Chapter 22: Prospective, collaborative approaches to meta-analysis

Anna Lene Seidler, Peter J Godolphin, Kylie E Hunter, Claire Vale, Luke Wolfenden, David J Fisher, Lisa M Askie, Jesse A Berlin, Jayne F Tierney on behalf of the Cochrane Prospective Meta-Analysis Methods Group

Cite this chapter as: Seidler AL, Godolphin PJ, Hunter KE, Vale C, Wolfenden L, Fisher DJ, et al, editor(s), on behalf of the Prospective Meta-Analysis Methods Group. Chapter 22: Prospective, collaborative approaches to meta-analysis (last updated November 2026). 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 6.5.1 Cochrane, 2026. Available from https://www.cochrane.org/handbook.

22.1 Introduction to prospective, collaborative meta-analysis

22.1.1 What is a prospective, collaborative meta-analysis?

A prospective meta-analysis (PMA) is a systematic review and meta-analysis of studies that are identified, evaluated, and determined to be eligible for the meta-analysis before the relevant results of any of those studies become known (Seidler et al 2019). While many of the methods are akin to those for standard reviews, the order in which they occur can vary and there are also some key differences (Seidler et al 2019). In particular, collaboration and liaison between the review team and investigators from each study is key to the success of PMA.

In this chapter, we mainly focus on PMAs using aggregate data or individual participant data (IPD) from randomized trials. Most of the same considerations will also apply to systematic reviews of other types of studies, covered in Section 22.6.4.

22.1.2 What are the advantages of PMA?

A central tenet of randomized trials is that they are designed prospectively, with the hypothesis, sample size, outcomes, and analyses specified before data collection and analysis, to avoid bias. In contrast, most systematic reviews and meta-analyses are undertaken retrospectively, after at least some of the eligible studies have been completed and reported. Consequently, methods such as trial selection, analysis plans, and interpretation can easily be influenced by researchers’ knowledge of existing study results, introducing potential bias (Table 1). For example, eligibility criteria might be defined to focus on trials with more favourable results, or the planned analyses may be influenced by the findings for one or more outcomes or subgroups of participants, which may skew findings. Such decisions may be one of the reasons that multiple meta-analyses on the same topic can reach conflicting conclusions (Siontis et al 2013). Also, many systematic reviews continue to rely on aggregate data extracted from publications. This means the conclusions of these reviews may need to be driven by the data available rather than the clinical questions that need to be addressed. They may also be affected by the long-known and extensive problems arising from publication (Dickersin et al 1987; Simes 1986) and other reporting biases (Page et al 2014; Showell et al 2024).

Adopting a prospective approach to meta-analysis reduces the potential for bias in the review and meta-analysis methods, because eligible studies are identified, and these methods are planned, before the results from the included studies are known (Seidler et al 2019). Thus, PMA has similar advantages to prospective multicentre clinical trials. Collaboration and liaison between the core project team and study investigators are central elements of the PMA approach and bring many other advantages to review and meta-analysis conduct. For example, the review team gain a better understanding of the design and conduct of individual studies, informing eligibility assessment, and improving the completeness and reliability of risk of bias assessments and interpretation of results (see Section 22.2.3 and Section 22.2.4 for more information on the project team). Designing a PMA whilst eligible studies are planned or in early data collection phases means investigators can work together to harmonize study designs, data collection, and other processes (Seidler et al 2019). Collaborating from such an early stage also means investigators can plan for data sharing from the start, which is why many prospective meta-analyses are performed with IPD (see Chapter 26). Even if a PMA is initiated when studies are nearer completion or in follow-up, steps can be taken to harmonize data definitions and analyses as far as possible.

22.1.3 When is PMA appropriate?

Prospective meta-analyses can be considered for any review question where evidence from new relevant studies is expected to emerge. For example, if a large multicentre trial was needed to achieve a sufficient sample size for an important clinical question, but was unfeasible to conduct, a PMA might be initiated to design and run similar, but separately funded trials. This was the case in the Neonatal Oxygenation Prospective Meta-analysis Collaboration (NeOProM) PMA, which investigated the optimal oxygen saturation range for preterm infants (Askie et al 2018).

Some PMAs are initiated after study investigators have learned that they are planning or conducting studies similar to others, or a funding/guideline body has realised that multiple similar ongoing studies exist. In these scenarios, systematic reviewers are enlisted to help collaboratively and prospectively plan to combine data across the planned studies. For example, once aware of the plethora of trials being conducted to assess the effects of various treatments for patients hospitalized with COVID-19, the World Health Organization (WHO) established the Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group to examine the effects of specific treatments in a series of PMAs (Vale 2024 et al; WHO 2020; WHO 2021; WHO 2025a) that would provide robust evidence to inform their clinical guideline (WHO 2025b). Alternatively, if systematic reviewers can identify similar studies in a particular clinical area before they have completed or produced results, they can take the opportunity to use prospective, collaborative methods to synthesize their data or results. This PMA approach has been particularly successful in assessing treatments for metastatic cancer (Burdett et al 2019; Rydzewska et al 2017; Vale et al 2016; Vale et al 2020).

A fully prospective meta-analysis may not be the method of choice in some situations, for example, because some important, large studies have already been published. However, even if some studies have already reported results, it may be possible to plan a partially prospective, collaborative meta-analysis, or to plan a prospective collaborative update, known as a ‘nested’ prospective meta-analysis (Section 22.5.2). Both approaches can still reduce bias and otherwise improve the quality of evidence synthesis.

22.1.4 How to choose between a PMA of individual participant or aggregate data?

PMAs based on IPD involve the collection of line-by-line data for each of the participants and their outcomes from all eligible studies (Tierney et al 2015; Tierney et al 2021a; Tudur Smith et al 2016). This helps circumvent the reporting biases associated with using published aggregate data (Stewart et al 2005) and permits detailed data checking and more thorough risk of bias and integrity assessments (Hunter et al 2024; Sotiropoulos et al 2025; Tierney et al 2015; Tierney et al 2021a). In addition, the IPD approach allows more flexible and detailed analysis, including the ability to explore how intervention effects might vary by participant characteristics (Tierney et al 2015; Tierney et al 2021a). Such investigations of variable intervention effects can otherwise be subject to ecological bias (also known as aggregation bias), i.e. systematic errors occurring through drawing conclusions using across-trial evidence about an individual-level relationship (Berlin 2002). IPD-PMA have been used across a number of health areas. For example, in early childhood obesity prevention, prospective accumulation of IPD enabled standardized measurement of body mass index (BMI) z-scores across countries, age groups, and time points, and enabled appropriate effect modification analysis with continuous variables (e.g. age) (Askie et al 2020; Hunter et al 2025). Such standardization and analyses would have been difficult or impossible with aggregate data alone.

An IPD-PMA might be most appropriate when there is the need to bring together detailed data for more thorough or complex analyses and time is less pressing (Tierney et al 2015; Tierney et al 2021a). However, for some reviews the granularity of IPD might not be needed or the collection of IPD might not be feasible, due to time or resource constraints. In these cases, or when there is a rapid or urgent need to assess whether a treatment works, such as during the COVID-19 pandemic (WHO 2020; WHO 2021), the collation of a pre-agreed set of aggregate data might be preferable because it offers a quicker, less resource-intensive alternative. One approach is the Framework for Adaptive Meta-analysis (FAME, see Section 22.5.1) (Tierney et al 2021b), which is similarly prospective and collaborative in nature, but unlike standard meta-analyses that tend to rely on published aggregate data, harmonized and detailed results (i.e. effect sizes and corresponding measures of variation) are sought directly from investigators. This includes the ability to assess appropriately whether effects vary by participant characteristics (Fisher et al 2017; Godolphin et al 2023), avoiding ecological bias (Berlin et al 2002). Analysis time points are carefully pre-planned based on accumulating information to assess the earliest possible timing for reliable meta-analysis. This lessens reporting biases, facilitates and increases the thoroughness of the analyses, and improves interpretation of findings, compared to standard reviews. See Chapter 26 (Section ‘When is an IPD review appropriate’) and Tierney et al 2021a for further guidance on when to choose an IPD approach.

Ideally, the decision to base a PMA on IPD, aggregate data, or both should be pre-planned when setting out the scope and research question, although this may change over time. For instance, a PMA collaboration may initially plan to collect IPD, but change to aggregate data, due to administrative or legislative barriers to data sharing or to get a quicker answer. This was the case in a PMA of radiotherapy scheduling for localized prostate cancer (Vale et al 2020). Alternatively, PMA collaborations may plan to collect aggregate data in the first instance to get an early indicator of any overall or subgroup effects, subsequently proceeding to collect IPD if this is necessary to improve harmonization or to perform more complex or more in-depth analysis. It is important to carefully document these decisions, their time point, and reasons.

Table 1: Aspects of conventional, retrospective systematic reviews and meta-analyses that can be biased by prior knowledge of trial results (adapted from Tierney et al 2021b)

22.1.5 How do PMA and standard Cochrane review methods differ?

Many of the methods for PMA are very similar to those of standard Cochrane reviews. For example, PMA involves preparation of a detailed protocol that includes: the objectives; specific questions to be addressed; study and participant eligibility criteria; the descriptive, baseline, and outcome data to be collected and how this will be managed; and the planned analyses, as well as other methods, such as risk of bias assessments. In contrast to a standard Cochrane review, however, PMA involves forming a collaboration of study investigators in advance of study results being known and managing the collaboration throughout the project. These study investigators are then closely involved in a number of steps, including preparing the protocol, planning the timing of meta-analysis, and coordinating the reporting of study and meta-analysis results. Another difference is that the searches focus on finding ongoing and unpublished studies, which often precede the writing of the full protocol. This is because the core project team need to identify the researchers responsible for relevant trials to bring them into the wider PMA team. The process of collecting, checking, and analysing IPD or detailed aggregate data is more involved than extracting summary data from publications, but also leads to more comprehensive and harmonized data.

Therefore, a PMA team requires wider expertise than is typically needed for a standard review. This includes the ability to manage a collaboration of study investigators, oversee governance, and, if IPD are to be used, to manage and analyse IPD (see Section 22.2.3, Chapter 26 (Section ‘How are IPD reviews organized?’), and Stewart et al 2021).

