Objectives
This is a protocol for a Cochrane Review (intervention). The objectives are as follows:
To evaluate the benefits and harms of metformin therapy initiated prior to conception and continued through the first trimester for women with polycystic ovary syndrome, compared with placebo or no metformin treatment, on pregnancy outcomes.
Background
Description of the condition
Polycystic ovary syndrome (PCOS) is the most common endocrine disorder among women of reproductive age, and is characterised by a combination of reproductive and metabolic abnormalities [1]. PCOS is associated with delayed diagnosis, subfertility, and substantial psychological burden, including anxiety and depression related to infertility and pregnancy loss [2]. The diagnosis of PCOS is commonly based on the revised Rotterdam criteria, endorsed by the European Society of Human Reproduction and Embryology (ESHRE) and the American Society for Reproductive Medicine (ASRM) [3]. According to these criteria, a diagnosis is confirmed when two of the following three features are present: clinical or biochemical evidence of hyperandrogenism, oligo‐ovulation or anovulation, and polycystic ovarian morphology on ultrasound, provided that other differential diagnoses have been excluded [3]. The metabolic features of PCOS are equally relevant and include an increased risk of developing type 2 diabetes mellitus and cardiovascular disease, which are evidenced by dyslipidaemia, elevated blood pressure, and impaired vascular function [4, 5, 6, 7]. Prevalence estimates based on these criteria suggest that up to 20% of women of reproductive age are affected by PCOS [8, 9, 10, 11].
Miscarriage, defined as the spontaneous loss of a pregnancy before 24 completed weeks of gestation, affects approximately 15% of all recognised pregnancies [12]. The risk is highest in the first trimester, with a marked decline beyond 14 weeks of gestation [12]. Miscarriage can lead to serious complications, including haemorrhage and infection, particularly in low‐resource settings [12]. The psychological impact is also considerable, with many women experiencing symptoms of anxiety, depression, and post‐traumatic stress disorder following pregnancy loss [12].
Women with PCOS are at heightened risk of miscarriage, with reported rates ranging from 30% to 50% [13, 14, 15, 16]. The underlying mechanisms for the increased miscarriage risk are not completely understood; however, insulin resistance has been implicated as an independent risk factor for early pregnancy loss [17]. In a systematic review evaluating in vitro fertilisation (IVF) and intracytoplasmic sperm injection (ICSI) outcomes in women with PCOS, those with insulin resistance had a significantly higher miscarriage rate compared to those without (odds ratio 1.11, 95% CI 1.02 to 1.22) [18].
Description of the intervention and how it might work
Metformin, a biguanide oral hypoglycaemic agent, is primarily used in the treatment of type 2 diabetes mellitus and exerts its effect by improving insulin sensitivity [19]. Given the central role of insulin resistance in the pathophysiology of PCOS, metformin has been hypothesised to have the potential to improve reproductive outcomes in this population [1, 19]. Metformin reduces insulin resistance through several mechanisms, including inhibition of hepatic gluconeogenesis and lipogenesis, and enhancement of peripheral glucose uptake [20]. Its use during pregnancy is widespread and is generally considered safe, with current evidence showing no increased risk of congenital anomalies associated with prenatal exposure [21].
Clinical practice regarding the use of metformin in women with PCOS who are trying for a pregnancy is variable. One common strategy involves discontinuing metformin at the point of a confirmed pregnancy, following its use as an adjunct to ovulation induction or assisted reproductive technology (ART), an approach supported by current international clinical guidelines [22]. Alternatively, as there is a theoretical concern that abrupt cessation of metformin in early pregnancy may lead to a rebound in insulin resistance, potentially adversely affecting pregnancy outcomes [23], some clinicians continue metformin therapy throughout the first trimester, and others advocate for ongoing treatment for the full duration of the pregnancy in order to potentially reduce the risk of miscarriage. Given this, there is a need for a review of available high‐quality evidence to determine whether continuation of metformin beyond conception, particularly during the first trimester, confers protection against miscarriage in women with PCOS.
In this review, we will explore sources of clinical heterogeneity that may modify the effect of metformin on miscarriage risk in women with PCOS: miscarriage history, body mass index (BMI), ethnicity, PCOS phenotype, method of conception; and metformin dose. Previous miscarriage history reflects baseline risk and may identify women with differing underlying pathophysiology and treatment response. We consider miscarriage history a subgroup of special interest because one or more previous miscarriages is an established clinical risk factor already used to guide miscarriage prevention interventions, including progesterone treatment recommended in NICE (National Institute for Health and Care Excellence) guidance [24]. We will conduct detailed subgroup analysis by the number of previous miscarriages (1, 2, and 3 or more) to explore whether there is a biological gradient in treatment effect with increasing miscarriage burden. We will look at the impact of BMI, as it is closely linked to insulin resistance and metabolic dysfunction, key mechanisms through which metformin may act. We will consider ethnicity and PCOS phenotype, as they may reflect variation in metabolic profile and treatment response. As described in the Background, PCOS is classified into four phenotypes under the Rotterdam Criteria, with diagnosis requiring at least two of three features: hyperandrogenism, ovulatory dysfunction, and polycystic ovaries [22]. We will explore method of conception because hormonal and early pregnancy characteristics differ between unassisted and assisted conception. We will also consider different doses of metformin because treatment effects may vary according to drug exposure and tolerability, with daily doses varying from 500 to 1500 mg. These subgroup analyses will assess whether any effects we find are consistent across clinically relevant groups or are concentrated within specific populations.
