composite primary*	Hypoglycaemia (child)	Blood glucose <35 mg/dL OR treatment <24 hours after birth with either IV, PO, or gel glucose therapy	<24 hours of birth
Composite primary*	Hyperbilirubinaemia (child)	Treated with phototherapy or exchange transfusion and treatment in the first postnatal week or kernicterus	Delivery to first postnatal week
Composite primary*	Fetal or neonatal death (child)	Death after randomisation and before hospital discharge or 30 days postnatal age if still hospitalised (excluding voluntary pregnancy termination)
Primary*	BMI (child)	kg/m2, continuous measureMeasurement using standardised scale and protocolUS CDC reference adjusted for child sex	2 years follow-up
Secondary	Hypertensive disorder of pregnancy (participant)	Diagnosis per ACOG criteria	Randomisation to delivery
Secondary	Gestational weight gain (participant)	Per Institute of Medicine guidelines (excess, within and less than weight gain recommendations and standardised z-scores)	Randomisation to delivery
Secondary	Preterm birth (child)	<37 weeks, ACOG best OB estimate	Delivery
Secondary	Small-for-gestational-age (child)	Birthweight <10th percentile for gestational ageUS birth certificate reference adjusted for parity and/or fetal sex48	Delivery
Secondary	Mode of delivery (participant)	Caesarean delivery, including primary vs repeat and by indication	Delivery
Secondary	Obstetric anal sphincter injuries (participant)	Per ACOG guidelines	Delivery
Secondary	Mechanical ventilation (child)	Requiring mechanical ventilation, by type and duration	<72 hours birth
Secondary	NICU admission (child)	Admitted to NICU or intermediate nursery	Discharge
Secondary	Oxygen support (child)	Requiring oxygen support, by type and duration	Discharge
Secondary	Respiratory distress syndrome (child)	Signs of respiratory distress on the basis of clinical features and oxygen or respiratory support for ≥6 hours	<72 hours birth
Secondary	Overweight, obesity (child)	BMI ≥85th percentile, BMI ≥95th percentile for age/sex.10	2 years follow-up
Secondary	Adiposity (child)	Triceps/subscapular skinfold thickness >90th percentile for age/sex; individual and sum of measures76	2 years follow-up
Secondary	Anthropometry (child)	Abdominal circumference; age-adjusted/sex-adjusted WHO z-scores for arm circumference.25	2 years follow-up
Secondary	Anthropometry (participant)	Waist and hip circumference, and ratio.	2 years follow-up
Secondary	BMI/obesity (participant)	Continuous, dichotomous (BMI ≥30 kg/m2).	2 years follow-up
Secondary	Adiposity (participant)	Thigh, triceps, subscapular, suprailiac skinfolds.	2 years follow-up
Secondary	Type 2 diabetes (participant)	A1c >6.5% OR fasting plasma glucose >126 mg/dL OR OGTT >200 mg/dL OR prior diagnosis per patient report	2 years follow-up
Secondary	Pre-diabetes (participant)	A1c 5.7%–6.4% OR fasting plasma glucose 100–125 mg/dL OR OGTT 140–199 mg/dL	Year follow-up
Secondary	Hypertension (participant)	Blood pressure (≥130 mm Hg systolic or ≥80 mm Hg diastolic at study visit), EHR or medication	2 years follow-up
Secondary	Treatment adherence/satisfaction (PRO)	Treatment Satisfaction Questionnaire for Medication58; Acceptability of Treatment20	6 weeks follow-up

^*The sample size of n=1572 provides 90% power, one-sided significance level of 0.025, to detect a non-inferiority margin of 8% of the composite RCT outcome (aim 1). For the follow-up (aim 2), a sample size of n=1415 provides 90% power to rule out an effect size of at least 0.172 (ie, 0.172 SD) or a 0.29 kg mean difference in child weight.

