Meta-Analysis of Individual Participant Data: Rationale, Conduct, and Reporting

Individual participant data meta-analysis (IPD-MA) is a systematic review method in which researchers obtain, harmonize, and reanalyze the raw participant-level data from multiple studies instead of relying only on published summary statistics. This approach allows investigators to examine treatment effects in subgroups, explore treatment-covariate interactions, and address questions that cannot be answered from aggregate data alone. For researchers planning or evaluating such projects, the core value lies in the ability to standardize definitions across studies, conduct consistent analyses, and generate evidence that directly informs clinical and policy decisions. This article explains when IPD-MA is advantageous, how to conduct one, and what reporting standards apply, with practical guidance for each stage of the process.

What Distinguishes IPD Meta-Analysis from Conventional Meta-Analysis

Conventional meta-analysis combines effect estimates extracted from published reports. Each study contributes summary statistics such as means, proportions, or hazard ratios. IPD-MA instead collects the raw data for every participant in every included study. This distinction changes what questions can be asked and how confidently those questions can be answered.

The most important advantage is the capacity to examine individual-level treatment effect modification. When only aggregate data are available, subgroup analyses are limited to characteristics reported by the original study authors, and those subgroup definitions often differ across studies. With IPD, a researcher can apply uniform definitions of age groups, disease severity categories, or biomarker thresholds across all participants in all trials. A 2021 systematic review of IPD meta-analyses on intervention effects found that only 31% of published IPD-MAs prespecified methods for assessing participant-intervention interactions, which suggests that even among IPD-MA practitioners, the potential for interaction analysis is frequently underused (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review).

IPD also enables consistent handling of missing data, standardized outcome definitions, and reanalysis of outcomes that were not reported in the original publications. For example, a meta-analysis of 19 placebo-controlled statin trials with 123,940 participants was able to examine muscle symptoms using uniform definitions applied across all trials, something that would have been impossible with published summary data alone (Effect of statin therapy on muscle symptoms: an individual participant data meta-analysis of large-scale, randomised, double-blind trials).

A further advantage is the ability to estimate within-study correlations between outcomes. Published reports rarely provide these correlations, but they are required for multivariate meta-analysis. When IPD are available, correlations can be calculated directly from the data, enabling joint synthesis of multiple correlated outcomes and reducing selective outcome reporting bias (Multivariate meta-analysis using individual participant data).

When IPD Meta-Analysis Is Justified

IPD-MA requires substantially more time, resources, and collaboration than conventional meta-analysis. It is justified when the research question depends on individual-level information that cannot be obtained from aggregate data.

Examining Treatment-Covariate Interactions

The most common justification is the need to identify which participants benefit most from an intervention. A 2024 methods paper describes IPD meta-analysis projects as being frequently initiated to identify treatment effect modifiers at the individual level, requiring statistical modeling of interactions between treatment effect and participant-level covariates (Individual participant data meta-analysis to examine linear or non-linear treatment-covariate interactions at multiple time-points for a continuous outcome). For example, an IPD-MA of blood-pressure-lowering treatment stratified 358,707 participants from 51 trials into age groups and blood pressure categories to examine whether benefits varied by baseline characteristics (Age-stratified and blood-pressure-stratified effects of blood-pressure-lowering pharmacotherapy for the prevention of cardiovascular disease and death: an individual participant-level data meta-analysis).

Standardizing Definitions Across Studies

When studies use different definitions of outcomes, exposures, or participant characteristics, IPD allows the analyst to recode all data according to a single set of definitions. A meta-analysis of parenting program effects on disruptive behavior pooled data from 14 European trials with 3,252 families and applied uniform definitions of parenting behaviors including praise, tangible rewards, physical discipline, harsh verbal discipline, and following through on discipline (Individual Participant Data Meta-Analysis: Individual Differences in Mediators of Parenting Program Effects on Disruptive Behavior). This standardization would not have been possible with published summaries.

Addressing Time-to-Event Questions

IPD is particularly valuable for survival analyses because the timing of events can be reanalyzed consistently. A study of mortality after release from incarceration combined data from 1,471,526 people in eight countries and examined mortality rates in specific time windows after release, including the first week, weeks 3 through 12, and later periods (Rates and causes of death after release from incarceration among 1 471 526 people in eight high-income and middle-income countries: an individual participant data meta-analysis). This level of temporal detail requires event dates for each participant.