Why it is important to do this review
Despite its biological rationale and widespread clinical use, the effect of metformin on miscarriage risk in women with PCOS is uncertain. This ambiguity may be largely attributable to heterogeneity in study populations, and inconsistency in the timing and duration of metformin administration across existing trials [25, 26]. The absence of such evidence has contributed to divergent clinical practices and an ongoing lack of consensus among healthcare providers regarding the continuation of metformin beyond the point of pregnancy confirmation.
As mentioned above, current international guidance, including the 2023 global guideline on PCOS management, is that metformin, when used as an adjunct to fertility treatment, should be discontinued once pregnancy is confirmed [22]. However, this recommendation is based on limited evidence and does not address whether continuing metformin into early gestation might reduce miscarriage risk [22].
The aim of this review is to evaluate the effect of metformin exposure, initiated before conception and continued through the first trimester, on the outcome of miscarriage in women with PCOS. Through systematic study identification, data extraction, and meta‐analysis, we aim to generate a high‐quality summary and evaluation of the data available from randomised controlled trials, in order to inform clinical decision‐making and potentially guide future updates to practice guidelines on the use of metformin for miscarriage prevention in this high‐risk population.
Objectives
To evaluate the benefits and harms of metformin therapy initiated prior to conception and continued through the first trimester for women with polycystic ovary syndrome, compared with placebo or no metformin treatment, on pregnancy outcomes.
Methods
Criteria for considering studies for this review
Types of studies
We will include randomised controlled trials (RCTs) with individual allocation. For cross‐over trials, we will only include data from the first intervention period (before cross‐over). Quasi‐randomised trials, defined as studies that allocate participants using a predictable rule, such as alternation, date of birth, or record number, will be excluded. We will include studies whether they are published as full‐text articles or as abstracts only, as well as unpublished data.
Types of participants
We will include RCTs of women with polycystic ovary syndrome (PCOS) who are attempting to become pregnant. We will accept the diagnosis of PCOS as determined in each individual study. If studies include mixed populations (i.e. that include women with other conditions or with no conditions, along with those who have PCOS), we will include them only if data for women with PCOS can be extracted separately, or can be obtained from study authors on request.
Types of interventions
We will include any RCT that compares metformin with no metformin (control) or with placebo if both are given at the same time, either before or soon after pregnancy has been confirmed, and continued until at least 12 weeks of gestation.
Outcome measures
We will exclude studies if none of our outcomes of interest are measured, and we have confirmed this with supporting evidence (e.g. communication with study authors or access to the original protocol). We will include any eligible study that evaluates at least one of our critical or important outcomes of interest.
Critical outcomes
Miscarriage: defined as pregnancy loss < 24 weeks (up to and including 23+6 weeks (i.e. 23 weeks and 6 days) of gestation)
Live birth, greater than or equal to 34 weeks of gestation
Important outcomes
Gastrointestinal side effects: these include nausea, vomiting, diarrhoea, and abdominal discomfort, assessed throughout the active treatment period.
Clinical pregnancy: defined as the presence of an intrauterine gestational sac confirmed on ultrasound, with or without foetal cardiac activity
Live birth, greater than or equal to 24 weeks of gestation
Preterm birth: defined as birth at less than 37 weeks of gestation
First‐trimester loss: defined as pregnancy loss up to and including 13+6 weeks (i.e. 13 weeks and 6 days) of gestation (amongst all women with positive pregnancy tests)
Second‐trimester loss: pregnancy loss from 14+0 weeks until 23+6 weeks inclusive (amongst all women with positive pregnancy tests)
Foetal growth restriction: as defined by the study investigators
Birthweight, in grams
Head circumference, in cm
Congenital abnormalities: defined as any structural or chromosomal anomaly identified antenatally or within 28 days of birth
Stillbirth: defined as the delivery of a baby after 24 weeks of gestation showing no signs of life
Gestational diabetes: diagnosed during pregnancy according to the criteria used in each included trial. Differences in diagnostic criteria and thresholds will be documented and explored as a potential source of clinical heterogeneity.
Multiple pregnancy (i.e. more than one foetus)
Death within 28 days of neonatal life
Search methods for identification of studies
Electronic searches
We will identify trials through systematic searches of the following bibliographic databases.
Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library
MEDLINE (Ovid; from 1946 onwards)
Embase (Ovid; from 1980 onwards)
CPCI‐S (Conference Proceedings Citation Index‐Science) on the Web of Science (Clarivate Analytics; from 1990 onwards)
We will also search the following trials registries for ongoing or unpublished trials.
US National Institutes of Health Ongoing Trials Registry ClinicalTrials.gov (https://clinicaltrials.gov)
WHO International Clinical Trials Registry Platform (ICTRP) Search Portal (https://trialsearch.who.int)
We will search all databases from their inception and impose no restrictions on publication language or publication status. We will not perform a separate search for adverse effects of metformin; we will only consider adverse effects described in the included studies. Detailed search strategies are provided in Supplementary material 1.
