Comparing ovulation induction strategies for unexplained infertility: protocol for a systematic review and network meta-analysis.

As per WHO, approximately one in six people of the reproductive age group worldwide will experience infertility in their lifetime. 1 Among couples seeking help for infertility, about 30% of them get classified as unexplained infertility despite normal ovulation, tubal patency and semen parameters. 2 As the aetiology is undefined, the treatment is often empirical rather than targeted, creating uncertainty for both patients and clinicians. Ovulation stimulation (OS), usually in combination with intrauterine insemination (IUI), remains the cornerstone of empirical therapy in unexplained infertility. Oral agents, such as clomiphene citrate and letrozole, remain the most common first-line drugs. Clomiphene citrate induces ovulation but may impair endometrial receptivity, whereas letrozole may avoid this drawback. Gonadotropins can increase pregnancy rates but are associated with higher risks of multiple gestation and ovarian hyperstimulation syndrome. 3 Professional guidelines generally recommend OS-IUI with oral agents, while discouraging gonadotropin-IUI due to safety concerns, but real-world practice continues to vary, including the use of oral–injectable combinations and differing procedural pairings. 2 4 Multiple randomised controlled trials, including a large multicentre trial by Diamond et al , have directly compared clomiphene, letrozole and gonadotropins in unexplained fertility and reported important differences in live birth rates and multiple pregnancies. 3 Targeted systematic reviews have evaluated specific contrasts, such as letrozole vs clomiphene 5 and gonadotropins vs oral agents. 6 A broader network meta-analysis by Wang and Mol included unexplained infertility but pooled across multiple treatment types, limiting its direct applicability to ovulation induction strategies. 4 More recently, Danhof et al. (2020) conducted a network meta-analysis of intrauterine insemination with ovarian stimulation, 7 and Wessel et al. (2022) performed an individual participant data meta-analysis of IUI-OS strategies. 8 While informative, these do not comprehensively evaluate combination regimens or estimate the marginal contribution of pharmacological vs procedural components across strategies. We therefore propose a protocol for systematic review and network meta-analysis focused exclusively on ovulation induction strategies for unexplained infertility, evaluating the comparative effectiveness and safety of pharmacological ovulation induction strategies (CC, letrozole and gonadotropins), used alone or in combination, with or without IUI compared against each other as well as against expectant management or placebo, where trial data exist. If sufficient data permit, we will undertake exploratory component network meta-analysis (CNMA) to disentangle the contributions of pharmacological and procedural components. Such approaches have been shown to yield more interpretable and clinically relevant evidence when evaluating complex interventions. 9 10 This review aims to provide more precise comparative estimates of benefits and harms, which may support future refinement of guidelines and help reduce uncertainty in clinical decision making. The primary objective of the study is to compare the efficacy and safety of letrozole, clomiphene citrate and gonadotropins for ovulation induction on live birth rates in women with unexplained infertility through systematic review and network meta-analysis. The secondary objectives were (1) to assess effects on multiple pregnancy, ectopic pregnancy, neonatal outcomes and to evaluate any adverse effects; (2) to explore whether intervention effects differ across subgroups defined by age, BMI, infertility duration, use of IUI, dose/regimen and cancellation policy; (3) to perform a CNMA, if feasible, to disentangle the effects of drug components (letrozole, clomiphene and gonadotropins) and procedural components (IUI and timed intercourse).

This systematic review and meta-analysis protocol is based on the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 statement 11 and Preferred Reporting Items for Systematic reviews and Meta-Analysis -Network Meta Analysis (PRISMA-NMA). 12 The completed PRISMA-P checklist and PRISMA-NMA are provided as onlinesupplemental appendices 1 2 respectively. This review has been registered in PROSPERO (registration number: CRD420251145492) We will include articles reporting results of clinical trials of ovulation induction strategies in couples with unexplained infertility as described in table 1 . We will include randomised controlled trials (RCTs) enrolling women aged 18–40 years with unexplained infertility as per ASRM guidelines, 2 defined as ≥12 months of unprotected intercourse with at least one patent fallopian tube, normal semen analysis by WHO criteria and documented ovulation. Trials will be eligible if ≥70% of participants met unexplained infertility criteria or if outcomes were separately reported for an unexplained subgroup. This was done to maximise available evidence while preserving clinical relevance. Eligible interventions include ovulation induction with letrozole, clomiphene citrate, gonadotropins (recombinant FSH, urinary FSH, hMG, etc.) and combination regimens (eg, letrozole+FSH and clomiphene+FSH) with or without IUI. Comparators may be any active agent, combination regimen, placebo or expectant management. The primary outcome is live birth per woman randomised; or if not present, then clinical and ongoing pregnancy. Secondary outcomes include ovulation, multiple pregnancy, miscarriage, ovarian hyperstimulation syndrome (OHSS), cycle cancellation, ectopic pregnancy, neonatal outcomes, time to pregnancy and cost metrics. Only English language articles will be included, and the same shall be acknowledged as a limitation. Only parallel-group RCTs (including multi-arm trials) will be included. Articles can be from any country provided they fit the inclusion criteria. Any trial conducted prior to 