When IPD-MA Is Not Necessary

IPD-MA is not required when the research question can be answered with aggregate data. If the goal is simply to estimate an overall treatment effect and no subgroup analyses are planned, conventional meta-analysis is more efficient. Similarly, if the studies used nearly identical definitions and reported all relevant outcomes, the additional effort of obtaining IPD may not be justified. Researchers should also consider whether the original data are likely to be available. A 2021 review found that up to 39% of IPD meta-analyses failed to obtain IPD from 90% or more of eligible participants or trials, and among those, only 48% provided reasons and 17% undertook strategies to account for the unavailable data (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Feasibility assessment should occur before committing to an IPD-MA.

Core Principles of IPD Meta-Analysis Conduct

Prospective Protocol Development

A protocol should be written and registered before data collection begins. The 2021 systematic review found that only 31% of published IPD-MAs established an a priori protocol and only 44% prespecified methods for assessing overall effects (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). A protocol should specify the research question, eligibility criteria for studies, search strategy, data items to be requested, harmonization rules, statistical analysis plan, and methods for handling missing data and unavailable IPD.

The EQUATOR Network maintains reporting guidelines and resources that can inform protocol development. While EQUATOR primarily hosts reporting guidelines, its library includes materials relevant to systematic review methodology that researchers can consult when designing their studies.

Comprehensive Literature Search

The same review found that only 19% of IPD-MAs conducted a comprehensive literature search (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). A comprehensive search should include multiple bibliographic databases, trial registries, and reference lists of relevant papers. For example, a protocol for an IPD-MA of frozen versus fresh embryo transfer planned to search Medline, Embase, PsycINFO, CENTRAL, ClinicalTrials.gov, and the International Clinical Trials Registry Platform (Individual participant data meta-analysis of trials comparing frozen versus fresh embryo transfer strategy (INFORM): a protocol).

The National Center for Biotechnology Information provides access to literature databases including PubMed, which is operated by the National Library of Medicine. These resources are commonly used for systematic review searching.

Data Acquisition and Data Sharing Agreements

Obtaining IPD requires contacting study investigators and establishing data sharing agreements. These agreements should specify the purpose of the data use, data security requirements, publication rights, and conditions for data destruction or return. The Research Data Framework developed by the National Institute of Standards and Technology provides a structure for describing data management practices, including aspects relevant to data sharing and reuse.

A protocol for an IPD-MA of PTSD interventions describes the process of contacting authors to contribute participant-level datasets and merging them into a master dataset (Protocol for individual participant data meta-analysis of interventions for post-traumatic stress). This process requires careful documentation of which data were received, from whom, and under what conditions.

Data Harmonization

Harmonization is the process of transforming data from different studies into a common format. This includes recoding variable names, standardizing units of measurement, applying uniform definitions of outcomes and covariates, and resolving inconsistencies. The 2024 methods paper on non-linear treatment-covariate interactions emphasizes the need for a set-up phase to identify relevant knot positions for spline functions at identical locations in each trial and a common reference group for each covariate (Individual participant data meta-analysis to examine linear or non-linear treatment-covariate interactions at multiple time-points for a continuous outcome).

Harmonization decisions should be documented in a data dictionary that records the original variable definitions, the harmonized definitions, and any transformations applied. This documentation is essential for transparency and reproducibility.

Statistical Analysis Approaches

Two main approaches exist for analyzing IPD: one-stage and two-stage methods.

In a one-stage approach, all participant data are analyzed in a single model that accounts for clustering within studies. For example, the blood-pressure-lowering meta-analysis used a fixed effects one-stage approach with Cox proportional hazard models stratified by trial (Age-stratified and blood-pressure-stratified effects of blood-pressure-lowering pharmacotherapy for the prevention of cardiovascular disease and death: an individual participant-level data meta-analysis). A one-stage IPD-MA of tuberculosis prevalence surveys analyzed 602,863 participants from 12 surveys in a single model (Prevalence of subclinical pulmonary tuberculosis in adults in community settings: an individual participant data meta-analysis).

In a two-stage approach, each study is analyzed separately and the study-specific estimates are then combined using meta-analytic methods. The 2024 methods paper proposes a two-stage multivariate approach in which restricted cubic spline functions are fitted in each trial separately and the parameter estimates are jointly synthesized in a multivariate random-effects model (Individual participant data meta-analysis to examine linear or non-linear treatment-covariate interactions at multiple time-points for a continuous outcome). A case study of continuous exposures demonstrates a two-stage process in which response curves are estimated separately for each study using fractional polynomials and then averaged pointwise over all studies (Meta-analysis for individual participant data with a continuous exposure: A case study).