Searching other resources
We will check reference lists of all included studies and any relevant systematic reviews to identify additional references to trials. We will explore citation‐based search methods for selected articles, including forward citation searches via Scopus and Google Scholar, as well as searches for related articles in MEDLINE through PubMed. We will also conduct a grey literature search using Google Scholar and the database search results. Additionally, we will examine any relevant retraction statements and errata using RetractionWatch (https://retractiondatabase.org).
Data collection and analysis
Selection of studies
We will use Covidence software for study selection [27]. Two review authors (AD, JC) will independently evaluate the titles and abstracts of all search results, categorising them as 'include' (eligible or potentially eligible or unclear) or 'exclude'. In the case of any discrepancies, a third review author (RS) will adjudicate. We will obtain full‐text articles of all eligible and potentially eligible or unclear records, and two review authors (AD, NP) will assess them based on our eligibility criteria, documenting reasons for excluding ineligible studies. Any disagreements will be resolved through discussion or consultation with a third review author (RS). Duplicates will be identified and excluded, and we will compile multiple reports of the same study to ensure each study, rather than each article, is the primary unit of interest in the review. We will document the selection process in detail, in order to create a comprehensive PRISMA flow diagram [28]. We will document all articles excluded after full‐text assessment in the 'Characteristics of excluded studies' table, along with the reasons for exclusion [28]. In light of the involvement of some review authors in an ongoing potentially eligible trial [29], all stages of study selection, data extraction, risk of bias assessment, and certainty of evidence assessment for that study will be undertaken independently by review authors with no potential conflict of interest (JC, NP, AG).
Screening eligible studies for scientific integrity and trustworthiness
Two review authors (AC, NP) will evaluate all studies meeting our inclusion criteria against predefined criteria, to select studies that, based on available information, are deemed sufficiently trustworthy to be included in the analysis. These criteria have been developed by Cochrane Pregnancy and Childbirth, and are as follows [30].
Research governance
No prospective trial registration for studies published after 2010 without plausible explanation
When requested, trial authors refuse to provide/share the protocol and/or ethics approval letter.
Trial authors refuse to engage in communication with the Cochrane review authors.
Trial authors refuse to provide individual patient data (IPD) data upon request with no justifiable reason.
Baseline characteristics
Characteristics of the study participants that appear implausibly similar, as indicated by distributions of means (standard deviations (SDs)) that are either excessively narrow or excessively wide [31]
Feasibility
Implausible numbers (e.g. 500 women with severe cholestasis of pregnancy recruited in 12 months)
(Close to) zero losses to follow‐up without plausible explanation
Results
Implausible results (e.g. massive risk reduction for main outcomes with small sample size)
Unexpectedly even numbers of women ‘randomised’, including a mismatch between the numbers and the methods. For example, if the study authors state no blocking was used but still end up with equal numbers, or they state that blocks of four were used, but the final numbers differ by six.
Where we classify a study as being at ‘high risk’ for one or more of the above criteria, we will attempt to contact the study authors to ask for further information to address concerns. If adequate information remains unavailable, we will categorise the study as ‘awaiting classification’, and we will describe our concerns and communications with the author (or lack thereof) in the review.
Data extraction and management
We will use a data collection form that has undergone pilot testing on at least one included study to document study characteristics and outcome data. Two review authors will be responsible for extracting the specified study characteristics. For the potentially eligible study that some of the review authors are involved with, review authors JC, NP, and AG will independently perform data extraction.
-
Methods
Study design: total duration of study (start date to end date; absence of this information will be reported), number of study centres and location, study setting, study definition of PCOS
-
Participants
Inclusion and exclusion criteria
Participant details, baseline demographics (mean age, age range), number randomised, number lost to follow‐up or withdrawn, number analysed
-
Interventions
Dose of metformin
Time points for treatment commencement and cessation
-
Outcomes
Definitions of relevant outcomes, including the methods and timing of their measurement, along with any pertinent subgroups for the review
Study sources of funding
Notable conflicts of interest of study authors
We will present these data in the 'Characteristics of included studies' table and summarise them in the 'Results' section of the review. If necessary, we will contact the authors of the included studies to request additional information or clarification about their research. We will document any such communications.
Handling duplicate and companion publications
In the case of duplicate publications, companion documents, or multiple reports of a primary study, we will maximise the data available by combining all relevant information. We will use the most comprehensive data set, aggregated from all known publications (including any data available in clinical trial registers, such as ClinicalTrials.gov or other similar sources).
In cases where published and unpublished data differ, we will seek clarification from the study authors. If there is no response, we will highlight the discrepancies in the review. If an included study is marked as completed in a clinical trial register but no additional information (such as study results or a publication) is available, we will include this study in the 'Characteristics of studies awaiting classification' table.
We will cite duplicate publications, companion documents, multiple reports of a primary study, and trial documents (such as trial registry information) from included studies as secondary references under the study ID of the relevant included study. Similarly, duplicate publications, companion documents, multiple reports of a study, and trial documents (such as trial registry information) from excluded studies will be listed as secondary references under the study ID of the excluded study.
Risk of bias assessment in included studies
Two review authors (AD, NP) will evaluate the risk of bias in each study using the Cochrane RoB 2 tool for all outcomes [32, 33]. Any discrepancies will be resolved through discussion or by seeking input from another review author (RS). Review authors JC, NP, and AG will independently perform the risk of bias assessment where the other review authors have declared a conflict of interest for a potentially eligible study.