ACOGAmerican College of Obstetricians and GynecologistsBMIbody mass indexNICUneonatal intensive care unitOGTToral glucose tolerance testPROpatient-reported outcomeRCTrandomised controlled trial

In the follow-up at 2 years, the primary child outcome is BMI as a continuous measure, which is consistent with prior GDM RCT follow-up studies and meta-analyses.^{18 19 40} An updated sex-adjusted US reference will be used for standardisation of height and weight for age.⁴⁹ Anthropometry will be measured with standardised protocols that have been successfully implemented in prior GDM cohorts at birth (co-I PC).^{37 38 50}

Secondary clinical outcomes

Secondary pregnancy outcomes (table 3) include neonatal outcomes (preterm birth, small-for-gestational-age, NICU admission, mechanical ventilation by duration, oxygen support by type and duration, and respiratory distress syndrome by clinical features and oxygen or respiratory support for any time during the first 72 hours after birth and participant outcomes (hypertensive disorder of pregnancy, mode of delivery, total gestational weight gain and obstetric anal sphincter injuries).

In the follow-up at 2 years, secondary child outcomes include obesity and measures of adiposity and anthropometry. Secondary participant outcomes at 2 years will include type 2 diabetes, obesity, pre-diabetes, hypertension, metabolic profile, and measures of anthropometry and adiposity.

Patient-reported measures

At randomisation, baseline assessments will include mental and physical health (PROMIS Global Short Form^{51 52}; DKQ,⁵³ DDS,⁵⁴ DMSES⁵⁵ and social determinants of health (Accountable Health Communities Health-Related Social Needs Screening Tool⁵⁶ and Williams Everyday Discrimination Scale⁵⁷ (table 3).

At ~6 weeks postpartum follow-up visit, PROs will include treatment adherence and satisfaction Treatment Satisfaction Questionnaire for Medication⁵⁸ and Acceptability of Treatment²⁰; infant feeding practices (CDC Infant Feeding Practices, selected questions)⁵⁹; pregnancy and childbirth experience (Birth Satisfaction Scale-Revised Indicator)⁶⁰; MIBS⁶¹; International Physical Activity Questionnaire (IPAQ), short-form⁶²; Mini-EAT (Eating Assessment Tool)⁶³ and the Brief Infant Sleep Questionnaire-Revised Short Form (selected questions).⁶⁴

At 2 years, PROs of the postpartum individual will include IPAQ, long-form⁶²; Mini-EAT⁶³; social determinants of health⁵⁶; mental and physical health⁵¹ and of the child will include CDC Child Health and Diet Survey (selected questions)⁶⁵; Movement Behaviour Questionnaire (selected questions)⁶⁶; Brief Infant Sleep Questionnaire-Revised Short Form (selected questions)⁶⁴ and the Child Eating Behaviour Questionnaire.⁶⁷

Procedures

Pharmacotherapy management

Study guidelines for metformin and insulin management including initiation, dosing, titration and monitoring have been developed based on current clinical guidelines.⁶⁸ Pharmacotherapy will be initiated when ≥30% of fasting and/or postprandial glucose values are elevated in the past week. Given this is a pragmatic trial, clinical practice may vary slightly across sites based on local standard-of-care and individualised provider–patient decision-making.

Treatment initiation and titration

Metformin (either extended or immediate release) will be started at 500 mg two times per day and titrated to a maximum daily dose of 2500 mg. Participants randomised to metformin will be given uniform advice by study personnel on how to minimise gastrointestinal distress, such as taking study tablets prior to meals and using antiemetics.

Providers will be encouraged to use trimester-specific and weight-based insulin dosing criteria for both basal and prandial insulins for up to a total of 4 daily injections. Consistent with clinical practice, some participants may be managed with a single dose of intermediate-acting or long-acting insulin at night to treat isolated fasting hyperglycaemia, while others may require additional treatment of postprandial hyperglycaemia with shorter-acting insulin. The sites’ insulin formularies may include rapid (Novolog, Humalog and Aspart), intermediate (Humulin N, Novolin N and NPH) and long-acting insulins (detemir, Lantus and degludec) with comparable efficacy in pregnancy. Participant insurance coverage will be considered when selecting insulin type.