It is possible to conduct a PMA within the Cochrane framework. For example, the NeOProM Collaboration published a Cochrane review based on a prospective meta-analysis (Askie et al 2017). However, published examples of PMA within Cochrane are few to date, which may be due partly to Cochrane systems and policies being designed around more conventional review workflows. PMA often requires greater flexibility, particularly in searching for studies, and in developing a protocol in collaboration with trial investigators (Seidler et al 2020a). As Cochrane continues to support newer methods such as living approaches to evidence synthesis, opportunities to conduct prospective meta-analyses within the Cochrane Library may improve.

22.1.6 What are the different types of prospective, collaborative meta-analysis?

Whilst PMA cover the broad concept of prospectively and collaboratively planning the synthesis of accumulating evidence, there are now several related variants (Box 1).

22.1.7 Chapter overview

In this chapter, we focus on key features that are unique to and unite all prospective, collaborative meta-analyses in contrast with standard Cochrane reviews. We cover three main phases: designing and planning (Section 22.2), conducting (Section 22.3), and reporting and disseminating (Section 22.4). Where appropriate, we highlight differences pertaining to the nature of the data (IPD versus aggregate data) across these phases. We discuss related approaches along the continuum from fully and partially prospective PMA through to other collaborative approaches such as collaborative network trials (Section 22.5). Finally, we describe other PMA uses, for example to combine observational studies or test systematic review methods, and the benefits of a PMA approach in a health emergency setting (Section 22.6).

22.2 Designing and planning a prospective, collaborative meta-analysis

In PMA, defining the broad question to be addressed and identifying eligible studies tends to precede detailed protocol development. This is because PMA is predicated on a question that needs to be addressed with multiple relevant planned or ongoing studies. In addition, this will help with the formation of a collaboration of relevant study investigators. This makes it possible to work with these investigators to develop a protocol prospectively. The key clinical questions and the data and analyses needed to address these questions can be pre-specified, free from influence of known study results.

22.2.1 Defining the overarching research question and eligibility criteria

Planning the research question, eligibility criteria, and studies to be included before their results are known is a critical part of any PMA. It minimizes the risk of publication, study selection, and other biases related to knowledge of the results of a potentially eligible study. That said, determining the scope of PMAs of interventional studies and their inclusion criteria is very similar to a standard Cochrane review (see Chapter 2 and Chapter 3) or a retrospective review of IPD. The PMA inclusion criteria do not need to be the same as the inclusion criteria of the included studies: they may be narrower or broader. For example, the PMA may only include a sub-population of participants included in a study, such as participants within a certain age range.

22.2.2 Searching for eligible studies

As for any Cochrane review, systematic searches to identify all eligible studies are vital (see Chapter 4). The key difference is that in PMA, the focus is on identifying ongoing or otherwise not yet completed studies. Clinical trial registries are therefore a key resource for identifying eligible studies (Hunter et al 2022a). General bibliographic databases should also be searched to identify existing studies, published protocols, and registration records (see Chapter 4 (Section ‘Bibliographic databases’).

While the Cochrane Central Register of Controlled Trials (CENTRAL; https://www.cochranelibrary.com/central) has included registration records sourced from ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP) since 2019, searching CENTRAL has low sensitivity for identifying these, and thus should not substitute direct registry searches (Banno et al 2020). In addition to these systematic searches, reaching out to researchers in the field can identify other planned or ongoing studies, or may encourage researchers to plan new studies to add to the PMA. Thus, identifying studies for a PMA can also be quite a proactive process, which is a key difference to a standard Cochrane review. If investigators of identified studies do not agree to collaborate and harmonize their protocols, published aggregate data from such studies may still be included either in the main meta-analysis or in sensitivity analysis by extracting their data from publications, where possible.

22.2.2.1 Searching trial registries

As of 2026, 18 clinical trials registries provide data to the WHO ICTRP (https://trialsearch.who.int) and are recognized by the International Committee of Medical Journal Editors (ICMJE). Comprehensive guidance on searching clinical trial registries is available (Hunter et al 2022a) (see Chapter 4 (Sections ‘Ongoing studies and unpublished data sources’ and ‘Trials registers and trials results registers’) and Chapter 4 S1: Technical Supplement). As a minimum, searches should include the WHO ICTRP and ClinicalTrials.gov (Hunter et al 2022a; Glanville et al 2014), but other registries may also be appropriate to a particular research question, for example, region- or disease-specific registries.

Typically, search strategies for clinical trial registries are much simpler than for bibliographical databases, due to the limited availability and functionality of their search tools. One or two key concepts should be selected from the PICO framework (e.g. population and intervention), and a list of synonyms or alternative terms should be generated for each. It may be useful to review the search strategies of similar studies for ideas. Review authors should start with the most specific concept, with synonyms combined using the ‘OR’ operator or similar. If too many results are retrieved, the search can be refined by adding a second concept and combining it with the first, for example using the ‘AND’ operator. Search interfaces vary substantially across registries and therefore search strategies need to be tailored to the specific registry (Hunter et al 2022a). If some eligible studies are already known, it is useful to test the search strategy to see if these are identified. If not, iterative refinements should be made until they are retrieved.

Searches filters (e.g. by study type, participant age, or recruitment status) should be avoided because study categorization in registries is not always accurate. For instance, randomized controlled trials (RCTs) may be incorrectly categorized and would be missed if an RCT filter was applied.

Since registration records are not in bibliographical format, information may be lost when exporting them to screening software, such as Covidence. This issue can be mitigated by using registry-specific Endnote filters (see https://endnote.com/downloads/filters/). For access to the most detailed and recent information, registration records should be screened at the source registry website.

22.2.2.2 Searching for completed studies

Although PMA only include studies that are ongoing or in follow-up, searches should attempt to identify all studies of relevance, irrespective of publication status. Completed studies may be summarized to provide context for the interpretation of results or included in a nested PMA (Section 22.5.2).

22.2.2.3 Timing and frequency of searches

Ideally, searches should be repeated on a regular basis to identify new eligible studies, and to update information on those found already, as registry records are updated. Investigators of newly identified studies should be invited to contribute and harmonize their collected outcomes to the PMA protocol. The appropriate timing and frequency should be determined by available resources and whether the research field is fast or slow-moving, and this should be detailed in the PMA protocol. Updating can be limited to specific dates to reduce the records needed to screen and avoid duplication of effort.

22.2.3 Establishing the PMA project team

In the early stages, a core project team should be assembled, which has the expertise needed to lead a PMA, including developing and overseeing the collaboration and managing the project on a day-to-day basis. The team should include individuals who specialize in evidence synthesis, data collection, management, and statistical analysis, and who have experience of managing collaborative groups of clinicians and other researchers. It is also advantageous to include members with a track record in the relevant healthcare area. As collaboration agreements and data sharing agreements may be required (see Section 22.2.4), individuals with legal or contract management expertise may also need to be called upon, though not necessarily required as part of the project team.

Whilst for some PMA the core project team may include some of the trial investigators, potential conflicts of interest need to be handled carefully. For this reason, core project teams may decide not to include the trialists or at least ensure that data collation and analysis are performed independently of them, and that the role and influence of trial investigators need to be reported on. However, close communication and involvement of trial investigators is an important part of the PMA process.

22.2.4 Initiating and maintaining the PMA collaboration

Key to the success of a PMA is the formation of a collaborative group, with the structure and roles and responsibilities of those involved being well-defined. An example of different roles and interest holders in a collaboration structure can be found in Box 2.

The project team should engage with study investigators as early as possible, and importantly, while studies are either ongoing or are completed. This means results are unknown to anyone, with the exception of, potentially, the trial data safety monitoring committee. Invitations to collaborate should clearly articulate the collaborative nature of the endeavour that seeks to complement included studies and include a brief outline of the project. Transparent processes and a priori agreements are crucial for building and maintaining trust within a PMA collaboration.

A collaborators’ agreement and governance plan can cement the PMA collaboration. Such an agreement ensures that expectations between the project team and study investigators are clear and agreed upon in advance. It is important that all those engaged understand and have clearly documented roles and responsibilities (and resource requirements). As a PMA includes ongoing studies that are yet to report results, agreements that protect the confidentiality of data and other information pertaining to the ongoing studies may be required. See Box 3 for details.

Regular meetings are vital for sustaining a collaborative group and for the success of PMAs, some of which may span many years. These meetings can be used to develop the PMA protocol, resolve disagreements, update investigators on progress, and prompt study investigators for data/information that is required. Ensuring that collaborators are fully engaged in protocol development can abate requests for post hoc analyses once the PMA results are presented. This can be particularly helpful where conclusions arising from the PMA differ from those of some included studies.  

Occasionally, investigators may decline to participate in a PMA collaboration or provide data from their studies. It remains important to track the progress of such studies through updates to trial register listings, study websites, or via other collaborators. Including published results from such studies to supplement the PMA findings (see Section 22.3.1) or, at the very least, noting and considering results from such studies, is a vital part of interpreting PMA results.

22.2.5 Developing a PMA protocol

A crucial feature of a PMA is that the methods are pre-specified in detail in a protocol, before eligible study results are known. This reduces the potential for bias and means that those conducting the studies can participate fully in shaping the plans and incorporating their clinical knowledge. It is essential to ‘date-stamp’ the protocol and make it publicly available. For Cochrane reviews with PMA, protocols may be published in the Cochrane Library. For reviews conducted outside of Cochrane, this can happen either on a study website, as part of the registration for the review, or as a publication. As formal publication of the protocol can take considerable time, registering an outline of the protocol on, for example, PROSPERO is an important first step, provides a prospective registration number, and makes the PMA findable in the PROSPERO database. Currently, PROSPERO is the most widely used registry for systematic reviews and enables explicit labelling of a review as a PMA. Initial registration and a full protocol are both important to the planning of a PMA, but neither of those steps replaces the other.

Conceptually, developing a protocol for a PMA is similar to other systematic reviews. The preferred reporting items for systematic reviews and meta-analyses extension for protocols (PRISMA-P) offers a useful template (Moher et al 2015), but key adaptations are needed to accommodate the prospective elements (see Box 4) and whether the PMA is based on IPD or aggregate data.

Some of these items can only be finalized at a later stage in the PMA process. For instance, after completing searches and initiating the collaboration, study investigators are commonly asked to provide input into the protocol before it is finalized and published to ensure their commitment to the protocol, and agreement to share their results when their studies have been completed.