We will assess the risk of bias for specific study results based on the following domains.
Bias arising from the randomisation process
Bias due to deviations from intended interventions
Bias due to missing outcome data
Bias in measurement of the outcome
Bias in selection of the reported result
The risk of bias in each domain for all outcomes included in the summary of findings table will be evaluated using the RoB 2 tool. The focus will be on quantifying the effect of assignment to the intervention at baseline, known as the intention‐to‐treat (ITT) effect [32, 34]. We will rate each domain as 'low risk of bias', 'some concerns', or 'high risk of bias' using the signalling questions in the RoB 2 tool. We will create risk of bias tables to display the rationale for the judgements we make. We will include relevant quotes from studies where appropriate to support our risk of bias assessments. The overall risk of bias for the result will be determined by the least favourable assessment across the bias domains. When analysing treatment effects, we will take into consideration the risk of bias for the studies contributing to each outcome. We will use the RoB 2 Excel tool for the assessment process [34]. If cluster‐randomised trials are included in the review, we will use the RoB 2 tool with the cluster‐trial extension. We will list consensus decisions for the signalling questions in an appendix. We will generate a figure in RevMan to present a visual summary of the risk of bias assessment.
Measures of treatment effect
In order to measure the treatment effects, we will report dichotomous outcomes as risk ratios (RRs) with 95% confidence intervals (CIs). For continuous data, we will calculate mean differences (MDs) with 95% CIs when the same scale is used to measure an outcome, and as standardised mean difference (SMD) with 95% CIs where different scales are used to measure the same outcome. If any studies report odds ratios (ORs) or hazard ratios (HRs), we will estimate an RR for each trial before pooling the data in meta‐analysis [35]. We will narratively describe skewed data reported as medians and interquartile ranges [36]. Where appropriate, we will report the number needed to treat for an additional beneficial outcome (NNTB) with 95% CIs and the number needed to treat for an additional harmful outcome (NNTH) with 95% CIs.
Unit of analysis issues
The individual participant will be the unit of analysis. We acknowledge the challenges related to unit of analysis when dealing with composite outcomes. Composite outcomes, such as adverse events, are typically reported as the number of participants with specific outcomes, as opposed to the number of participants with one or more outcomes. Our intention is to extract the number of participants with one or more complications, but if this is not possible, we will report the number of cases for each component of the composite outcome only.
We do not expect to identify any cluster‐randomised trials or multi‐arm studies. However, if a cluster‐randomised trial is included, we will assess whether the analysis appropriately accounted for the clustered design. Where clustering has not been adequately addressed, we will adjust the analysis, including the use of an effective sample size and an intracluster correlation coefficient, where available. The likely allocation of the intervention made at a group level will be all inpatients in a particular unit. We will carefully consider any potential confounding factors that may impact the results that are unrelated to the intervention, including differences in practice across units, or heterogeneity in cohorts. We will use the intracluster (or intraclass) correlation coefficient (ICC) to convert trials to their effective sample size before incorporating them into meta‐analysis [37]. If we include a multi‐arm study, we will include the pairwise comparison data of participants. We will not include data from any intervention group that is not relevant to the review [37].
For trials that include multiple intervention arms with a single control group (e.g. arms with different metformin doses), we will combine the relevant intervention groups into a single group for the primary analysis to avoid double‑counting participants in the control arm, in accordance with guidance in the Cochrane Handbook for Systematic Reviews of Interventions (Chapter 23.3.4) [37]. Where comparisons between specific doses are of interest, we will retain the individual intervention arms in the prespecified subgroup analysis by metformin dose.
Dealing with missing data
We will attempt to contact the authors of the included studies to request any unreported or missing data, or to clarify unclear data. For each study, we will record the number of participants lost to follow‐up and the number of dropouts. In instances where standard deviations are missing for continuous data, we will calculate them from standard errors, P values, or 95% CIs. If we calculate any data, we will report these in the 'Characteristics of included studies' table [33]. If it is not feasible to calculate missing data in this manner and where we are unable to retrieve missing data, we will describe them for each included study in the risk of bias table, and we will discuss the potential impact of missing data in the 'Results' and 'Author conclusions' of this review. We will perform a sensitivity analysis to explore the effects of missing data (see Sensitivity analysis).
Reporting bias assessment
We will conduct comprehensive literature searches without any restrictions on publication date or language, in order to minimise reporting bias. We will also use study protocols and trial registrations to evaluate studies for selective reporting. When more than 10 studies are included in a meta‐analysis, we will use Egger's weighted regression test for continuous outcomes and the Harbord test for binary outcomes. Our assessment will include identifying potential publication bias and investigating potential biases arising from small studies. Additionally, we will examine outcome reporting bias by comparing the results extracted from trial protocols with their published reports. Our evaluation will aim to determine the presence of selective outcome reporting [38].
Synthesis methods
Using Review Manager (RevMan) [39], where possible, we will perform meta‐analysis with trials that provide data for a particular outcome. We will use the random‑effects model as we anticipate clinical and methodological variation across trials, including differences in metformin dose, gestational timing of treatment initiation, and comparator interventions. Between‑study variance (τ²) will be estimated using the restricted maximum likelihood (REML) method. Confidence intervals for pooled effect estimates will be calculated using the Hartung–Knapp–Sidik–Jonkman (HKSJ) adjustment when at least three studies contribute to a meta‑analysis and τ² is greater than zero. In meta‐analyses including only two studies, or where the estimated heterogeneity is zero, we will use the standard Wald‑type method. In addition to the pooled effect estimate and its confidence interval, we will report 95% prediction intervals for all random‑effects meta‑analyses.