We will standardise the incorporation of best practices regarding metformin and insulin titration per ADA and ACOG guidelines. Providers at each site will be instructed on the study protocol and trained on study procedures, including glycaemic monitoring ~1–2 weeks and uptitration. Glucose assessment by those participants electing self-monitored blood glucose monitoring will be performed at fasting and three times postprandial; those who elect continuous glucose monitoring will be asked to similarly document their fasting and postprandial values. Adherence to these goals will be monitored by research staff monthly from participant interviews and medical record review. Concerns regarding protocol adherence will be discussed with site PIs. Weekly participant glucose logs and total metformin and insulin doses and type of insulin will be recorded and considered in data analysis.

Treatment supplementation

Participants receiving metformin will have insulin supplemented (ie, addition of insulin to base regimen of metformin) only if they have not achieved euglycaemia for at least 30% of glucose values after approaching the maximum daily dose of metformin (>2000 mg or the maximum tolerated dose). Participants will be asked to continue taking metformin after treatment supplementation with insulin, which is generally the current clinical practice. In rare circumstances (0%–2%) in which severe gastrointestinal distress or intolerable side effects are present with metformin, participants may be prescribed insulin before reaching the maximum daily dose of metformin (2500 mg) or switched to insulin entirely.⁶⁹ Reasons that patients and providers decide on treatment supplementation will be collected.

Data safety and monitoring

An independent data safety monitoring board (DSMB) has been created to provide oversight of trial accrual and of privacy and safety of study participants. DSMB members have appropriate expertise (obstetrics and gynaecology, maternal–fetal medicine, endocrinology, neonatology, bioethics and biostatistics). The DSMB will meet to review the protocol prior to study initiation and then yearly to review study progress. The DCC will provide reports to the DSMB that include recruitment, protocol adherence and safety outcomes.

Detailed information about AEs and SAEs will be collected and evaluated throughout the trial. If a patient develops an SAE, the primary clinician in collaboration with the site PI will ascertain the safety of continuing the intervention. All unanticipated and possibly study-related AEs and SAEs will be reported to the IRB per regulatory reporting guidelines. Metformin may be temporarily stopped in the setting of acute kidney injury or intravenous contrast administration. Metformin has been reported to be very rarely associated with lactic acidosis (<10 cases per 100 000 patient-years), although the validity of this association has been challenged.⁷⁰ We will include lactic acidosis on metformin as a safety stopping rule.

Statistical analysis plan

Sample size and power

Published data suggest that upwards of 30% of individuals with GDM have an associated adverse neonatal outcome. Using data from recent meta-analyses that compared the two treatment regimens,^{18 19 71} and the most recent RCT (although comparing glyburide to insulin) that assessed the same primary composite outcome as in our study,²³ we estimate the frequency of the primary composite perinatal outcome to be 28% with insulin. To be conservative, we have used an estimate of 25%.

We have chosen a non-inferiority trial design because metformin’s advantages in terms of cost and ease (eg, oral, no refrigeration needed, less costly) suggest that metformin may be the preferred first-line treatment for GDM if it were found to be non-inferior to insulin in terms of efficacy and safety.⁴¹ A non-inferiority margin of 8% was selected for the primary outcome based on a survey and interviews we conducted in January to June 2021 with each of our 20 site PIs, all of whom are maternal-fetal medicine specialists, as well as interviews with 144 patients. This conservative margin is also consistent with recent non-inferiority RCTs for GDM.²³ Additionally, we estimate that 20% of individuals who are randomised to metformin will require supplemental insulin,²¹ which is lower than prior trials because we will exclude those with fasting hyperglycaemia (>115 mg/dL for >50% in the prior week) who are at the highest risk of failing metformin.²⁰

Based on the above assumptions, we plan to enrol 1572 individuals to determine if metformin is non-inferior to insulin for the composite primary outcome, with 90% power, one-sided significance level of 0.025, a loss to follow-up at delivery of 2% and 20% supplementation with insulin in addition to metformin.