Box 4. Additional items to include in a prospective meta-analysis protocol

Several key items should be included in a PMA protocol in addition to those outlined in PRISMA-P (Shamseer et al 2015):

Search methods for planned and ongoing studies

This would include how planned and ongoing studies are identified (for example, by searching prospective trial registries) and how potential collaborators will be or have been contacted to participate (Section 22.2.2). If applicable, plans to regularly update the search and actions to be taken if subsequent studies are identified while the PMA is in progress should be detailed.

List of included studies

For studies already identified, their registration identifiers, brief study characteristics, anticipated (or actual) number of participants, and timelines for each participating study should be provided, along with a statement that their results related to the PMA research question are not yet known to anyone outside data safety monitoring committees.

Data to be collected

The protocol should specify which outcomes and baseline characteristics will be collected for studies in the PMA, and when and how these should be collected, measured, and defined. These may be affected by the nature of the data being sought (e.g. aggregate data or IPD), which should also be clearly outlined (Section 22.2.7).

The protocol should also describe what approach will be taken if some study investigators do not provide participant-level data (for an IPD-PMA) or detailed aggregate data (for a PMA of aggregate data), and if summary data will be accepted for these studies (Section 22.3.1).

Statistical analysis

It is important to specify sample size calculations or anticipated statistical power, since this informs and provides justification for pre-specification of the timing of analysis, including any triggers that will be used to determine when to proceed with analysis (see Section 22.2.8.3)

An outline of all the analysis methods should be clearly specified and include: methods for checking the requested aggregate data or IPD; methods to assess the risk of bias and integrity of included studies; synthesis methods for evaluating treatments effects, including those for exploring variations in effects by study or participant characteristics, and methods for quantifying and accounting for heterogeneity (Section 22.2.8, Section 22.3.2, and Section 22.3.3).

Future updates

Details on whether there is an intention to update the PMA, for example, at regular intervals via ongoing cycles of data collection (e.g. five yearly) or by other triggers (e.g. accumulating information size) should be provided.

Collaboration management and publication policy

Collaboration management and publication policy (Section 22.2.4, Section 22.4.1, and Section 22.4.2) should be specified, including details of the project team and any data monitoring committees (see Section 22.2.9).

22.2.6 Ethical approval

For PMA using IPD, some institutions will require ethical approval for holding participant data (see Chapter 26, Section ‘Data security and confidentiality’). This should be explored with the relevant ethics board. For PMA based on aggregate data, no ethical approval is required, since the individual included studies will have obtained local ethical approval and are only sharing summary level information.

22.2.7 Harmonizing included studies

PMA provides an opportunity for study investigators to agree on harmonization of the design of studies and/or their data, thereby reducing clinical heterogeneity and potentially the variability of study findings. In addition, this can lead to greatly increased availability of outcome data and thus enhanced statistical power in meta-analysis (Seidler et al 2021a).

22.2.7.1 Harmonizing data collection

If a PMA is initiated early enough, core data collection can be standardized. By harmonizing relevant elements of their protocols and data collection methods, for example, the same outcomes and definitions, measurement instruments, and analysis time points can be used across all or most of the identified studies. If studies are at a later stage, and cannot be standardized fully, then outcomes and baseline information, as agreed by the collaborative group, should be requested in a harmonized fashion.

22.2.7.2 Harmonizing outcomes and related data

When considering outcome harmonization, a first step should be to check whether a relevant core outcome set exists already (https://www.comet-initiative.org/), and if a core outcome measurement set exists for how to measure these outcomes (COSMIN). If it does, then core outcomes should be prioritized for the PMA, so long as included studies are evaluating these consistently. If a core outcome set does not exist, then outcomes can be discussed and prioritized among the PMA collaborators, and a list of these and other variables and their definitions agreed upon. A similar approach should be adopted for subgroup definitions, particularly for PMAs based on aggregate data (see Section 22.3.1), since this will have a direct impact on data collection and analysis.

If it is not possible to get fully harmonized data, the PMA project team may be able to map existing study outcomes to the desired PMA outcomes with a sufficient degree of accuracy, so that these data can be included.

Importantly, studies can be included in the PMA even if not all outcomes (or other data) have been collected and harmonized. Moreover, individual investigators are always free to collect additional data of relevance to their study setting or questions (e.g. locally important questions, or explorations into biological mechanisms).

22.2.8 Planning statistical analyses

22.2.8.1 General approach

A major advantage of PMA is that planned analyses are driven by pre-specified key clinical questions that have been agreed by the project team. This enables flexible and comprehensive analyses, such as analyses based on multiple outcomes, analysis of rare but important adverse effects, analyses by participant characteristics, and a range of sensitivity analyses. Whilst this applies to both PMA based on aggregate data and IPD, PMA based on IPD offers a range of additional analytical opportunities (see Chapter 26 (Section ‘Analysis advantages’)). The planned analyses inform the PMA data requirements (see Section 22.2.7 and Section 22.3.1); in contrast, analyses in standard Cochrane reviews are typically restricted to whichever data are reported and the limitations these impose.

Although PMAs use similar analysis methods to those employed in retrospective reviews of aggregate data (see Chapter 10) or IPD (see Chapter 26, Section ‘Analysis of IPD’), a major strength is that statistical hypotheses are pre-specified free of influence from the accumulating data. To minimize bias in PMA, all analytical methods must be pre-specified in sufficient detail in the review protocol (see Box 4) before the results of any of the included studies are known. If deficiencies with the analysis plan are uncovered after the protocol is completed, then amendments should still be made, if possible, before generating final (especially comparative) results and these amendments should be described in the report of the PMA.

The collaborative relationships developed during the PMA ensure a priori selection of primary and key secondary outcomes for the PMA and specification of subgroup analyses that are clinically relevant and feasible. These also encourage acceptance and agreement of the study investigators for the prospectively planned analyses (see below, and Section 22.3.1), allow collective planning for data collection (see Section 22.3.1), and help determine the time point of analyses (see Section 22.2.8.3).

22.2.8.2 Developing a statistical analysis plan

Much like prospective randomized trials (Gamble et al 2017), the protocol for a PMA should be supplemented with a comprehensive Statistical Analysis Plan (SAP) (Tierney et al 2015), which contains clear details on all planned analyses (Debray et al 2015), so they may be replicated without ambiguity. An SAP typically expands on data and analytical aspects of the review protocol, for example detailing any specific considerations and hypotheses for subgroup analyses, planned sensitivity analyses, how studies that do not agree to collaborate will be dealt with, and determination of the optimal time point for meta-analysis (see Section 22.2.8.3). Pre-specification of statistical methods before results are known is an important instrument to reduce bias. Whilst a detailed SAP is particularly important for an IPD-PMA due to the breath of analyses possible, SAPs are also recommended for an aggregate data PMA to pre-specify hypotheses and analyses as mentioned above. Much like the PMA protocol, an SAP should be made publicly available in a time-stamped and version-controlled document, before data are merged and any analyses undertaken. Unplanned, post hoc analyses may be used to further explain or add to the results, but such exploratory analyses should be justified and clearly reported as such.

22.2.8.3 Determining the time point for meta-analysis

A major advantage of PMA is the ability to anticipate how data are accruing within and across studies over time, which can be used to inform a priori decision-making regarding the timing of meta-analysis, importantly before any study results are known. Planning can then be optimized to balance both the strength of evidence and timeliness. Consensus on timing should be reached with study investigators, using information concerning sample size, anticipated (and actual) accrual rates, and timing of planned interim or final analysis. Prospective calculations of statistical power relating to the main effects and effect modification analyses are an important tool to inform these decisions (Riley and Ensor 2021). It is important to pre-define any decision rules and triggers and include them in the protocol (Section 22.2.5) and SAP.

In the FAME approach to aggregate data PMA, a key additional consideration is anticipating the earliest future time point when there will likely be sufficient evidence from eligible studies to produce reliable meta-analysis (Tierney 2021b). This is based on when the information from these studies would likely represent a substantial proportion of the total potential information (‘relative’ information size) and give the pre-defined power for assessing effects on the main outcome (‘absolute’ information size) (Tierney 2021b). Using both types of information size allows the timing of meta-analysis to coincide with a potentially definitive result and may avoid the need to wait many months or years for small amounts of study data or results that would provide only modest gains in certainty. Again, it is important that decisions and analysis triggers are pre-specified in the protocol, in advance of study results being known.

There is a potential risk of time-lag bias if a PMA final analysis occurs earlier or includes fewer studies than originally planned, for example because some have not published or are unable to provide data until much later than expected (Page et al 2021a). Similarly, if studies that are available have stopped early for benefit this could also risk bias. In each case, the pooled estimate could overestimate the true treatment effect (Showell 2024), even when nominal sample-size or power thresholds have been met. To mitigate this risk, decisions regarding the timing of the PMA should not be based solely on pre-specified power criteria but ideally pre-specify which studies it will include and wait until these have data available. However, where it is not feasible to include all the studies specified in the protocol, this limitation should be clearly explained in the PMA report. In addition, sensitivity analysis could be used to explore the influence of studies that have stopped early on the pooled estimate. Note that these issues may also affect both standard and living systematic reviews.

22.2.9 Establishing an overarching data monitoring committee for a PMA

Randomized trials frequently include a plan for interim analyses of data, particularly to determine when results are likely to be reliable and to monitor the safety of the interventions. PMA offers a unique opportunity to perform similar interim analyses using data from all studies. Data may be combined across studies usually as summary data or even as individual participant data (depending on what is feasible for the collaboration) for an interim analysis under the auspices of an overarching data safety monitoring committee (DSMC) or assessed separately by each study and the results then shared amongst the DSMCs of all the participating studies.

The ability to perform combined interim analyses raises some ethical issues. If an overall net benefit of an intervention on the primary outcome has been demonstrated, issues may arise including whether it would be appropriate to continue randomization within individual studies, or whether investigators proceed with the PMA to clarify treatment effects in subgroups of clinical interest or for less common endpoints. If each study has its own DSMC, then communication amongst committees, for example in deciding whether or not to close an individual study early because of evidence of efficacy from the combined interim data, would be beneficial (Hillman and Louis 2003). It could be argued that knowledge of combined safety data from all participating studies might reduce the chances of spurious early stopping of an individual study. Studies may also consider limiting enrolment in continuing studies to participants in a particular subgroup of interest if such a decision makes clinical and statistical sense (Chalmers et al 2013).