The primary analyses will include all eligible studies, regardless of their risk of bias. For dichotomous outcomes, we will calculate RR with 95% CI. For continuous outcomes measured with the same validated scale and units of measure, we will pool the data using MD with 95% CI. For continuous outcomes measured with different scales and units of measure, we will use SMD with 95% CI.
We will not pool data when there is considerable heterogeneity (I2 ≥ 75%) that cannot be explained by the diversity of clinical and methodological features amongst studies, or by the limited available data. We will instead present their results in a narrative synthesis, following the Synthesis Without Meta‐analysis (SWiM) reporting guideline [40], which promotes transparent reporting of study grouping, the standardised metric used for the synthesis, the synthesis method, how data are presented, a summary of the synthesis findings, and limitations of the synthesis.
We will summarise the findings by outcome in a summary of findings table, which will include information on sample size, effect size and direction, and our assessment of the certainty of the evidence (see Certainty of the evidence assessment).
Investigation of heterogeneity and subgroup analysis
We will evaluate the clinical, methodological, and statistical heterogeneity within the trials. It is important to note that clinical heterogeneity can be present even in the absence of statistical heterogeneity. Therefore, we will conduct a thorough analysis of the trials to identify variations in participant characteristics, interventions administered, and outcomes assessed. We will also assess methodological heterogeneity by examining variations in study design and conduct.
We will assess statistical heterogeneity of the estimate of effects between the included studies by visually inspecting the forest plots, and calculating the Chi2 test and the I2 statistic. We will quantify the possible magnitude of inconsistency (i.e. heterogeneity) using the following ranges of I2 to guide interpretation [32].
0% to 40% might not be important.
30% to 60% may represent moderate heterogeneity.
50% to 90% may represent substantial heterogeneity.
75% to 100% represents considerable heterogeneity.
If we identify considerable heterogeneity (I2 ≥ 75%), we will investigate possible causes by comparing the characteristics of each study and undertaking subgroup analyses where possible, including an investigation of interactions [41].
Subgroup of special interest
Miscarriage history (0; 1 or more; number of miscarriages not known)
Additional subgroups
Number of previous miscarriages (0; 1 or 2; 3 or more; number of miscarriages not known)
BMI (< 35 kg/m2 versus ≥ 35 kg/m2)
Ethnicity (white; black; south Asian; other; ethnicity not known)
PCOS phenotype (according to PCOS phenotype classifications used in eligible studies)
Method of conception (unassisted versus ovulation induction versus intrauterine insemination versus in vitro fertilisation)
Dose of metformin
In the event that a particular characteristic appears to hold significance for subgroup analysis, we will conduct post‐hoc analysis and document our rationale for doing so. We will use the formal test for subgroup differences in RevMan [39], and base our interpretation on the result.
Sensitivity analysis
We intend to conduct the following sensitivity analyses for the critical outcomes (miscarriage and live birth) to examine whether crucial methodological factors or decisions have influenced the observed effect sizes [42].
Risk of bias: removing studies with an overall high risk of bias
Missing data: excluding studies judged to be at high risk of bias arising from missing outcome data (RoB 2 domain 3)
Publication status: excluding unpublished data
Meta‐analysis model: repeating the analyses with the fixed‐effect model, using Mantel‐Haenszel methods for dichotomous outcomes and inverse variance methods for continuous outcomes
Certainty of the evidence assessment
Two review authors (JC, AG) will independently evaluate the certainty of the evidence based on the five GRADE considerations: study limitations, consistency of effect, imprecision, indirectness, and publication bias in relation to the studies providing data for the outcome. Any discrepancies will be resolved through discussion or by consulting a third review author (RS). In footnotes, we will provide justifications for any downgrading of evidence certainty, and we will include comments to assist readers in understanding the review as needed [43]. We will use GRADEpro GDT software and will follow the methods and recommendations outlined in chapter 14 of the Cochrane Handbook for Systematic Reviews of Interventions [43, 44, 45].
We will use the overall RoB 2 judgement to address risk of bias in the GRADE assessment. We will classify the risk of bias as: high risk of bias, some concerns, or low risk of bias. A low risk of bias will indicate no limitation (certainty will not be downgraded). Some concerns will indicate either no limitation, or serious limitation (certainty will be downgraded by one level). A high risk of bias will indicate either serious limitation, or very serious limitation (certainty will be downgraded by two levels). The certainty of evidence will be graded as high, moderate, low, or very low based on all the GRADE considerations.
We will create a summary of findings table for the comparison of preconception and first trimester metformin treatment versus placebo or no treatment in the first trimester. We will present our critical outcomes, which are miscarriage and live birth (greater than 34 weeks' gestation), and the following important outcomes: gastrointestinal side effects, clinical pregnancy, preterm birth, first‐trimester loss, and birthweight. Details of outcome definitions and measures are presented above in Outcome measures.
Equity considerations
We do not plan to undertake a formal equity‐focused analysis, as we anticipate there will be a lack of relevant information in the review's included studies; however, we will extract and report study setting and population characteristics and use these to comment on applicability.