The choice between one-stage and two-stage approaches depends on the research question, the number of studies, the complexity of the data, and the statistical model being used. Both approaches can be valid when implemented correctly.

At a Glance: Key Decisions in IPD Meta-Analysis

Practical Workflow for Conducting an IPD Meta-Analysis

Step 1: Define the Research Question and Eligibility Criteria

The research question should specify the population, intervention, comparator, outcomes, and time frame. Eligibility criteria for studies should be defined in advance and should include the minimum data requirements. For example, the blood-pressure-lowering meta-analysis included trials with at least 1,000 person-years of follow-up in each treatment group and excluded participants with a previous history of heart failure (Age-stratified and blood-pressure-stratified effects of blood-pressure-lowering pharmacotherapy for the prevention of cardiovascular disease and death: an individual participant-level data meta-analysis).

Step 2: Register the Protocol

Protocol registration provides transparency and helps prevent duplication of effort. The protocol should be registered in a publicly accessible registry before data collection begins. The EQUATOR Network provides access to reporting guidelines that can inform protocol content.

Step 3: Conduct the Literature Search

Search multiple databases including PubMed, Embase, and CENTRAL. Search trial registries such as ClinicalTrials.gov. Check reference lists of relevant papers and previous reviews. Document the search strategy, including the date of search, search terms, and number of records retrieved. The National Library of Medicine provides access to PubMed for literature searching.

Step 4: Screen Studies and Assess Eligibility

Two independent reviewers should screen titles and abstracts, then full texts, against the eligibility criteria. Disagreements should be resolved by discussion or by a third reviewer. The screening process should be documented using a flow diagram that records the number of records identified, screened, excluded, and included.

Step 5: Contact Investigators and Establish Data Sharing Agreements

Contact the lead investigators of eligible trials to invite them to join the collaboration and share deidentified IPD. The INFORM protocol describes inviting lead investigators to join the collaboration and share deidentified IPD of their trials (Individual participant data meta-analysis of trials comparing frozen versus fresh embryo transfer strategy (INFORM): a protocol). Data sharing agreements should be signed before data transfer.

Step 6: Receive, Validate, and Harmonize Data

Upon receiving data, check for completeness, internal consistency, and alignment with the published reports. Create a data dictionary that documents variable names, definitions, coding schemes, and any transformations. Recode variables to a common format across all studies.

Step 7: Assess Risk of Bias

Assess the risk of bias of each included trial using an appropriate tool. The 2021 systematic review found that only 43% of IPD-MAs used a satisfactory technique to assess risk of bias and only 40% accounted for risk of bias when interpreting results (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Risk of bias assessment should be conducted by two independent reviewers.

Step 8: Conduct the Statistical Analysis

Follow the prespecified analysis plan. Conduct the primary analysis, then sensitivity analyses to test the robustness of the findings. If IPD are unavailable for some eligible studies, consider strategies to account for the unavailable data and document these strategies.

Step 9: Report the Findings

Report the IPD-MA according to established reporting guidelines. The EQUATOR Network provides access to reporting guidelines including PRISMA-IPD. The report should describe the search, study selection, data acquisition, harmonization, statistical methods, and results.

Records and Measurements in IPD Meta-Analysis

Documentation Requirements

Complete documentation is essential for transparency and reproducibility. The following records should be maintained throughout the project:

• Search strategies and results for each database
• Screening decisions for each record
• Eligibility decisions for each full-text article
• Data sharing agreements for each contributing study
• Data dictionaries for each original dataset and the harmonized dataset
• Analysis scripts and output
• Correspondence with study investigators

Measuring Data Completeness

Track the proportion of eligible studies and participants for whom IPD were obtained. The 2021 systematic review found that up to 39% of IPD-MAs failed to obtain IPD from 90% or more of eligible participants or trials (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Document the reasons for unavailable data and any strategies used to account for it.

Measuring Harmonization Success

Record the number of variables that required recoding, the number of discrepancies identified during data validation, and the number of participants affected by each harmonization decision. This information helps readers understand the degree of standardization achieved.

Common Failure Patterns in IPD Meta-Analysis

Incomplete Data Acquisition

The most common failure is the inability to obtain IPD from all eligible studies. Investigators may decline to participate, may no longer have access to the data, or may be unable to share data due to contractual or regulatory restrictions. The 2021 review found that up to 39% of IPD-MAs failed to obtain IPD from 90% or more of eligible participants or trials (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). This failure can introduce selection bias if the studies that provide data differ systematically from those that do not.

