Use of metformin to prolong gestation in preterm pre-eclampsia: Randomised, double blind, placebo controlled trial

Objective To evaluate whether extended release metformin could be used to prolong gestation in women being expectantly managed for preterm pre-eclampsia. Design Randomised, double blind, placebo controlled trial. Setting Referral hospital in Cape Town, South Africa. Participants 180 women with preterm pre-eclampsia between 26+0 to 31+6 weeks’ gestation undergoing expectant management: 90 were randomised to extended release metformin and 90 to placebo. Intervention 3 g of oral extended release metformin or placebo daily, in divided doses, until delivery. Main outcome measure The primary outcome was prolongation of gestation. Results Of 180 participants, one woman delivered before taking any trial drug. The median time from randomisation to delivery was 17.7 days (interquartile range 5.4-29.4 days; n=89) in the metformin arm and 10.1 (3.7-24.1; n=90) days in the placebo arm, a median difference of 7.6 days (geometric mean ratio 1.39, 95% confidence interval 0.99 to 1.95; P=0.057). Among those who continued to take the trial drug at any dose, the median prolongation of gestation in the metformin arm was 17.5 (interquartile range 5.4-28.7; n=76) days compared with 7.9 (3.0-22.2; n=74) days in the placebo arm, a median difference of 9.6 days (geometric mean ratio 1.67, 95% confidence interval 1.16 to 2.42). Among those who took the full dosage, the median prolongation of gestation in the metformin arm was 16.3 (interquartile range 4.8-28.8; n=40) days compared with 4.8 (2.5-15.4; n=61) days in the placebo arm, a median difference of 11.5 days (geometric mean ratio 1.85, 95% confidence interval 1.14 to 2.88). Composite maternal, fetal, and neonatal outcomes and circulating concentrations of soluble fms-like tyrosine kinase-1, placental growth factor, and soluble endoglin did not differ. In the metformin arm, birth weight increased non-significantly and length of stay decreased in the neonatal nursery. No serious adverse events related to trial drugs were observed, although diarrhoea was more common in the metformin arm. Conclusions This trial suggests that extended release metformin can prolong gestation in women with preterm pre-eclampsia, although further trials are needed. It provides proof of concept that treatment of preterm pre-eclampsia is possible. Trial registration Pan African Clinical Trial Registry PACTR201608001752102 https://pactr.samrc.ac.za/. To assess the pharmacokinetics of extended release metformin in women with preterm pre-eclampsia, we performed an open label study of 15 women with preterm pre-eclampsia before the trial started (see supplementary file material and figure A for details). We performed a double blind, placebo controlled clinical trial to compare prolongation of gestation among women with a diagnosis of preterm pre-eclampsia at 26+0 to 31+6 weeks’ gestation. The women were treated with either 3 g of extended release metformin (1 g three times daily) or matching placebo. The trial site was Tygerberg Hospital in Cape Town, South Africa; a large academic referral centre situated in a region with high rates of pre-eclampsia. We recruited women with a diagnosis of preterm pre-eclampsia at 26+0 to 31+6 weeks’ gestation. Only those who were candidates for expectant management and had no clinical indication for immediate delivery, based on the clinical assessments of the attending doctors, were invited to participate. Eligible women were those who were able to give informed consent, were not currently using metformin, and had a fetus without structural anomalies. Exclusion criteria included established maternal or fetal compromise that necessitated immediate delivery, diabetes or gestational diabetes, contraindications to metformin (baseline creatinine >124 μmol/L, hypersensitivity to metformin, or a metabolic acidosis), use of drugs that might interact with metformin (glyburide, furosemide, or cationic drugs), and multiple gestations. Pre-eclampsia was defined according to the International Society for the Study of Hypertension in Pregnancy Classification14, but we only included those who had proteinuria‎ (>300 mg of protein measured on a 24 hour urine collection). Women with either new onset hypertension after 20 weeks’ gestation or women receiving treatment for chronic hypertension, with loss of blood pressure control were invited to participate provided they also fulfilled our definition of pre-eclampsia. Expectant management involved hospital admission with close maternal and fetal surveillance. Maternal surveillance included blood pressure measurements every four hours, twice daily clinical assessments, daily urinalysis, and twice weekly biochemical testing. Biochemical tests measured haemoglobin, urea, and creatinine levels and platelet count. If biochemical abnormalities were detected, then further investigations were