1980, quasi-randomised, crossover, non-randomised studies and animal trials will be excluded. ART, Assisted Reproductive Technology; hMG, Human Menopausal Gonadotropin; IUI, intrauterine insemination; IUI, Intrauterine Insemination; IVF, In-Vitro Fertilization; OHSS, Ovarian Hyperstimulation Syndrome; PCOS, Polycystic Ovary Syndrome; RCTs, randomised controlled trials; rFSH, Recombinant-Follicle Stimulating Hormone. A comprehensive search will be conducted on databases like MEDLINE (PubMed), Embase, Scopus, Web of Science, COCHRANE Library and trial registry ClinicalTrials.gov. Databases will be searched from inception to September 2025. The review commenced on 11 September 2025, and the tentative completion date is 3 March 2026. The PubMed search strategy is shown below; full strategies for all databases are provided in online supplemental appendix 3 . The PubMed search strategy is shown in table 2 ; full strategies for all databases are provided in online supplemental appendix 3 . Additionally, to minimise publication bias, we will search through grey literature sources like postgraduate theses, conference abstracts and dissertations. These sources will be included if sufficient methodological details are available to permit risk of bias assessment. Previously done systematic reviews and meta-analyses will be checked for additional references; relevant references will be screened as supplemental sources. The studies that are recognised will be imported into a web-based systematic review management platform ( Nested knowledge ) for de-duplication and screening. Zotero will be used as a backup for citation management. Two reviewers will independently screen the titles and the abstracts to identify the eligible studies. Next, each of the selected studies will be reviewed comprehensively by two reviewers to check if they truly fit the inclusion criteria and their references will be reviewed for additional relevant studies. Any discrepancies will be resolved by discussion or, if unresolved, adjudication with a third reviewer. Studies screened through the above process will undergo a thorough data extraction process. Data will be independently extracted by two reviewers using a pre-piloted data extraction form. Any discrepancies will be resolved by discussion with a third reviewer. The domains under which the data will be extracted are given in detail in table 3 . BMI, Body Mass Index; OHSS, Ovarian Hyperstimulation Syndrome; RCT, randomised controlled trial; USG, Ultrasonography. The primary outcome reported will be live birth per woman randomised, as it is the ideal ending point for treatment of infertility. Live birth will be defined as delivery of a live born infant beyond or at 24 weeks of gestation. But due to limited reporting of live births in several trials, alternative primary outcomes will be considered. If live birth is unavailable, ongoing pregnancy ( > 12 weeks of gestation) will be given preference over clinical pregnancy (gestational sac confirmed by ultrasound). If multiple overlapping outcomes are reported in the same trial, the predetermined outcome hierarchy will be followed. Secondary outcomes include ovulation rate, multiple pregnancy (number and order of pregnancy), miscarriages, neonatal outcomes, OHSS, ectopic pregnancy, time to pregnancy, cost outcomes, cycle cancellations and maternal adverse events. OHSS definitions will vary greatly across trials, and the trialist’s definition shall be extracted verbatim, and where feasible, grouped into standardised categories, according to Golan criteria. Cost data will be extracted in the original currency and year reported. Neonatal outcomes will be extracted and prioritised as pre-term birth, mean birth weight, neonatal death and congenital anomalies. To ensure consistency throughout the study, the operational definition of each outcome is specified in detail in online supplemental appendix 4 . The risk of bias for randomised controlled trials will be assessed using the Cochrane Risk of bias 2.0 (RoB-2) 13 tool across five domains: randomisation process, deviation from intended interventions, missing outcome data, measurements of outcomes and selection of the reported result. Two reviewers will independently apply the tool for each outcome, with any discrepancies being resolved by thorough discussion or arbitration with a third reviewer. Judgements will be categorised as low risk, some concerns or high risk of bias, and results will be summarised in tabular form and traffic-light plots. Trials classified with a high risk of bias will be downweighted in the certainty of evidence assessment and excluded for sensitivity assessment to examine the robustness of results. Publication bias will be assessed using funnel plots, comparison-adjusted funnel plots and Egger’s test where feasible. When there are fewer studies, we will interpret funnel plot asymmetry cautiously and consider alternative explanations, such as heterogeneity or selective outcome reporting. We will first perform pairwise random-effects meta-analysis for direct comparisons wherever more than two RCTs are available using risk ratio (RRs) with 95% CIs for dichotomous outcomes and mean or standardised mean differences (MD/SMD) for continuous outcomes. Restricted Maximum Likelihood will be the default estimator for between-study variance, with alternative estimators and Bayesian hierarchical models for sensitivity analysis. Heterogeneity will be assessed with τ², I² and χ² statistics. Analysis will proceed in a stepwise manner, beginning with pairwise meta-analysis, followed by network meta-analysis where feasible, and component NMA, if assumptions are met. For multi-arm trials, intervention groups will be combined where clinically appropriate. Where a single comparator is used against multiple distinct interventions, the comparator arm will be proportionately split to avoid double counting. For rare outcomes (eg, ectopic pregnancy, neonatal outcomes, etc.), we will use robust methods to sparse data (eg, Peto OR, beta binomial or Bayesian hierarchical rare event models) rather than conventional continuity corrections. Neonatal