For the 2-year follow-up, if outcomes are obtained on 1415 participants (ie, a loss to follow-up rate of 10%), there will be 90% power to rule out an effect size of at least 0.172 SD. This translates to a 0.31 unit difference in BMI or a 0.29 kg mean difference in child weight.²⁵ There will be 80% power to rule out an effect size of at least 0.149 SD, or a 0.27 unit difference in BMI or a 0.25 kg mean difference.

Analyses for pregnancy outcomes (aim 1)

We will use descriptive statistics to characterise participants to determine comparability of treatment groups at baseline. As an intention-to-treat analysis, the comparison is between individuals randomised to start metformin regardless of whether they later required supplemental insulin or stopped metformin due to side effects and switched to insulin versus individuals randomised to start on insulin. Analyses of the primary outcome will consist of summarising the proportions of trial participants with the primary endpoint for each group and calculating the corresponding between-group risk difference (insulin minus metformin) with 95% CIs.

Data analyses will adhere to the CONSORT (Consolidated Standards of Reporting Trials) guidelines and follow the intention-to-treat principle in which patients are analysed in the group to which they were randomised, regardless of whether they received the assigned intervention or altered their assigned medication prior to delivery. Metformin will be determined as non-inferior if the lower 95% confidence limit for the risk difference is −8 percentage points or greater (ie, closer to 0). If treatment groups differ on a pretreatment factor known to be a risk factor for the outcome, the analysis will adjust for these differences and an adjusted risk difference will be reported. If metformin is determined to be non-inferior to insulin, a superiority test will be conducted without adjusting the type I error, with metformin considered superior if the lower 95% confidence limit for the risk difference is more than 0.

Interim analyses

Since the sample size estimate is based on the assumption that the primary endpoint rate will be 25% in the insulin group, it is important to evaluate this proportion in the study after 20% of the participants (N=315) have delivered. In addition, the proportion of patients in the metformin group who require supplemental insulin will be reported. Once 50% of the participants have delivered (N=786), a formal interim analysis will be performed to determine whether metformin is inferior to insulin, with an upper boundary for the stopping rule for harm based on a one-sided type I error of 0.025 and the Lan-DeMets generalisation of the O’Brien-Fleming boundary. If the upper confidence bound for the risk difference is less than 0, the DSMB will evaluate this in the context of the other safety outcomes. We also plan to calculate conditional power given the observed data and conditional on future data showing no difference between treatment strategies. If the conditional power is high (>90%) that the neonatal composite rate will be more that 8% higher in the metformin arm, the DSMB will consider termination for futility, although any decision to terminate the study would not be reached solely on statistical grounds but on a number of clinical and statistical considerations.

Analyses for postpartum follow-up through 2 years (aim 2)

Child BMI is the primary outcome at 2 years of age. Analyses will consist of summarising the mean BMI standardised for age and sex for each group and calculating the corresponding between-group mean difference with 95% CIs using generalised linear models. Metformin will be determined as non-inferior to insulin if the lower 95% confidence limit for the mean difference is 0.31 units or greater (ie, closer to 0). Additional analyses as detailed above for the primary neonatal composite in the RCT will be performed, including for measures of child adiposity and anthropometry. Fetal sex will be evaluated for predefined interaction analyses with treatment group, and anthropometry will be standardised by sex-specific standards.⁴⁸

Mixed-methods analyses (aim 3)

We will use the constant comparative method and a grounded theory approach to analyse interview data.⁷² This iterative approach to analysis will include reading interview transcripts and discussing findings among investigators as the study progresses. Our approach will enable exploration of emergent themes and ensure saturation in data collection. Analysis will prioritise the elucidation of key concepts from individuals’ interview statements (extraction), conceptual development based on constant comparative analysis, and classification of data through code development.^{72 73} The coding team (co-I ASM) will create a preliminary coding dictionary based on the interview guide, defining broad categories of findings to enable coding of responses to interview questions. Frequent discussions among coding team members will allow the characterisation of emergent codes and ensure agreement about identified themes and subthemes. ATLAS.ti software will be used to support the analysis process.