done, including assessing levels of liver enzymes and lactate dehydrogenase. Fetal surveillance included ultrasonography for fetal size and Doppler velocimetry soon after admission, followed by reviews twice a week or more, as indicated (such as the presence of absent end diastolic flow in the umbilical artery). Cardiotocography was performed every six hours. To enhance fetal lung maturity, two doses of betamethasone 24 hours apart were given to participants when first admitted, followed by a single repeat dose one week later if no delivery had occurred, according to hospital protocol.15 Delivery before 34 weeks’ gestation was a clinical decision made by the woman’s treating team. Of the authors, only DRH was involved in clinical decisions for some participants. At Tygerberg hospital, expectant management is stopped at 34 weeks’ gestation and delivery takes place. Owing to the high demand for neonatal beds at Tygerberg hospital, neonates are transferred to a step-down facility (primary or secondary level hospital), or home as soon as allowed according to the neonate’s condition. This is an individualised clinical decision based on the neonate’s weight, gestation at delivery, feeding schedule, and progress. This trial was prospectively registered with the Pan African Clinical Trial Registry on 29 August 2016. The trial protocol has been published.16 The study was conducted with fidelity to the protocol, which is available along with the prespecified statistical analysis plan (see supplementary material). All participants provided written informed consent. Study data were collected and managed using REDCap electronic data capture tools.17 The women were randomly assigned to receive extended release metformin or placebo using a 1:1 online, web based sequence generator. Because gestation might affect the length of pregnancy prolongation, we stratified randomisation according to gestation at recruitment (stratum 1: 26+0 to ≤28+6 weeks; stratum 2: 29+0 to 31+6 weeks) using blocks of size four to six. The South Australian Health and Medical Research Institute in Adelaide, Australia (https://www.sahmri.org/) set up the randomisation. CAC and SS selected and randomised the participants. Both the metformin and the placebo tablets were manufactured by Merck Healthcare (Darmstadt, Germany) and provided as a research donation. Each metformin tablet contained 500 mg of metformin hydrochloride. The placebo tablets contained no active ingredient. The tablets were identical in size, thickness, physical properties, and appearance. The trial pharmacy at TREAD Research, Tygerberg Hospital (https://www.tread.co.za/) manufactured identical packages, each labelled with unique identifiers, according to a randomly generated sequence known only to the randomisation centre. If follow-up packs were required, the randomisation centre dispatched packs from the same allocation. Trial participants, doctors, care providers, and investigators had no access to the randomisation list, and allocation concealment was maintained throughout the trial. The women initially received one trial tablet three times a day, with the dose increased to two tablets three times a day as tolerated. If side effects developed, the dose could be decreased to one tablet twice a day and then increased if symptoms improved, to a maximum of 3000 mg/day. We recommended that treatment should be taken from randomisation until delivery. After delivery, the treatment packs were collected and the remaining tablets counted by CAC and SS. The trial midwives reviewed participants daily for drug compliance and adverse events. Serious adverse events were reported to both the data monitoring and safety committee and the health research ethics committee and were handled in accordance with good clinical practice guidelines. Plasma samples were collected from the women at randomisation and twice a week until birth. The samples were processed and frozen at −80°C. After recruitment was complete, the frozen samples were shipped in one batch to Melbourne, Australia where ST, TJK, and FB measured levels of circulating soluble fms-like tyrosine kinase-1, placental growth factor, and soluble endoglin. (See supplementary material for details on how samples were collected, assayed, and reported.) The primary outcome was prolongation of gestation, measured from time of the first dose of trial drug until delivery in days and hours. We prespecified an analysis for treatment received (women who took the trial drugs at any dose) and a further analysis limited to women who took the full dosage (ie, from randomisation to delivery). Secondary outcomes included composite maternal, fetal, or neonatal outcomes and circulating levels of anti-angiogenic biomarkers associated with pre-eclampsia. The maternal composite outcome included any of the following: maternal death, eclampsia, pulmonary oedema (oxygen saturation ≤90%, with clinical symptoms requiring