Consumer involvement
Consumers were not involved in this review due to limited resources. However, we did use core outcome sets for the review's outcomes, which were developed with consumer input.
Supporting Information
Supplementary materials are available with the online version of this article: 10.1002/14651858.CD016141.
Supplementary materials are published alongside the article and contain additional data and information that support or enhance the article. Supplementary materials may not be subject to the same editorial scrutiny as the content of the article and Cochrane has not copyedited, typeset or proofread these materials. The material in these sections has been supplied by the author(s) for publication under a Licence for Publication and the author(s) are solely responsible for the material. Cochrane accordingly gives no representations or warranties of any kind in relation to, and accepts no liability for any reliance on or use of, such material.
Supplementary material 1 Search strategies
New
Additional information
Acknowledgements
Editorial and peer‐reviewer contributions
The following people conducted the editorial process for this article.
Sign‐off Editor (final editorial decision): Dr Zarko Alfirevic, University of Liverpool, UK
Managing Editor (selected peer reviewers, provided editorial guidance to authors, edited the article): Leanne Jones, Central Editorial Service
Editorial Assistant (conducted editorial policy checks, collated peer‐reviewer comments, supported editorial team): Joshua Guinoo, Central Editorial Service
Copy Editor (copy editing and production): Laura MacDonald, Cochrane Central Production Service
Peer reviewers (provided comments and recommended an editorial decision): Jack Wilkinson, Centre for Biostatistics, University of Manchester (clinical/content review); Joelcio Francisco Abbade, Obstetrics and Gynecology Department, Botucatu Medical School, UNESP – São Paulo State University, Botucatu City, São Paulo State, Brazil (clinical/content review); Dr Rebecca Harmston (patient and public review); Tom Patterson, Cochrane Evidence Production and Methods Directorate (methods review); Jo Platt, Central Editorial Information Specialist (search review).
Contributions of authors
Arri Coomarasamy (AC), Rima Dhillon‐Smith (RDS), James Cheshire (JC), and Adam J Devall (AD) conceived the idea for this review. AD drafted the first version of the protocol and updated the cited literature. Nishanthi Periyathambi (NP) drafted the search strategy. Pedro Melo (PM), Akanksha Garg (AG), and Kugajeevan Vigneswaran (KV) provided critical input. All review authors commented on the protocol draft, provided critical input, and approved the final version.
Declarations of interest
AD: has declared that he is a co‐applicant on a randomised trial of letrozole or clomifene, with or without metformin, for ovulation induction in women with polycystic ovary syndrome that may be eligible for inclusion in this review. AD will not be involved in study selection, data extraction, risk of bias assessment, GRADE assessment, or data analysis for any trial in which he had a direct role.
JC: none known
NP: none known
Akanksha Garg (AG): none known
Kugajeevan Vigneswaran (KV): has declared that he has received a speaker fee from Merck for a lecture on individualised ovarian stimulation in IVF and is an author on a randomised trial of letrozole or clomifene, with or without metformin, for ovulation induction in women with polycystic ovary syndrome that may be eligible for inclusion in this review. KV will not be involved in study selection, data extraction, risk of bias assessment, GRADE assessment, or data analysis for any trial in which he had a direct role.
PM: has declared that he is the national clinical coordinator for a randomised trial of letrozole or clomifene, with or without metformin, for ovulation induction in women with polycystic ovary syndrome that may be eligible for inclusion in this review. PM will not be involved in study selection, data extraction, risk of bias assessment, GRADE assessment, or data analysis for any trial in which he had a direct role.
RS: has declared that she is a co‐applicant on a randomised trial of letrozole or clomifene, with or without metformin, for ovulation induction in women with polycystic ovary syndrome that may be eligible for inclusion in this review. RS will not be involved in study selection, data extraction, risk of bias assessment, GRADE assessment, or data analysis for any trial in which she had a direct role.
AC: has declared that he is Chief Investigator for a randomised trial of letrozole or clomifene, with or without metformin, for ovulation induction in women with polycystic ovary syndrome that may be eligible for inclusion in this review. RS will not be involved in study selection, data extraction, risk of bias assessment, GRADE assessment, or data analysis for any trial in which he had a direct role.
Sources of support
Internal sources
-
None, Other
No internal sources of support
External sources
-
Tommy's, Other
This worked is supported by funding from the Tommy's charity to the Tommy's National Centre for Miscarriage Research.
Registration and protocol
Cochrane approved the proposal for this review in March 2024.
Data, code and other materials
Data sharing not applicable to this article as it is a protocol, so no datasets were generated or analysed.
Disclosure of artificial intelligence use
We did not use AI (artificial intelligence) platforms, systems, or tools in the preparation of the manuscript.