It is important to pre-specify how studies in a PMA can work together in a shared DSMC. A PMA of oxygen saturation targets for preterm infants provides an example of the impact of these questions on data safety monitoring. Participating studies made different decisions on whether to share interim data or not, which then led to early stopping of some studies but not others (Chalmers et al 2013). To ensure ethical obligations are fulfilled, and participants are not continuing to be enrolled to an intervention known to be not effective or harmful, PMA collaborations should carefully plan their model of data safety monitoring in advance.

22.2.10 Other impacts of PMA included studies

PMA can impact on the design and conduct of individual studies, beyond the harmonization of protocols. For example, similar studies may adapt their planned sample sizes to obtain sufficient power for the detection of subgroup effects within the PMA.

For PMA using IPD, studies need to ensure that they incorporate plans for data sharing into study design and conduct. Planning for data sharing should commence early, to ensure appropriate participant consent is sought to share data, and other potential barriers to data sharing are circumvented early. This can make data sharing in an IPD-PMA more efficient than in a conventional retrospective IPD meta-analysis (Hunter et al 2023; Tan et al 2021). It is important that each study’s ethics application and consent form outlines plans for data sharing. If a PMA is initiated after some eligible studies are already underway, amendments to ethics applications may be required to facilitate data sharing.

The decision to collaborate in a PMA may also be useful for convincing funders of the need to continue supporting studies, for example with expensive interventions or rare diseases where individual funders have insufficient funds to support one single adequately powered study. For the study teams involved in the ARTISTIC PMA, prospectively planning a meta-analysis helped study investigators reassure patients and funders that there was value in continuing their studies (Vale et al 2020), and provided justification for an independent data monitoring committee for one of the studies (Parker et al 2020) to amend the primary outcome.

22.3 Conducting a prospective, collaborative meta-analysis

22.3.1 Data collection

A key difference between a standard Cochrane review and one including a PMA would be that the data for the PMA are likely to be collected directly from study teams (or data repositories) rather than being extracted from reports or publications. Precisely which data and how they are collected depends on whether the PMA is based on IPD or aggregate data as well as on the outcomes and subgroups of interest to the specific research question.

22.3.1.1 Collecting study design and conduct information

For PMA based on either IPD or aggregate data, it is common to seek information directly from the study investigators on study design, participants, interventions, and outcomes to better understand each eligible study, and to later populate a table of study characteristics and aid risk of bias assessments. Requesting a full study protocol is ideal for this purpose, as study or trial registry entries may not be sufficiently detailed.

22.3.1.2 Collecting study baseline and outcome data

For PMA based on IPD, baseline, treatment allocation, and outcome data are sought for each eligible participant in each study and used directly in the descriptive and outcome analyses, similar to a standard IPD meta-analysis (see Chapter 26, Section ‘Deciding which data items to collect’). Suggested formats, coding, and detailed data definitions are usually supplied via a data dictionary and the participant data requested need to be transferred and stored securely (see Chapter 26, Sections ‘Data security and confidentiality’ and ‘Re-coding and re-defining data’ and Tierney et al 2021c). Study-level data and IPD can also be sought to help assess data integrity and enhance risk of bias assessments (Hunter et al 2024) (see Section 22.3.2, Chapter 26, Section ‘Managing and checking IPD’ and Tierney et al 2021c).

If it is not possible to obtain IPD from some of the identified studies, then detailed aggregate-level data can be sought for those studies instead. Compared to a standard Cochrane review, this enables better descriptions of study characteristics, analyses of more outcomes, and investigation of effects by participant-level subgroups. Baseline data (e.g. median age, percentage female) and results for each outcome, both overall and for participant subgroups or subsets, would be requested in a consistent way (Vale 2013), ideally via structured data collection forms or spreadsheets supplied by the PMA project team. This also reduces the burden on study teams and aids the automation of data processing in standard statistical software packages by the PMA core project team. Many studies collect and report equity-related variables heterogeneously and sparsely making equity-related assessments difficult in standard Cochrane reviews. PMAs can prospectively address this, by using standardized frameworks for collecting and reporting participant demographic and equity-relevant characteristics across included trials. Established frameworks such as diversity and inclusion measures (e.g. Diversity Minimal Item Set (Stadler et al 2023)) or harmonized indicator sets available through the WHO Health Equity Assessment Toolkit (HEAT 2024) can support alignment with international statistical standards, thereby also decreasing heterogeneity across the wider evidence base. Other initiatives, such as PRO EDI, may also offer guidance (PRO EDI).

The structure of aggregate data will depend on the outcome and choice of effect measure as defined in the PMA protocol. For example, for binary or continuous outcomes the ‘raw’ data (e.g. frequency counts, means and standard deviations) or model-based output using an appropriate effect measure (e.g. odds ratios, risk ratios, mean differences), alongside an appropriate measure of the variability (e.g. 95% confidence interval), could be requested. For other outcome types (e.g. time-to-event outcomes) it may be that model-based output is required to enable meta-analysis.

Additional information relevant to risk of bias assessments including, for example, whether outcome assessors were blinded, or to what degree participants adhered to their randomized interventions, can be collected similarly, and if sought in advance of outcome results, the assessments can be completed ahead of the meta-analysis. Similarly, patient characteristics at baseline may be collected on completion of recruitment for each study to increase efficiency in the process. For any studies that do not provide detailed aggregate data (e.g. a study that is not part of the PMA), then standard aggregate data can be extracted from publications, where appropriate, to supplement this.

22.3.2 Checking study data, risk of bias, and integrity

The collaborative nature of PMA fosters open, regular, and direct communication with study investigators, which is valuable for resolving uncertainties that arise during data checking, and risk of bias and integrity assessments. For example, communication can clarify sequence generation methods, confirm ethics approval details, or explain data anomalies. However, final judgements in these assessments should be made independently of the investigators to reduce the influence of potential conflicts of interests.

For PMA based on IPD, rigorous data checking is essential, which is the same for standard IPD meta-analysis (see Tierney et al 2021c and Chapter 26, Section ‘Checking the completeness and integrity of incoming data’). This process may be supported by adherence to a standardized checklist, such as the Checklist for Individual Participant data Processing of Randomised Trials (CHIPPR, in development (Hunter et al 2022b)). Guidance for how IPD can inform risk of bias assessments is also available (Sotiropoulos et al 2025; Tierney et al 2021c). Note that a key strength of PMA is that the collaboration might be able to work together to harmonize important protocol elements at the study design stage to reduce the risk of bias, for example blinding of outcome assessments.

Similarly, adherence to key integrity items related to governance can be proactively monitored. For instance, the collaboration can require and ensure that all studies are registered prospectively in a WHO-recognized clinical trials registry, and verify that appropriate ethical approvals have been sought. Given that rising cases of fabrication, falsification, plagiarism, and questionable research practices are being uncovered, integrity checks are increasingly recognized as important in systematic reviews, including PMA. There are several tools available to support such integrity assessments, such as TRACT and INSPECT-SR (Mol et al 2023; Wilkinson et al 2025) for aggregate-level data and the IPD Integrity Tool (Hunter et al 2024) for IPD. Ongoing engagement with investigators and tracking the progress of studies over time may also safeguard against nefarious or questionable research practices (Godolphin et al 2022).

22.3.3 Statistical analysis

Most IPD-PMAs use a similar, comprehensive analytical approach and workflow to retrospective IPD meta-analyses (see Chapter 26, Section ‘Analysis of IPD’), whereas the contrast between a PMA of aggregate data and a standard Cochrane review will depend on the type and quantity of data obtained. In some cases, analysis workflows, depth, and detail will more closely mimic those of an IPD meta-analysis (e.g. WHO 2021) and specialist software packages (e.g. R or Stata) may be required to carry out such analyses.

Due to the prospective nature of PMA, with hypotheses generated in advance of knowledge of results, there is the possibility of making stronger causal statements than may be possible in a standard, retrospective meta-analysis. Therefore, it is important that PMA statisticians adopt good practice as would be expected in clinical trial statistics (EMA 1998). For instance, PMA statisticians should ensure that they remain blind to associations between treatment allocation and principal outcomes until the SAP has been finalized (and ideally made available online and date stamped, see Section 22.2.8.2) to preserve the prospective nature of the PMA (Iflaifel et al 2022).

22.3.4 Assessing certainty of evidence

The overall certainty of evidence arising from a PMA can be appraised in a similar way to a standard Cochrane review (e.g. using GRADE, see Chapter 14, Section ‘The GRADE approach’). However, since the planning and statistical analyses are defined a priori, potential bias in review and meta-analysis methods are reduced. In addition, having more complete and consistent outcome and subgroup data due to harmonization can provide higher precision, avoid reporting bias, and limit clinical and methodological heterogeneity. Together, this can lead to higher certainty of evidence ratings. Similar to standard Cochrane reviews, assessments of certainty of evidence are an important tool to inform the reporting of the PMA and, due to their prospective nature and common design features, it might be possible to interpret the level of evidence from PMA of randomized trials as akin to a multicentre trial.

22.4 Reporting and dissemination of a prospective, collaborative meta-analysis

22.4.1 Coordinating publication of a PMA and included studies

An issue that arises for PMAs, which would generally not arise for standard systematic reviews, is how to coordinate publication of included studies and the PMA. Investigators may prefer for their studies to be published before, or at the same time as, the PMA results. Where possible, these decisions should be discussed and agreed in advance and documented in collaboration agreements (see Section 22.2.4) and reviewed as needed.

Aligning publication of the PMA results with (all, or some, of) the included studies can benefit both the included studies and the meta-analysis. For example, through negotiation with journal editors, it may be possible that studies included in the PMA are published simultaneously with the publication of the PMA in the Cochrane Library, thereby increasing exposure and the impact of the joint work, or for PMA published outside of Cochrane they may appear in the same journal as the included trials. A recent example of this is the ARTISTIC PMA in prostate cancer (Kneebone et al 2020; Sargos et al 2020; Vale et al 2020). In addition, it may be possible to work with guideline developers to align a PMA publication with release of a related guideline, as was achieved in COVID-19 (Agarwal et al 2020; WHO 2021; WHO 2025b) (see Section 22.4.4).