outcomes will be prioritised as specified in the outcomes and prioritisation section, and results will be narratively summarised or pooled if only more than three RCTs report comparable outcomes. Cost data would be converted to 2025 US dollars using purchasing power parity. Due to anticipated heterogeneity, results will be summarised descriptively, and pooling will only be attempted if more than three RCTs report directly comparable measures. If the network of interventions is sufficiently connected and clinical transitivity appears plausible across trials, we will attempt a random-effects network meta-analysis (NMA) using both direct and indirect evidence. Transitivity will be tested by comparing between studies distributions of potential effect modifiers (eg, mean age, cancellation policies and proportion with unexplained infertility), while inconsistency will be assessed using both local (node-splitting) and global (design-by-treatment interaction) approaches. Results will be reported as league tables, forest plots and ranking probabilities (surface under the cumulative ranking curve, SUCRA), with absolute risk estimates. We will describe the geometry of the evidence network for each outcome using network plots (nodes representing interventions and edges representing direct comparisons), tabulate the number of studies and participants per pairwise comparison, and report disconnected nodes or sparse areas of the network. Pooling will be limited to comparisons involving at least two RCTs; otherwise, findings will be narratively summarised. Where sufficient permutations of component combinations across trials are available and additive assumptions appear plausible, we will also carry out a component network meta-analysis (CNMA) for the estimation of marginal effects of drug and procedural components. Regimens will be broken down into components (eg, letrozole, clomiphene, gonadotropins, IUI), modelled under additive log-odds assumptions, with interaction terms explored if feasible. Bayesian hierarchical CNMA models will be run in JAGS with vague priors (βk~N(0,10²); τ~Uniform (0,2)), and results cross-checked against frequentist implementations. Outputs will be component and regimen effect estimates with 95% credible intervals, and contribution plots. If the network is too sparse or assumptions of additivity are not met, results will not be interpreted and presented as narrative summaries. The certainty of evidence will be appraised using the CINeMA framework, which applies GRADE principles to network meta-analysis. Confidence in effect estimates will be assessed across six domains: within-study bias, reporting bias, indirectness, imprecision, heterogeneity and incoherence. For each primary and key secondary outcome, summary of findings tables for primary and key secondary outcomes will be generated to provide transparent ratings of certainty (high, moderate, low or very low). Sensitivity analyses excluding trials at high risk of bias will be undertaken, and interpretations will be framed considering the overall certainty. Leave-one out sensitivity analyses will be conducted to evaluate the influence of individual trials on pooled estimates if feasible. Subgroup analyses will first focus on primary modifiers (age, BMI and cancellation criteria). Secondary variable (dose regimen, region and duration of infertility) will only be analysed if sufficient data (> 5 trials) are available. We will evaluate effect modification using meta-regression (random effects). Interaction p-values and change in between-study variance (τ²) will be reported; subgroup findings will be interpreted cautiously and only when consistent across models and clinically plausible. We will limit subgroup testing to the pre-specified list above. This review and meta-analysis will be a secondary analysis of data from previously published randomised controlled trials. No new human data will be collected, and ethics approval is therefore not required. If individual participant data are shared with the review team, they will be stored and analysed under data-user agreements; only de-identified datasets will be used, and analysis will be performed on secure institutional servers in accordance with the original consent and institutional policies, and results will be reported in aggregate. The results of the study will be published in peer-reviewed journals and presented at academic conferences. All extraction sheets (CSV/Excel), analysis scripts (R markdown) and output figures will be archived. De-identified IPD (if acquired) will be stored under controlled access and only shared if PI permissions permit on request to the corresponding author. Patients or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

This review is designed to provide a comprehensive synthesis of ovulation-induction strategies in unexplained infertility. This review is restricted to English-language publications, and we acknowledge the potential for language bias. Anticipated challenges include inconsistent reporting of live birth, our preferred primary outcome, which may necessitate reliance on ongoing or clinical pregnancy as prespecified surrogates. Variability in trial definitions of outcomes such as ovulation, OHSS or cancellation criteria may further limit comparability. Some interventions, especially combination regimens or historical approaches, may have sparse data, reducing the precision of effect estimates or creating disconnected evidence networks. Finally, network meta-analysis relies on assumptions of transitivity and consistency, which may not always be met. To mitigate these issues, we will apply RoB-2 for risk of bias and CINeMA to assess certainty of evidence, and conduct prespecified sensitivity and subgroup analyses. The application of component network meta-analysis, if feasible, represents a novel contribution, allowing estimation of the marginal effects of pharmacological vs procedural components. These analyses will generate clinically interpretable evidence to guide counselling, treatment selection and guideline development in unexplained infertility.