References
- 1.Teede HJ, Misso ML, Deeks AA, Moran LJ, Stuckey BG, Wong JL, et al; Guideline Development Groups. Assessment and management of polycystic ovary syndrome: summary of an evidence-based guideline. Medical Journal of Australia 2011;195(6):S65-112. [DOI] [PubMed] [Google Scholar]
- 2.Almhmoud H, Alatassi L, Baddoura M, Sandouk J, Alkayali MZ, Najjar H, et al. Polycystic ovary syndrome and its multidimensional impacts on women's mental health: a narrative review. Medicine (Baltimore) 2024;103(25):e38647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group. Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome (PCOS). Human Reproduction 2004;19(1):41-7. [DOI] [PubMed] [Google Scholar]
- 4.Guan C, Zahid S, Minhas AS, Ouyang P, Vaught A, Baker VL, et al. Polycystic ovary syndrome: a 'risk-enhancing' factor for cardiovascular disease. Fertility and Sterility 2022;117(5):924-35. [DOI] [PubMed] [Google Scholar]
- 5.Glintborg D, Henriksen JE, Andersen M, Hagen C, Hangaard J, Rasmussen PE, et al. Prevalence of endocrine diseases and abnormal glucose tolerance tests in 340 Caucasian premenopausal women with hirsutism as the referral diagnosis. Fertility and Sterility 2004;82(6):1570-9. [DOI] [PubMed] [Google Scholar]
- 6.Balen A. Polycystic ovary syndrome (PCOS). Obstetrician & Gynaecologist 2017;19(2):119-29. [Google Scholar]
- 7.Visser JA. The importance of metabolic dysfunction in polycystic ovary syndrome. Nature Reviews. Endocrinology 2020;17(2):77-8. [DOI] [PubMed] [Google Scholar]
- 8.Hart R. Polycystic ovarian syndrome–prognosis and treatment outcomes. Current Opinion in Obstetrics & Gynecology 2007;19:529-35. [DOI] [PubMed] [Google Scholar]
- 9.March WA, Moore VM, Willson KJ, Phillips DI, Norman RJ, Davies MJ. The prevalence of polycystic ovary syndrome in a community sample assessed under contrasting diagnostic criteria. Human Reproduction 2010;25(2):544-51. [DOI] [PubMed] [Google Scholar]
- 10.Yildiz BO, Bozdag G, Yapici Z, Esinler I, Yarali H. Prevalence, phenotype and cardiometabolic risk of polycystic ovary syndrome under different diagnostic criteria. Human Reproduction 2012;27(10):3067-73. [DOI] [PubMed] [Google Scholar]
- 11.Wolf WM, Wattick RA, Kinkade ON, Olfert MD. Geographical prevalence of polycystic ovary syndrome as determined by region and race/ethnicity. International Journal of Environmental Resesearch and Public Health 2018;15(11):2589. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Quenby S, Gallos ID, Dhillon-Smith RK, Podesek M, Stephenson MD, Fisher J, et al. Miscarriage matters: the epidemiological, physical, psychological, and economic costs of early pregnancy loss. Lancet 2021;397(10285):1658-67. [DOI] [PubMed] [Google Scholar]
- 13.Mills G, Badeghiesh A, Suarthana E, Baghlaf H, Dahan M. Associations between polycystic ovary syndrome and adverse obstetric and neonatal outcomes: a population study of 9.1 million births. Human Reproduction 2020;35(8):1914-21. [DOI] [PubMed] [Google Scholar]
- 14.D'Alterio MN, Sigilli M, Succu AG, Ghisu V, Laganà AS, Sorrentino F, et al. Pregnancy outcomes in women with polycystic ovarian syndrome. Minerva Obstetrics and Gynecology 2022;74(1):45-59. [DOI] [PubMed] [Google Scholar]
- 15.Abruzzese G, Silva A, Velazquez M, Ferrer M-J, Motta A. Hyperandrogenism and polycystic ovary syndrome: effects in pregnancy and offspring development. WIREs Mechanisms of Disease 2022;14(5):e1558. [DOI] [PubMed] [Google Scholar]
- 16.Pan ML, Chen LR, Chen KH. The risk of subsequent miscarriage in pregnant women with prior polycystic ovarian syndrome: a nationwide population-based study. International Journal of Environmental Research and Public Health 2021;18(16):8253. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Tian L, Shen H, Lu Q, Norman RJ, Wang J. Insulin resistance increases the risk of spontaneous abortion after assisted reproduction technology treatment. Journal of Clinical Endocrinology and Metabolism 2007;92(4):1430-33. [DOI] [PubMed] [Google Scholar]
- 18.Li F, Qi JJ, Li LX, Yan TF. Impact of insulin resistance on IVF/ICSI outcomes in women with polycystic ovary syndrome: a systematic review and meta-analysis. European Journal of Obstetrics, Gynecology and Reproductive Biology 2024;299:54-61. [DOI] [PubMed] [Google Scholar]
- 19.Le J, Chen Y, Yang W, Chen L, Ye J. Metabolic basis of solute carrier transporters in treatment of type 2 diabetes mellitus. Acta Pharmaceutica Sinica. B 2024;14(2):437-54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Foretz M, Guigas B, Viollet B. Metformin: update on mechanisms of action and repurposing potential. Nature Reviews. Endocrinology 2023;19(8):460-76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Raperport C, Chronopoulou E, Homburg R. Effects of metformin treatment on pregnancy outcomes in patients with polycystic ovary syndrome. Expert Review of Endocrinology & Metabolism 2021;16(2):37-47. [DOI] [PubMed] [Google Scholar]
- 22.Teede HJ, Tay CT, Laven JJE, Dokras A, Moran LJ, Piltonen TT, et al. Recommendations from the 2023 international evidence-based guideline for the assessment and management of polycystic ovary syndrome. Journal of Clinical Endocrinology and Metabolism 2023;108(10):2447-69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Luo Z, Wang L, Wang Y, Fan Y, Jiang L, Xu X, et al. Impact of insulin resistance on ovarian sensitivity and pregnancy outcomes in patients with polycystic ovary syndrome undergoing IVF. Journal of Clinical Medicine 2023;12(3):818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.National Institute for Health and Care Excellence. Ectopic pregnancy and miscarriage: diagnosis and initial management. https://www.nice.org.uk/guidance/ng126 17 April 2019 (last updated 23 August 2023). [PubMed]
- 25.Sharpe A, Morley LC, Tang T, Norman RJ, Balen AH. Metformin for ovulation induction (excluding gonadotrophins) in women with polycystic ovary syndrome. Cochrane Database of Systematic Reviews 2019, Issue 12. Art. No: CD013505. [DOI: 10.1002/14651858.CD013505] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Tso LO, Costello MF, Albuquerque LET, Andriolo RB, Macedo CR. Metformin treatment before and during IVF or ICSI in women with polycystic ovary syndrome. Cochrane Database of Systematic Reviews 2020, Issue 12. Art. No: CD006105. [DOI: 10.1002/14651858.CD006105.pub4] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Covidence. Version accessed 28 January 2025. Melbourne, Australia: Veritas Health Innovation, 2025. Available at https://www.covidence.org.