Note that coordinating publication of included studies and a PMA may be easier if the meta-analysis is based on aggregate-level data (e.g. using the FAME approach) because the time lag between publication of studies and collection and analysis of IPD meta-analysis often makes this more difficult. Hence, studies need to share data very early if an IPD-PMA is to be publish simultaneously with the individual study results. Either way, a PMA using participant-level data can still align with clinical or public health guideline reporting.

Even if managed well, in some cases a PMA publication may include study outcomes or results that are unpublished. Hence, it is recommended that PMA publications clearly indicate the sources of included data and refer to any prior publications of the individual studies.

22.4.2 Determining authorship of the PMA publication

A PMA is a major collaborative undertaking, whether based on IPD or aggregate data, so PMA publications should recognize the involvement of study teams as well as the core project team. This may involve offering study representatives the opportunity to contribute as co-authors for the PMA (consistent with International Committee of Medical Journal Editors (ICMJE) guidelines for authorship (ICMJE). Sometimes, PMA results are published under a collaborative group name, in which case the contributions of each collaborator in this group should be clearly described, alongside any relevant conflicts of interest.

22.4.3 Reporting guidelines for PMA

PMA-specific reporting standards are, as of May 2026, currently in development, but have not been released, so authors are advised to follow PRISMA-2020 (Page et al 2021b) and appropriate extensions, where applicable (e.g. PRISMA-IPD (Stewart et al 2015)), and to check the Equator network for PRISMA-PMA extension updates. In addition, items specific to the PMA that were documented in the protocol (see Section 22.5, Box 4) should be reported – including searching for unpublished and ongoing studies, time of study inclusion, and outcome and protocol harmonization, and triggers for the PMA. Any deviations from the protocol should be clearly explained. In addition, it is important to report studies that were identified but could not provide information within the required timeframe for analysis, and how decisions were made to proceed nonetheless with analysis. It is also important to report on collaboration management, how involved the included investigators were, and how independent assessments of the studies were.

22.4.4 Engaging with clinical guideline groups and/or policy makers

As with other Cochrane reviews (see Chapter 2, Section ‘Engaging stakeholders to help define the review questions’ and Chapter 1, ‘Who should do a systematic review?’), PMAs may benefit from engagement with relevant stakeholders, including policy makers and guideline developers, where appropriate. In PMA compared to standard Cochrane reviews, the prospective timing may create opportunities for earlier engagement so that the review question, outcomes, and timing of synthesis are well aligned with decision-making needs. To ensure that results from the PMA can impact clinical or public health practice as quickly as possible, contact should be made with relevant guideline development groups in the early stages of a PMA. Timelines can be shared, enabling the guideline group to convene and review the PMA evidence at the earliest appropriate time point.

Often, results from the PMA can be provided, in confidence, to guideline bodies at (or even prior to) submission for publication, helping to reduce time-lag, and enabling the alignment of publication of the PMA results and the guideline(s). Guideline bodies, including policy makers, may even have representatives as part of the PMA project team to provide input on the protocol and statistical analysis plan and ensure the PMA addresses research questions relevant to the guidelines and their recommendations. For example, in the Individual Participant Data on Cord Management at Preterm Birth (iCOMP) study (Seidler et al 2023a), which included a nested prospective meta-analysis, the Chair of the International Liaison Committee (ILCOR) on the Resuscitation Neonatal Life Support Task Force was a member of the PMA advisory group. This enabled the study design to align with ILCOR’s objectives. Results were shared with the ILCOR taskforce ahead of publication, enabling the evidence to be considered at the earliest appropriate time point and helping to align the timing of treatment recommendations with publication of the study findings. In the WHO REACT PMAs (Vale et al 2024; WHO 2021; WHO 2025a; WHO 2020), the WHO Clinical Management lead was part of the core PMA team and provided a link between the PMA and the WHO Guideline Panel, ensuring that outcomes and subgroups relevant to the panel were collected and analysed, and that results were provided confidentially to the panel, pre-publication, as soon as they were finalized.

22.5 Other types of prospective and/or collaborative meta-analysis

22.5.1 Framework for Adaptive Meta-analysis

The Framework for Adaptive Meta-analysis (FAME) is an approach that aims to produce timely, reliable, and thorough PMA based on aggregate data. It is both fully prospective and collaborative (Tierney et al 2021b). Many of the methods described for aggregate data PMA above derive from these principles, which, in brief, are:

1. Start early, whilst studies are ongoing or yet to report.
2. Liaise with study investigators to get information on studies including trial design, accrual, data maturity, and analysis and dissemination.
3. Predict the earliest possible timing of reliable meta-analysis based on having sufficient power and a substantial proportion of eligible information (see Section 22.2.8.3).
4. Develop a protocol and collect harmonized and detailed data from investigators.
5. Interpret meta-analysis results taking account of available and unavailable data, and assess the likely value of updating.

The full FAME approach has been successfully implemented to provide more timely, reliable, and thorough evaluation of prostate cancer treatments (Burdett et al 2019; Rydzewska et al 2017; Vale et al 2016; Vale et al 2020). This includes providing timely insight into adverse effects of abiraterone (Rydzewska et al 2017), subgroup effects associated with prostate radiotherapy (Burdett et al 2019; Rydzewska et al 2017) and effects of radiotherapy scheduling on an early outcome (Vale et al 2020).

It may not always be possible to follow all these principles for every health question. However, using some of them (‘FAME-lite’) still gives the ability to increase timeliness, reduce bias, and provide more nuanced results than standard approaches to evidence synthesis. For example, in the DADSPORT FAME-lite review of the duration of hormone therapy for non-metastatic prostate cancer (Burdett et al 2025), ongoing and reported studies were included (akin to nested PMA, see Section 22.5.2). Investigators from all studies formed part of the PMA collaboration and provided detailed and harmonized data that enabled analyses of more outcomes and participant subgroups and a network meta-analysis. These analyses would not have been possible in an equivalent review based on published aggregate data. This meant that the project team could show that short- or long-course hormone therapy prolonged cancer-specific and metastases-free survival, but not overall survival, and that the effects were consistent across studies, despite some differences in design and participant characteristics.

22.5.2 Nested prospective meta-analysis

For some clinical questions, there may already be relevant evidence available (e.g. multiple published studies and/or meta-analyses) whilst many relevant studies continue to be planned or are ongoing. Similarly, living reviews or updates of standard reviews may need to integrate previously published retrospective evidence. In such cases, combining prospective and/or collaborative elements with retrospective approaches can still reduce bias, improve data and analysis quality, and strengthen the interpretation and certainty of findings.

A nested PMA is a partially prospective approach in which some eligible studies have already been published, with results known, while others are ongoing (Seidler et al 2019). A nested PMA has a prospective and a retrospective component, which can be analysed combined and separately. Most of the methodological principles described in Sections 22.2 to 22.4 still apply. A FAME-lite approach (see Section 22.5.1) that relaxes the prospective FAME principle (number 1 above) could be seen as a specific type of nested PMA.

For studies still in progress (prospective evidence), harmonization of outcomes and other key protocol elements can enhance the ability to perform robust meta-analyses. This can be guided by how outcomes were defined in already completed studies. Retrospective evidence can also inform power calculations and safety monitoring. Where certain questions have already been answered by the existing evidence, for example the overall effect of an intervention or the effects in a particular population, the prospective component of a nested PMA may focus on subgroups for which evidence remains limited.

In a nested PMA, the protocol and SAP should pre-specify how the retrospective evidence will be handled in the overall analysis and analysis of participant subgroups. One simple approach would be for the primary analysis to include both the prospective and retrospective evidence, with a sensitivity analysis restricted to the prospective evidence. For analysis of participant subgroups, the retrospective evidence could be seen as hypothesis generating and the prospective evidence used to test this further.

For example, the iCOMP collaboration (Seidler et al 2023a; Seidler et al 2023b) initially aimed to conduct a fully prospective meta-analysis of randomized trials on umbilical cord management at preterm birth. However, when searching for eligible studies, the team identified many completed studies, and the volume of available evidence was considered too large for a wholly prospective approach. Moreover, initial funding delays meant that additional studies were completed before the PMA protocol could be finalized. As a result, a nested PMA was planned (Seidler et al 2020b). In the first round of analyses, iCOMP conducted a sensitivity analysis comparing results from prospective and retrospective evidence (Seidler et al 2023a). Investigators were able to include IPD for several completed studies that were unpublished, thus reducing publication bias (Seidler et al 2023a). The collaboration also identified key unanswered questions in specific subgroups, such as infants requiring resuscitation and those with congenital abnormalities, focussing ongoing work on prospectively planning, harmonizing, and combining further studies to address these subgroup-specific questions.

22.5.3 Anytime Live and Leading Interim (ALL-IN) meta-analysis

The ALL-IN meta-analysis approach links prospective and living systematic review methods (ter Schure and Grünwald 2025). It is designed whilst studies are ongoing and involves continuous sharing of IPD with a central team, enabling continuous monitoring that could lead to studies stopping early based on early meta-analysis results. A benefit of this approach would be that the meta-analysis arrives at a definitive answer as soon as it has enough information, without the need for any additional participants being recruited into the contributing studies. Conceptually, this would then lead to the studies halting recruitment, stopping unnecessary randomization. Yet, to reach that point the study would need to recruit more participants, since the ability to continuously monitor results reduces statistical power. In addition, in practice this may be challenging to implement, and it could prove difficult to get sufficient buy-in from study investigators. As such, this approach has only been fully implemented once so far (see below).

An ALL-IN approach uses an alternative statistical paradigm, making use of e-values (Grünwald 2024), which permits continuous monitoring and assessment without inflating Type I error (see Chapter 10, Section ‘Statistical issues arising from repeating meta-analyses’). This is a key distinction of the ALL-IN approach, enabling the meta-analysis to be updated at any time. However, in some cases there might be a cost in terms of an inflated Type II error (that is, reduced statistical power and thus larger required sample sizes, see Chapter 10, Sections ‘Statistical issues arising from repeating meta-analyses’ and ‘Sequential statistical methods for meta-analysis’). Furthermore, quite complex infrastructure and resources may be required to enable continuous data sharing whilst studies are ongoing, as well as full collaboration and data sharing between study teams and the meta-analysis project team to enable regular analysis updates.

For example, during the COVID-19 pandemic, an ALL-IN meta-analysis investigated the effect of vaccination of healthcare workers compared to placebo or no vaccination on COVID-19 infections and hospitalizations (ter Schure et al 2022). Six trials were included, with each trial regularly uploading data to a secure server for combined analysis.