treatment), severe renal impairment (serum creatinine level ≥125 μmol/L), a cerebral vascular event, placental abruption, and liver haematoma or rupture. The fetal composite outcome included major changes in heart rate patterns on the non-stress test or cardiotocograph, as defined by the attending clinician, that necessitated delivery; intrauterine fetal death; or fetal growth restriction at birth, defined as an estimated fetal weight less than the third centile on the GROW (gestation related optimal weight) chart or less than the third centile on the World Health Organization charts, or both.18 19 The neonatal composite outcome included any of the following: neonatal death within six weeks of the expected due date, grade III or IV intraventricular haemorrhage (on cranial ultrasonography), necrotising enterocolitis (on radiographic studies), or bronchopulmonary dysplasia (needing oxygen at 28 days of life). Exploratory outcomes included individual clinical outcomes (listed in table 2). The sample size was initially informed by a retrospective cohort study from our unit, where the mean prolongation for expectant management of preterm pre-eclampsia was 11 days (SD 7 days) and a coefficient of variation of 0.64.20 A normal distribution was assumed, and the sample size was uplifted by 15% to account for non-normality, resulting in a sample size of 120 women. In 2018 we completed the Pre-eclampsia Intervention Trial (evaluating esomeprazole), which was run at the same site and recruited a similar population of participants.9 The overall prolongation in the placebo arm of the Pre-eclampsia Intervention Trial was 13.1 (SD 12.2) days. We therefore recalculated the sample size for the current trial, and settings included a coefficient of variation set higher, at 1.0, a power of 90%, and a detectable geometric mean ratio of 1.6. Overall, 130 participants were required (65 in each group) for a power of 90% to detect a prolongation of gestation longer than five days, significant at a two sided 5% level. We increased our recruitment target to 75 women in each group to compensate for anticipated dropouts and amended the protocol (version dated 26 January 2018).15 During the trial, the incidence of side effects was higher than anticipated, resulting in a higher than expected number of participants who decreased or stopped the trial drug. While still blinded, the investigators made a request to the data monitoring and safety committee on 19 June 2019 to increase the sample size to 90 women in each arm, and this was approved. The final target recruitment was therefore 180 participants. We prepared a prespecified statistical analysis plan that was finalised before unblinding (see supplementary material). Personal and clinical data are presented as means (standard deviations), medians (interquartile ranges), minimums, maximums, and numbers (percentages), according to type and distribution. The primary outcome analysis was performed using ordinary least squares regression of log transformed prolongation. Bootstrap resampling (10 000 replications) was used to estimate the associated 95% confidence intervals. The use of geometric mean ratio followed input from the statistical reviewer for The BMJ during peer review, who highlighted the inadequate sample size if quantile median regression was used. The use of geometric mean ratio is concordant with the prespecified sample size and power calculations. Gestational strata are presented as covariates, and the results are presented as geometric mean ratios with corresponding 95% confidence intervals and P values. We performed supplementary survival analyses using Cox proportional hazards regression adjusted for strata of gestational age and presented as hazard ratios (95% confidence intervals), along with a Kaplan-Meier survival curve to illustrate the prolongation of gestation. The prespecified composite maternal and neonatal outcomes were tested using an exact logistic regression model, with group and strata as covariates. We considered a two sided P value <0.05 to be significant. For the exploratory outcomes, we used the same statistical techniques as those used to compare the primary and secondary outcomes, but without correction for strata. Software Stata v16 (Stata Statistical Software Release 16; Stata) was used. Although we did not actively seek patient or public involvement in the development of the protocol, the steering committee of international content experts included a woman who had experienced pre-eclampsia. Furthermore, we spoke to patients about the study and asked a member of the public to read our manuscript before submission. Participants are acknowledged and thanked for their contribution and participation. A dissemination strategy has been developed to work closely with stakeholders and knowledge users to facilitate transfer of the findings to relevant end users, including patients, patient organisations, and clinicians.