- 28.Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2020;372:n71. [DOI: 10.1136/bmj.n71] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.ISRCTN11828358. Comparison of letrozole or clomifene for ovulation induction in women with polycystic ovarian syndrome [Letrozole or clomifene, with or without metformin, for ovulation induction in women with polycystic ovary syndrome: a 2x2 factorial design randomised trial (the LOCI trial)]. https://www.isrctn.com/ISRCTN11828358 (first posted 7 January 2020).
- 30.Alfirevic Z, Kellie FJ, Weeks J, Stewart F, Jones L, Hampson L; on behalf of the Pregnancy and Childbirth Editorial Board. Identifying and handling potentially untrustworthy trials – Trustworthiness Screening Tool (TST) developed by the Cochrane Pregnancy and Childbirth Group. Cochrane Pregnancy and Childbirth, 2020. Available at https://documentation.cochrane.org/download/attachments/169771061/PGC TST FINAL.pdf?version=1&modificationDate=1694523757481&api=v2 (accessed 19 May 2025).
- 31.Carlisle JB. Data fabrication and other reasons for non-random sampling in 5087 randomised, controlled trials in anaesthetic and general medical journals. Anaesthesia 2017;72:944–5. [DOI] [PubMed] [Google Scholar]
- 32.Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook. [DOI] [PMC free article] [PubMed]
- 33.Higgins JP, Savović J, Page MJ, Elbers RG, Sterne JA. Chapter 8: Assessing risk of bias in a randomized trial (last updated October 2019). In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook.
- 34.Sterne JA, Savović J, Page MJ, Elbers RG, Blencowe NS, Boutron I, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ 2019;366:l4898. [DOI] [PubMed] [Google Scholar]
- 35.Higgins JP, Li T, Deeks JJ (editors). Chapter 6: Choosing effect measures and computing estimates of effect (last updated August 2023). In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook.
- 36.McKenzie JE, Brennan SE. Chapter 12: Synthesizing and presenting findings using other methods (last updated October 2019). In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook.
- 37.Higgins JP, Eldridge S, Li T, editor(s). Chapter 23: Including variants on randomized trials (last updated October 2019). In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook.
- 38.Page MJ, Higgins JP, Sterne JA. Chapter 13: Assessing risk of bias due to missing results in a synthesis (last updated August 2024). In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook.
- 39.Review Manager (RevMan). Version 11.0.0. The Cochrane Collaboration, 2026. Available at https://revman.cochrane.org.
- 40.Campbell M, McKenzie JE, Sowden A, Katikireddi SV, Brennan SE, Ellis S, et al. Synthesis without meta-analysis (SWiM) in systematic reviews: reporting guideline. BMJ 2020;368:l6890. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Altman DG, Bland JM. Interaction revisited: the difference between two estimates. BMJ 2003;326(7382):219. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Deeks JJ, Higgins JP, Altman DG, McKenzie JE, Veroniki AA. Chapter 10: Analysing data and undertaking meta-analyses (last updated November 2024). In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook.
- 43.Schünemann HJ, Higgins JP, Vist GE, Glasziou P, Akl EA, Skoetz N, et al. Chapter 14: Completing ‘Summary of findings’ tables and grading the certainty of the evidence (last updated August 2023). In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane, 2024. Available from https://www.cochrane.org/handbook.
- 44.GRADEpro GDT. Version accessed 28 January 2025. Hamilton (ON): McMaster University (developed by Evidence Prime), 2025. Available at https://gradepro.org.
- 45.Schünemann H, Brożek J, Guyatt G, Oxman A, editor(s). Handbook for grading the quality of evidence and the strength of recommendations using the GRADE approach (updated October 2013). GRADE Working Group, 2013. Available from https://gdt.gradepro.org/app/handbook/handbook.html.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary material 1 Search strategies
Data Availability Statement
Data sharing not applicable to this article as it is a protocol, so no datasets were generated or analysed.