22.5.4 Collaborative network trials

Collaborative network trials are another collaborative, prospective approach which combine and adopt many of the principles of PMA and master protocols (Bogin 2020). In contrast to multicentre trials, Collaborative Network Trials include several comparable but independent trials, designed as fit-for-purpose by each of the collaborating sites. This means they may address local evidence needs and questions whilst being harmonized enough to contribute to the overall research aim, just like individual trials contributing to a standard PMA.

They have been employed in the field of implementation science to examine the effects of different strategies to implement the same intervention (e.g. health service, procedure, or therapy) on the same health service outcomes (Barnes et al 2024). Like a PMA, they involve the conduct of independent trials that are formally networked, undertaken at different sites or by different health services, but employing harmonized methods (Barnes et al 2024). In contrast to most standard PMAs, collaborative network trials can be supported by shared trial expertise and infrastructure. Data are shared between health services and sites undertaking trials to support pooled analyses, often using advanced IPD meta-analysis methods (e.g. IPD network meta-analysis (NMA, see Chapter 11) or component IPD-NMA (Tsokani et al 2023)) and can be used by health services to guide local improvements to the implementation of the targeted health intervention. In contrast to a standard PMA, individual trials in a collaborative network trial may not complete their own independent analysis or publish their own results separately. Like the ALL-IN approach, all analyses are usually centralized.

Collaborative network trials adopt elements of master protocols to undertake research more efficiently (Bogin 2020). As such, they can incorporate multiple subgroups or sub-studies examined under a unifying trial infrastructure. They can accommodate adaptive designs, allowing ineffective interventions (or implementation strategies) to be dropped, or effective interventions to be further improved.

In Australia, a collaborative network of 10 randomized trials was undertaken by health promotion services in each of 10 local health districts to improve the implementation of a school nutrition programme (Barnes et al 2024). Each health promotion service tested the effects of a different strategy designed by health services in a way that addressed local barriers to programme implementation. A governance structure was established with representatives of all participating services and oversaw centralized data collection, allocation, harmonization of core outcomes, and analysis.

22.6 New developments and other considerations in prospective meta-analyses

22.6.1 Increased study registration

One key challenge for PMA is the identification of ongoing studies as they arise. For randomized trials, identification of all available evidence is facilitated by the high rate of registration of newly planned trials in WHO-recognized clinical trial registries. Interest holders across the trial life cycle, including regulatory bodies, funding agencies, journals, and institutions, have introduced and are enforcing requirements for prospective trial registration, making most trials findable before recruitment begins. Another key advance has been the progressive improvement of trial registry search functions. The availability of specific guidance on how to search clinical trial registries has further enhanced the ability to identify relevant studies (Hunter et al 2022a) (see Section 22.2.2).

22.6.2 Artificial intelligence and automation

In the future, artificial intelligence may enable an automatic searching and screening process for planned and ongoing studies. Such technologies may even automatically connect researchers who are planning similar relevant studies and suggest how they may collaborate and harmonize their planned studies within a PMA model. Increased automation may also make data processing and statistical analyses more streamlined and efficient, which may enable continuous monitoring approaches. Yet, just like for standard Cochrane reviews, it is important to consider the impact of potential bias and lack of transparency when introducing artificial intelligence solutions.

22.6.3 PMA for health emergencies

PMA is particularly suited to a public health emergency context, such as a pandemic. In these circumstances, there is a risk of substantial research waste if multiple small, and thus underpowered, individual randomized trials are initiated at speed and not designed or reported in ways that facilitate timely synthesis (Seidler et al 2021b). If the public health emergency relates to a new disease or condition, there may also be no or limited existing randomized evidence available for decision making, so a fully prospective approach to meta-analysis is possible and desirable. By planning the meta-analysis in advance and harmonizing key features across trials where feasible, PMA can help ensure that results from all trials contribute to a coordinated body of evidence, including where individual trials might otherwise remain unpublished or have limited influence on practice. The global community will require rapid, but reliable, evidence so collecting detailed aggregate rather than IPD might provide the most timely evidence to inform clinical practice worldwide.

For example, during the COVID-19 pandemic, the WHO REACT Working Group initiated four PMAs of randomized trials in patients hospitalized with COVID-19, to evaluate the effects of corticosteroids, interleukin-6 antagonists, sodium glucose co-transporter-2 inhibitors, and anticoagulants (Godolphin et al 2026; Vale et al 2024; WHO 2020; WHO 2021; WHO 2025a). For each PMA, detailed aggregate data were obtained to evaluate treatment effects on multiple outcomes, both overall and across pre-defined patient subgroups, as well as extensive information to inform risk of bias assessments, providing greater context on the reliability of the evidence. Results from each PMA were shared directly with WHO Guideline Development Groups before publication, to feed into living NMA and improve the efficiency of preparing recommendations. Although there were increased calls for sharing of IPD during the COVID-19 pandemic (Petkova et al 2020), aggregate data approaches enabled faster accumulation of evidence without the data sharing infrastructure and bureaucracy that can hinder collation of IPD.

22.6.4 PMA of studies other than randomized trials

Prospective meta-analyses can be extended to observational studies of the effects of interventions, studies of other human health questions (e.g. diagnostic test accuracy or prognosis), or other study types (e.g. animal studies). Many principles remain the same, but there are a few key differences:

• Identification of ongoing studies is more difficult for non-interventional studies, where there is no mandatory registration and thus there are far lower registration rates. Currently, such PMAs commonly rely on investigators deciding proactively to plan similar studies and connecting to each other via professional networks. Additionally, PMA plans may need to be publicized through presentations at relevant conferences, the publication of PMA protocols, and through a collaboration website or posts on relevant social media.
• Results need to be interpreted according to the level of evidence included in the PMA. If a PMA of the effects of an intervention includes observational studies, the findings should generally be interpreted as associations rather than as causal effects. Although modern causal inference methods applied to observational data can sometimes provide causal interpretations, particular caution is recommended if these are included within a wider evidence synthesis/PMA.
• In PMAs of observational studies, the prospectivity criterion, i.e. “no results related to the PMA research question being known”, may mean that large data collections (e.g. regulatory data) that have previously been used to answer different research questions may still be included as prospective evidence in the PMA because they have never been analysed to answer this specific research question (e.g. if certain outcomes or predictors in a cohort have not previously been analysed).
• In many situations, especially for assessing drug safety, where event rates may be low, it may be necessary to analyse data from multiple sources (e.g. multiple large administrative claims databases). In addition to increasing statistical power for infrequent events, using multiple databases allows the evaluation of consistency, since we generally put more faith in results that are consistent across multiple data sources than in results from single sources. However, when interpreting the results it is important to remember that the level of evidence from observational databases is lower than randomized evidence.

While meta-analyses of observational studies of the effects of interventions are frequently subject to major heterogeneity concerns, due to different populations, adjustments for different covariates, and different exposure measurements and analyses, PMA offers huge potential for the harmonization of data and analyses to mitigate these.

22.6.5 Embedding a study within a PMA

A study within a review (SWAR) is a study that is embedded within a systematic review, which aims to provide evidence about systematic review methodology (Devane et al 2022). SWARs are distinct from their parent review and aim to address a methodological concern regarding a specific part of the systematic review process. Examples of SWARs cover methods for screening (Devane and Clarke 2011), obtaining unpublished data (Godolphin et al 2019), and querying information with study investigators (Goossen et al 2021).

Given their prospective nature, PMA approaches offer a unique possibility to build in SWARs to test and improve PMA methods. Further, the time-lag from initiating a PMA to its completion, while awaiting the accumulation of data, gives ample opportunity to design and implement a SWAR. If the project team is considering embedding a SWAR in a PMA, then registering the SWAR (https://www.qub.ac.uk/sites/TheNorthernIrelandNetworkforTrialsMethodologyResearch/SWATSWARInformation/Repositories/SWARStore/) and linking with the SWAR Network (https://evidencesynthesisireland.ie/swar-network/) are useful steps to ensure that the proposed SWAR is both high priority and adds to the evidence base.

22.7 Chapter information

Authors: Anna Lene Seidler, Peter J Godolphin, Kylie E Hunter, Claire Vale, Luke Wolfenden, David J Fisher, Lisa M Askie, Jesse A Berlin, Jayne F Tierney on behalf of the Cochrane Prospective Meta-Analysis Methods Group

Funding: none

Declarations of interest: none known

22.8 References

Agarwal A, Hunt BJ, Stegemann M, Rochwerg B, Lamontagne F, Siemieniuk RA, et al. A living WHO guideline on drugs for covid-19. BMJ 2020;370:m3379. doi:10.1136/bmj.m3379

Askie LM, Darlow BA, Davis PG, Finer N, Stenson B, Vento M, et al. Effects of targeting lower versus higher arterial oxygen saturations on death or disability in preterm infants. Cochrane Database of Systematic Reviews 2017;4. doi:10.1002/14651858.CD011190.pub2

Askie LM, Darlow BA, Finer N, Schmidt B, Stenson B, Tarnow-Mordi W, et al. Association between oxygen saturation targeting and death or disability in extremely preterm infants in the neonatal oxygenation prospective meta-analysis collaboration. JAMA 2018;319(21):2190–201. doi:10.1001/jama.2018.5725

Askie LM, Espinoza D, Martin A, Daniels LA, Mihrshahi S, Taylor R, et al. Interventions commenced by early infancy to prevent childhood obesity—The EPOCH Collaboration: an individual participant data prospective meta-analysis of four randomized controlled trials. Pediatric Obesity 2020;15(6):e12618. doi:10.1111/ijpo.12618

Banno M, Tsujimoto Y, Kataoka Y. Using the Cochrane Central Register of Controlled Trials to identify clinical trial registration is insufficient: a cross-sectional study. BMC Medical Research Methodology 2020;20(1):200. doi:10.1186/s12874-020-01083-y

Barnes C, Jones J, Wolfenden L, Robertson K, Seidler AL, Norman J, et al. A collaborative network trial to evaluate the effectiveness of implementation strategies to maximize adoption of a school-based healthy lunchbox program: a study protocol. Frontiers in Public Health 2024;12:1367017. doi:10.3389/fpubh.2024.1367017