Subgroup analyses

Treatment effectiveness for subgroups may differ due to barriers related to social determinants of health (eg, race/ethnicity), bioavailability of medication, physiologic insulin resistance (eg, BMI) or factors related to GDM and its severity (eg, maternal age, gestational age at medication initiation) (online supplemental file 3). We will employ existing rigorous checklist for addressing the design, analysis and context of subgroup analyses.⁷⁴ These risk factors were selected based on differences in the frequency of GDM and adverse pregnancy and postpartum outcomes, and hence, at least a theoretical possibility as to why HTE may exist. We will formally assess for effect modification (interaction effect). Should we note significant heterogeneity of treatment effect across these prespecified groups (p<0.05), we will then systematically examine two-way effect modification. Should there be evidence of HTE, the proposed exploratory subgroup analyses will employ a non-inferiority approach consistent with the overall trial design and analysis plan.

Missing data and sensitivity analyses

We will investigate the robustness of the observed differences between the two groups with respect to any missing data. First, an inverse probability weighting (IPW) analysis will be conducted with each case weighted by the inverse probability of being a complete case. Under a missing-at-random mechanism, the IPW approach would result in an unbiased estimate of the difference between groups assuming a correctly specified model for the missing data. Second, a tipping-point analysis will describe the additional number of events in the insulin group versus the metformin group among the participants with missing data that would change the conclusion related to non-inferiority. In addition, a sensitivity analysis will be performed among participants in the metformin group who did not require supplemental insulin versus participants randomised to insulin only.

Retention

Participants will be asked to provide contact information (eg, phone, email and address) for themselves and two relatives who would know how to contact them. Research staff will actively maintain contact with participants throughout their pregnancies and by telephone, email or post, every 6 months after delivery. Participants will be asked to verify or update information at each contact. We will also maintain contact with participants and their families through flyers, cards and electronic communications in order to provide study updates.

Compensation

Participant reimbursement will be provided for completing assessments at multiple time points: randomisation (US$100), 6 weeks post partum (in person, virtual and/or telephone) (US$50) and 2-year follow-up visits for the participant and child (in person) (US$125). Participants selected for qualitative interviews will receive additional compensation (US$100).

Ethics and dissemination

The OSU Institutional Review Board (IRB), which will serve as the single IRB of record for all sites, has approved this protocol. All protocol amendments will be communicated for approval to the OSU IRB. Before a site may start the trial, it must be certified, which involves certification of research staff and an IRB reliance agreement with the single IRB.

We will submit study results for publication in peer-reviewed journals. The DCC and CCC will maintain access to the final trial dataset, and a limited deidentified dataset will be released via the online portal of the primary funder. A key component of our dissemination plan will be increasing patient and provider awareness about the comparative effectiveness results. Our partnership with DiabetesSisters and the Stakeholder Engagement Group will be leveraged for dissemination of results, including appropriate forums (eg, meetings, newsletters, social media communities, online videos). We will share accessible evidence-based factsheets and provide our primary publications for free download, including to study participants.

Discussion

In this two-arm, open-label, pragmatic, comparative effectiveness RCT, we will examine whether metformin is not inferior to insulin in reducing adverse pregnancy outcomes and is comparably safe for exposed mothers and children, and whether patient-reported factors including facilitators and barriers of medication use differ between metformin versus insulin use. The DECIDE trial will randomise 1572 pregnant individuals with GDM who need pharmacotherapy at 20 US sites—with uniform diagnostic and treatment criteria—to oral metformin versus injectable insulin and follow them and their children through delivery and then to 2 years post partum.