Berlin JA, Santanna J, Schmid CH, Szczech LA, Feldman HI. Individual patient- versus group-level data meta-regressions for the investigation of treatment effect modifiers: ecological bias rears its ugly head. Statistics in Medicine 2002;21(3):371–87. doi:10.1002/sim.1023

Bogin V. Master protocols: new directions in drug discovery. Contemporary Clinical Trials Communications 2020;18:100568. doi:10.1016/j.conctc.2020.100568

Burdett S, Boevé LM, Ingleby FC, Fisher DJ, Rydzewska LH, Vale CL, et al. Prostate radiotherapy for metastatic hormone-sensitive prostate cancer: a STOPCAP Systematic review and meta-analysis. European Urology 2019;76(1):115–24. doi:10.1016/j.eururo.2019.02.003

Burdett S, Fisher DJ, Tierney JF, Cook AD, Spratt DE, Dignam JJ, et al. Duration of Androgen Suppression with Postoperative Radiotherapy (DADSPORT) for nonmetastatic prostate cancer: a collaborative systematic review and meta-analysis of aggregate data. European Urology 2025;88(3):277–90. doi:10.1016/j.eururo.2025.05.013

Chalmers I, Altman DG, McHaffie H, Owens N, Cooke RW. Data sharing among data monitoring committees and responsibilities to patients and science. Trials 2013;14(1):102. doi:10.1186/1745-6215-14-102

COSMIN. I’m looking for available outcome measurement instruments [Internet]. [cited 2025 Oct 6]. Available from: https://www.cosmin.nl/finding-right-tool/available-outcome-measurement-instrument/

Debray TPA, Moons KGM, van Valkenhoef G, Efthimiou O, Hummel N, Groenwold RHH, et al. Get real in individual participant data (IPD) meta-analysis: a review of the methodology. Research Synthesis Methods 2015;6(4):293–309. doi:10.1002/jrsm.1160

Devane D, Burke NN, Treweek S, Clarke M, Thomas J, Booth A, et al. Study within a review (SWAR). Journal of Evidence-Based Medicine 2022;15(4):328–32. doi:10.1111/jebm.12505

Devane D, Clarke M. SWAR 01: Citation screening in systematic reviews: two approaches, two authors and time taken [Internet]. 2011 [cited 2025 Oct 7]. Available from: https://www.qub.ac.uk/sites/TheNorthernIrelandNetworkforTrialsMethodologyResearch/FileStore/SWARFileStore/Filetoupload1519796en.pdf

Dickersin K, Chan S, Chalmers TC, Sacks HS, Smith H. Publication bias and clinical trials. Controlled Clinical Trials 1987;8(4):343–53. doi:10.1016/0197-2456(87)90155-3

European Medicines Agency (EMA). ICH E9 statistical principles for clinical trials - Scientific guideline [Internet]. 1998 [cited 2025 Oct 15]. Available from: https://www.ema.europa.eu/en/ich-e9-statistical-principles-clinical-trials-scientific-guideline

Fisher DJ, Carpenter JR, Morris TP, Freeman SC, Tierney JF. Meta-analytical methods to identify who benefits most from treatments: daft, deluded, or deft approach? BMJ 2017; 356:j573. doi:10.1136/bmj.j573

Gamble C, Krishan A, Stocken D, Lewis S, Juszczak E, Doré C, et al. Guidelines for the content of statistical analysis plans in clinical trials. JAMA 2017;318(23):2337–43. doi:10.1001/jama.2017.18556

Glanville JM, Duffy S, McCool R, Varley D. Searching ClinicalTrials.gov and the International Clinical Trials Registry Platform to inform systematic reviews: what are the optimal search approaches? Journal of the Medical Library Association: JMLA. 2014;102(3):177–83. doi:10.3163/1536-5050.102.3.007

Godolphin PJ, Bath PM, Montgomery AA. Short email with attachment versus long email without attachment when contacting authors to request unpublished data for a systematic review: a nested randomised trial. BMJ Open 2019;9(1):e025273. doi:10.1136/bmjopen-2018-025273

Godolphin PJ, Fisher DJ, Bradbury CA, Murthy S, Vale CL. Intermediate- versus prophylactic-dose heparin for hospitalised patients with COVID-19: an updated meta-analysis of randomised trials. Intensive Care Medicine 2026;52(4):800–2. doi:10.1007/s00134-025-08128-0

Godolphin PJ, Rogozińska E, Fisher DJ, Vale CL, Tierney JF. Meta-analyses based on summary data can provide timely, thorough and reliable evidence: don’t dismiss them yet. Nature Medicine 2022;28(3):429–30. doi:10.1038/s41591-021-01675-1

Godolphin PJ, White IR, Tierney JF, Fisher DJ. Estimating interactions and subgroup-specific treatment effects in meta-analysis without aggregation bias: a within-trial framework. Research Synthesis Methods 2023;14(1):68–78. doi:10.1002/jrsm.1590

Goossen K, Rombey T, Kugler CM, De Santis KK, Pieper D. Author queries via email text elicited high response and took less reviewer time than data forms – a randomised study within a review. Journal of Clinical Epidemiology 2021;135:1–9. doi:10.1016/j.jclinepi.2021.02.006

Grünwald PD. Beyond Neyman–Pearson: E-values enable hypothesis testing with a data-driven alpha. Proceedings of the National Academy of Sciences of the United States of America 2024;121(39):e2302098121. doi:10.1073/pnas.2302098121

Health Equity Assessment Toolkit (HEAT): Software for exploring and comparing health inequalities in countries. Built-in database edition. Version 7.0. [Internet]. Geneva: World Health Organization; 2024 [cited 2026 Apr 27]. Available from: https://whoequity-heat-1.share.connect.posit.cloud

Hillman DW, Louis TA. DSMB case study: decision making when a similar clinical trial is stopped early. Controlled Clinical Trials 2003;24(1):85–91. doi:10.1016/s0197-2456(02)00274-x

Hunter KE, Aberoumand M, Libesman S, Sotiropoulos JX, Williams JG, Aagerup J, et al. The Individual Participant Data Integrity Tool for assessing the integrity of randomised trials. Research Synthesis Methods. 2024;15(6):917–39. doi:10.1002/jrsm.1738

Hunter KE, Nguyen D, Libesman S, Williams JG, Aberoumand M, Aagerup J, et al. Parent-focused behavioural interventions for the prevention of early childhood obesity (TOPCHILD): a systematic review and individual participant data meta-analysis. Lancet 2025;406(10509):1235–54. doi:10.1016/S0140-6736(25)01144-4

Hunter KE, Tan AC, Webster AC, Hamilton DG, Barnett A, Jones L, et al. Responsibilities for receiving and using individual participant data. Cochrane Evidence Synthesis and Methods 2023;1(9):e12028. doi:10.1002/cesm.12028

Hunter KE, Webster AC, Clarke M, Page MJ, Libesman S, Godolphin PJ, et al. Development of a checklist of standard items for processing individual participant data from randomised trials for meta-analyses: protocol for a modified e-Delphi study. PLOS One 2022b;17(10):e0275893. doi:10.1371/journal.pone.0275893

Hunter KE, Webster AC, Page MJ, Willson M, McDonald S, Berber S, et al. Searching clinical trials registers: guide for systematic reviewers. BMJ 2022a;377:e068791. doi:10.1136/bmj-2021-068791

Iflaifel M, Partlett C, Bell J, Cook A, Gamble C, Julious S, et al. Blinding of study statisticians in clinical trials: a qualitative study in UK clinical trials units. Trials 2022;23(1):535. doi:10.1186/s13063-022-06481-9

International Committee of Medical Journal Editors. Defining the Role of Authors and Contributors [Internet]. [cited 2025 Oct 7]. Available from: https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html

Kneebone A, Fraser-Browne C, Duchesne GM, Fisher R, Frydenberg M, Herschtal A, et al. Adjuvant radiotherapy versus early salvage radiotherapy following radical prostatectomy (TROG 08.03/ANZUP RAVES): a randomised, controlled, phase 3, non-inferiority trial. Lancet Oncology 2020;21(10):1331–40. doi:10.1016/S1470-2045(20)30456-3

Moher D, Shamseer L, Clarke M, Ghersi D, Liberati A, Petticrew M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Systematic Reviews 2015;4(1):1. doi:10.1186/2046-4053-4-1

Mol BW, Lai S, Rahim A, Bordewijk EM, Wang R, van Eekelen R, et al. Checklist to assess Trustworthiness in RAndomised Controlled Trials (TRACT checklist): concept proposal and pilot. Research Integrity and Peer Review 2023;8(1):6. doi:10.1186/s41073-023-00130-8

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 2021b;372:n71. doi:10.1136/bmj.n71

Page MJ, McKenzie JE, Kirkham J, Dwan K, Kramer S, Green S, et al. Bias due to selective inclusion and reporting of outcomes and analyses in systematic reviews of randomised trials of healthcare interventions. Cochrane Database of Systematic Reviews 2014;10. doi:10.1002/14651858.MR000035.pub2

Page MJ, Sterne JAC, Higgins JPT, Egger M. Investigating and dealing with publication bias and other reporting biases in meta-analyses of health research: a review. Research Synthesis Methods 2021a;12(2):248–59. doi:10.1002/jrsm.1468

Parker CC, Clarke NW, Cook AD, Kynaston HG, Petersen PM, Catton C, et al. Timing of radiotherapy after radical prostatectomy (RADICALS-RT): a randomised, controlled phase 3 trial. Lancet 2020;396(10260):1413–21. doi:10.1016/S0140-6736(20)31553-1

Petkova E, Antman EM, Troxel AB. Pooling data from individual clinical trials in the COVID-19 era. JAMA 2020;324(6):543–5. doi:10.1001/jama.2020.13042

PRO EDI: improving how equity, diversity and inclusion are handled in trials and evidence synthesis. Trial Forge [Internet]. [cited 2026 Apr 27]. Available from: https://www.trialforge.org/trial-diversity/pro-edi/

Riley RD, Ensor J. Power calculations for planning an IPD meta-analysis. In: Individual Participant Data Meta‐Analysis: A Handbook for Healthcare Research. 2021:291–309. doi:10.1002/9781119333784.ch12

Rydzewska LHM, Burdett S, Vale CL, Clarke NW, Fizazi K, Kheoh T, et al. Adding abiraterone to androgen deprivation therapy in men with metastatic hormone-sensitive prostate cancer: a systematic review and meta-analysis. European Journal of Cancer 2017;84:88–101. doi:10.1016/j.ejca.2017.07.003

Sargos P, Chabaud S, Latorzeff I, Magné N, Benyoucef A, Supiot S, et al. Adjuvant radiotherapy versus early salvage radiotherapy plus short-term androgen deprivation therapy in men with localised prostate cancer after radical prostatectomy (GETUG-AFU 17): a randomised, phase 3 trial. Lancet Oncology 2020;21(10):1341–52. doi:10.1016/S1470-2045(20)30454-X

Seidler AL, Aberoumand M, Hunter KE, Barba A, Libesman S, Williams JG, et al. Deferred cord clamping, cord milking, and immediate cord clamping at preterm birth: a systematic review and individual participant data meta-analysis. Lancet 2023a;402(10418):2209–22. doi:10.1016/S0140-6736(23)02468-6

Seidler AL, Aberoumand M, Williams JG, Tan A, Hunter KE, Webster A. The landscape of COVID-19 trials in Australia [Preprint]. Medical Journal of Australia [Internet]. 2021b [cited 2026 Apr 28]. Available from: https://www.mja.com.au/journal/2021/landscape-covid-19-trials-australia

Seidler AL, Duley L, Katheria AC, De Paco Matallana C, Dempsey E, Rabe H, et al; iCOMP Collaboration. Systematic review and network meta-analysis with individual participant data on cord management at preterm birth (iCOMP): study protocol.BMJ Open 2020b;10(3):e034595. doi:10.1136/bmjopen-2019-034595

Seidler AL, Hunter KE, Cheyne S, Berlin JA, Ghersi D, Askie LM. Prospective meta‐analyses and Cochrane’s role in embracing next‐generation methodologies. Cochrane Database of Systematic Reviews 2020a;10. doi:10.1002/14651858.ED000145

Seidler AL, Hunter KE, Cheyne S, Ghersi D, Berlin JA, Askie L. A guide to prospective meta-analysis. BMJ 2019;367: l5342. doi:10.1136/bmj.l5342

Seidler AL, Hunter KE, Espinoza D, Mihrshahi S, Askie LM, Askie LM, et al. Quantifying the advantages of conducting a prospective meta-analysis (PMA): a case study of early childhood obesity prevention. Trials 2021a;22(1):78. doi:10.1186/s13063-020-04984-x

Seidler AL, Libesman S, Hunter KE, Barba A, Aberoumand M, Williams JG, et al. Short, medium, and long deferral of umbilical cord clamping compared with umbilical cord milking and immediate clamping at preterm birth: a systematic review and network meta-analysis with individual participant data. Lancet 2023b;402(10418):2223–34. doi:10.1016/S0140-6736(23)02469-8

Shamseer L, Moher D, Clarke M, Ghersi D, Liberati A, Petticrew M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ 2015;349:g7647. doi:10.1136/bmj.g7647

Showell MG, Cole S, Clarke MJ, DeVito NJ, Farquhar C, Jordan V. Time to publication for results of clinical trials. Cochrane Database of Systematic Reviews 2024;11. doi:10.1002/14651858.MR000011.pub3

Simes RJ. Publication bias: the case for an international registry of clinical trials. Journal of Clinical Oncology 1986;4(10):1529–41. doi:10.1200/JCO.1986.4.10.1529

Siontis KC, Hernandez-Boussard T, Ioannidis JPA. Overlapping meta-analyses on the same topic: survey of published studies. BMJ 2013;347: f4501. doi:10.1136/bmj.f4501

Sotiropoulos JX, Hunter KE, Aagerup J, Williams JG, Libesman S, Aberoumand M, et al. Individual participant data–informed risk of bias assessments for randomized controlled trials in systematic reviews and meta-analyses. Journal of Clinical Epidemiology 2025;185: 111875. doi:10.1016/j.jclinepi.2025.111875

Stadler G, Chesaniuk M, Haering S, Roseman J, Straßburger VM, Martina S, et al. Diversified innovations in the health sciences: Proposal for a Diversity Minimal Item Set (DiMIS). Sustainable Chemistry and Pharmacy 2023;33:101072. doi:10.1016/j.scp.2023.101072

Stewart L, Tierney J, Burdett S. Do systematic reviews based on individual patient data offer a means of circumventing biases associated with trial publications? In: Publication Bias in Meta-Analysis. John Wiley & Sons, Ltd; 2005: 261–86. doi:10.1002/0470870168.ch14

Stewart LA, Clarke M, Rovers M, Riley RD, Simmonds M, Stewart G, et al. Preferred Reporting Items for a Systematic Review and Meta-analysis of Individual Participant Data: The PRISMA-IPD Statement. JAMA 2015;313(16):1657–65. doi:10.1001/jama.2015.3656

Stewart LA, Riley RD, Tierney JF. Planning and initiating an IPD meta-analysis project. In: Individual Participant Data Meta-Analysis: A Handbook for Healthcare Research. John Wiley & Sons, Ltd; 2021:21–43. doi:10.1002/9781119333784.ch3

Tan AC, Askie LM, Hunter KE, Barba A, Simes RJ, Seidler AL. Data sharing—trialists’ plans at registration, attitudes, barriers and facilitators: a cohort study and cross-sectional survey. Research Synthesis Methods 2021;12(5):641–57. doi:10.1002/jrsm.1500

ter Schure JA, Grünwald P. ALL-IN meta-analysis: breathing life into living systematic reviews and prospective meta-analyses [version 2; peer review: 1 approved, 2 approved with reservations]. F1000Research; 2025. doi:10.12688/f1000research.74223.2

ter Schure JA, Ly A, Belin L, Benn CS, Bonten MJM, Cirillo JD, et al. Bacillus Calmette-Guérin vaccine to reduce COVID-19 infections and hospitalisations in healthcare workers – a living systematic review and prospective ALL-IN meta-analysis of individual participant data from randomised controlled trials [Preprint]. medRxiv 2022. doi:10.1101/2022.12.15.22283474

Tierney JF, Fisher DJ, Vale CL, Burdett S, Rydzewska LH, Rogozińska E, et al. A framework for prospective, adaptive meta-analysis (FAME) of aggregate data from randomised trials. PLOS Medicine 2021b;18(5):e1003629. doi:10.1371/journal.pmed.1003629

Tierney JF, Riley RD, Rydzewska LHM, Stewart LA. Running an IPD meta-analysis project. In: Individual Participant Data Meta‐Analysis: A Handbook for Healthcare Research. 2021c:45–80. doi:10.1002/9781119333784.ch4

Tierney JF, Riley RD, Smith CT, Clarke M, Stewart LA. Rationale for embarking on an IPD meta-analysis project. In: Individual Participant Data Meta-Analysis: A Handbook for Healthcare Research. John Wiley & Sons, Ltd; 2021a:7–19. doi:10.1002/9781119333784.ch2

Tierney JF, Vale C, Riley R, Smith CT, Stewart L, Clarke M, et al. Individual participant data (IPD) meta-analyses of randomised controlled trials: guidance on their use. PLOS Medicine 2015;12(7):e1001855. doi:10.1371/journal.pmed.1001855

Tsokani S, Seitidis G, Mavridis D. Component network meta-analysis in a nutshell. BMJ Evidence-Based Medicine 2023;28(3):183–6. doi:10.1136/bmjebm-2021-111906

Tudur Smith C, Marcucci M, Nolan SJ, Iorio A, Sudell M, Riley R, et al. Individual participant data meta‐analyses compared with meta‐analyses based on aggregate data. Cochrane Database of Systematic Reviews 2016;9. doi:10.1002/14651858.MR000007.pub3

Vale C, Godolphin PJ, Fisher D, Horby PW, Kosiborod MN, Hochman JS, et al. Sodium-glucose co-transporter-2 inhibitors for hospitalised patients with COVID-19: a prospective meta-analysis of randomised trials. Lancet Diabetes & Endocrinology 2024;12(10):735–47. doi:10.1016/S2213-8587(24)00219-5

Vale CL, Burdett S, Rydzewska LHM, Albiges L, Clarke NW, Fisher D, et al. Addition of docetaxel or bisphosphonates to standard of care in men with localised or metastatic, hormone-sensitive prostate cancer: a systematic review and meta-analyses of aggregate data. Lancet Oncology 2016;17(2):243–56. doi:10.1016/S1470-2045(15)00489-1

Vale CL, Fisher D, Kneebone A, Parker C, Pearse M, Richaud P, et al. Adjuvant or early salvage radiotherapy for the treatment of localised and locally advanced prostate cancer: a prospectively planned systematic review and meta-analysis of aggregate data. Lancet 2020;396(10260):1422–31. doi:10.1016/S0140-6736(20)31952-8

Vale CL, Tierney JF, Burdett S. Can trial quality be reliably assessed from published reports of cancer trials: evaluation of risk of bias assessments in systematic reviews. BMJ 2013;346:f1798. doi:10.1136/bmj.f1798

WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group, Shankar-Hari M, Vale CL, Godolphin PJ, Fisher D, Higgins JPT, et al. Association between administration of il-6 antagonists and mortality among patients hospitalized for COVID-19: a meta-analysis. JAMA 2021;326(6):499–518. doi:10.1001/jama.2021.11330

WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group, Vale CL, Godolphin PJ, Fisher DJ, Higgins JPT, McAleenan A, et al. Anticoagulation among patients hospitalized for covid-19: a systematic review and prospective meta-analysis. Annals of Internal Medicine 2025a;178(1):59–69. doi:10.7326/ANNALS-24-00800

WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group, Sterne JAC, Murthy S, Diaz JV, Slutsky AS, Villar J, et al. Association between administration of systemic corticosteroids and mortality among critically ill patients with COVID-19: a meta-analysis. JAMA 2020;324(13):1330–41. doi:10.1001/jama.2020.17023

WHO. Clinical management of COVID-19: living guideline, June 2025. World Health Organization; 2025b. doi:10.2471/B09467

Wilkinson J, Heal C, Flemyng E, Antoniou GA, Aburrow T, Alfirevic Z, et al. INSPECT-SR: a tool for assessing trustworthiness of randomised controlled trials [Preprint]. medRxiv; 2025. doi:10.1101/2025.09.03.25334905