The proposed comparative effectiveness study is designed to inform one of the most frequent medication decisions in pregnancy. The clinical equipoise that currently exists in use of these medications for GDM underscores that a trial with pregnancy and postpartum follow-up in a diverse, representative and contemporary US population is necessary and will fill a key knowledge gap affecting everyday practice, patient experience and clinical outcomes.⁴¹ These themes, listed in bold below, have been identified as critical by stakeholders including patients, providers, researchers and professional societies.

Fill a critical evidence gap with regard to the optimal pharmacotherapy for individuals with GDM to prevent adverse pregnancy outcomes

Among the major limitations of the RCTs to date are (1) using varying GDM diagnostic criteria, (2) unclear criteria or guidelines for supplemental insulin, (3) lack of sufficient power for important outcomes, (4) insufficient long-term assessment of outcomes in exposed children, (5) unreported patterns of hyperglycaemia potentially influencing treatment effectiveness and (6) results from populations that do not reflect a contemporary US population. DECIDE will address each of these limitations with uniform diagnostic and treatment criteria and inclusion of 20 academic and community centres representative of major US geographical regions with diverse population characteristics.

Identify the long-term outcomes of metformin versus insulin on pregnant person and child health

Experts have cautioned that a GDM treatment trial without a plan for robust postnatal follow-up will not meaningfully fill the evidence gap and allow best practices to be determined.^{24 71} DECIDE embeds a seamless, preplanned and rigorous follow-up of all randomised mother–child dyads.

Characterise patient experiences of individuals with GDM requiring pharmacotherapy

An in-depth understanding of patient and other key stakeholder perspectives on barriers to and facilitators of metformin versus insulin use is necessary to identify opportunities to improve outcomes. DECIDE includes PROs and outcomes that focus on the same constructs to bolster patient and stakeholder confidence.⁷⁵ DECIDE also assesses patient experiences, such as medication side effects, whether patients take medicines as directed, how satisfied they are with their medication choice, and how their medication choice impacts their pregnancy and postpartum experience, which may explain observed HTEs.

Active patient and stakeholder engagement

The proposed study is designed with the goal of informing healthcare decisions, both by filling an important evidence gap and by ensuring that the evidence provided is aligned with and informed by patients and other healthcare partners. While conducting the study, we will engage with the patient advisory board and stakeholder engagement group, which includes patients, patient advocates, clinicians, researchers, purchasers, payors, industry, health systems and policy-makers. We will discuss the study protocol and startup in a cooperative learning environment, and these stakeholders will be invited to participate in data analysis to add their perspectives to promote authenticity.

Limitations and strengths

Limitations

First, while randomisation to pharmacotherapy minimises selection bias, lack of patient and provider blinding to treatment can introduce bias. Second, because this is a pragmatic RCT, variations in insulin formulary and differences in medication titration may result in heterogeneity in outcomes. To minimise the impact of variation of treatment effects across study sites, we have instituted uniform criteria for treatment initiation, defined as ≥30% elevated glucose values in the prior week. Also, the DECIDE manual of operations will contain guidelines for insulin and metformin management and standardised glycaemic targets for medication titration. We will stratify randomisation by site, and we will consider adjustment for site in analyses via both stratification and interaction effects. Finally, we include follow-up through 2 years postpartum, although longer follow-up may be necessary to assess the long-term impact of pharmacotherapy on outcomes.

Strengths

We have powered our study to a conservative non-inferiority margin, which is consistent with recent non-inferiority RCTs for GDM²³ and allows for substitution of supplemental insulin for those on metformin. Second, we examine postpartum safety following exposure to metformin versus insulin on child and maternal/paternal health. Third, we integrate rigorous assessment of patient preferences and values through PROs, standardised measures and qualitative interviews as part of the RCT and follow-up. Finally, DECIDE includes a racially, ethnically and geographically diverse patient population with broad inclusion criteria reflective of obstetric practice to maximise relevance, impact and generalisability.

All statements in this report, including its findings and conclusions, are solely those of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute (PCORI), its Board of Governors or Methodology Committee.