Inadequate Risk of Bias Assessment

Many IPD-MAs fail to assess risk of bias adequately. The 2021 review found that only 43% used a satisfactory technique to assess risk of bias and only 40% accounted for risk of bias when interpreting results (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Risk of bias assessment is important because the quality of the included trials affects the confidence that can be placed in the pooled estimates.

Insufficient Prespecification

Only 31% of IPD-MAs established an a priori protocol and only 31% prespecified methods for assessing participant-intervention interactions (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Without prespecification, there is a risk of selective reporting and data-driven analysis.

Incomplete Reporting of Excluded Studies

Only 32% of IPD-MAs provided a list of excluded studies with justifications (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Readers need this information to assess the completeness of the evidence base.

Inadequate Handling of Unavailable IPD

Among IPD-MAs that failed to obtain IPD from 90% or more of eligible participants or trials, only 48% provided reasons and only 17% undertook strategies to account for the unavailable IPD (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Failure to address unavailable data can bias the results.

Ethics and Governance Considerations

Ethics Approval for the IPD-MA Itself

IPD-MAs may require ethics approval, depending on the jurisdiction and the nature of the data. The INFORM protocol reports that ethics approval was granted by the Monash University Human Research Ethics Committee (Individual participant data meta-analysis of trials comparing frozen versus fresh embryo transfer strategy (INFORM): a protocol). Researchers should check with their institutional review board or research ethics committee about whether approval is required.

Ethics Documentation for Included Trials

A 2026 analysis of ethics reporting in IPD-MAs identified five ethics items that are inconsistently reported: whether approval from a research ethics committee was obtained for the included trials, whether participants of those trials provided informed consent, whether the IPD-MA itself was approved by a research ethics committee, the de-identification status of the data (pseudonymized versus anonymized), and whether participants in all included trials had given consent or had not objected to the secondary use and sharing of their data with third parties (Inconsistent ethics reporting in individual participant data meta-analyses: a call for updating PRISMA and PRISMA-IPD). These items should be reported in the text of the IPD-MA or in the disclosure section.

Data De-identification

Data shared for IPD-MA should be de-identified to protect participant privacy. The distinction between pseudonymized and anonymized data is important because pseudonymized data can potentially be re-identified if the key is available. Researchers should verify the de-identification status of all data received and document this in the report.

Data Security

Data sharing agreements should specify data security requirements, including storage, access controls, and transmission methods. Researchers should ensure that data are stored securely and that access is limited to authorized personnel.

Reporting Standards for IPD Meta-Analysis

PRISMA-IPD

The PRISMA-IPD statement provides reporting guidance specific to IPD-MA. The EQUATOR Network provides access to this guideline and other reporting resources. PRISMA-IPD extends the PRISMA statement to address issues specific to IPD, including the methods for obtaining and harmonizing data, the proportion of eligible studies for which IPD were obtained, and the methods for handling unavailable data.

Calls for Updated Reporting Standards

A 2026 analysis called for updating PRISMA and PRISMA-IPD to include five ethics items that are not currently included in either guideline (Inconsistent ethics reporting in individual participant data meta-analyses: a call for updating PRISMA and PRISMA-IPD). Researchers should consider reporting these items even before the guidelines are updated.

Reporting the Search

The report should describe the search strategy in sufficient detail to allow replication. This includes the databases searched, the search terms used, the date of the search, and the number of records retrieved. The National Library of Medicine provides access to PubMed, which is a primary database for biomedical literature.

Reporting Data Acquisition and Harmonization

The report should describe how investigators were contacted, how many agreed to participate, how many provided data, and how the data were harmonized. Any discrepancies between the IPD and the published reports should be described.

Reporting Statistical Methods

The report should describe the statistical methods in sufficient detail to allow replication. This includes the analysis approach (one-stage or two-stage), the statistical models used, the handling of missing data, and the methods for assessing heterogeneity and publication bias.

Limitations and Caveats

Data Availability Constraints

IPD-MA depends on the willingness and ability of study investigators to share data. Some investigators may be unable to share data due to contractual obligations, regulatory restrictions, or concerns about participant privacy. The 2021 review found that up to 39% of IPD-MAs failed to obtain IPD from 90% or more of eligible participants or trials (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). This limitation should be acknowledged and addressed in the analysis.

Resource Intensity

IPD-MA requires substantially more time and resources than conventional meta-analysis. The process of contacting investigators, negotiating data sharing agreements, receiving and validating data, and harmonizing variables can take many months or years. Researchers should plan for this timeline and secure adequate funding.

Statistical Complexity

The statistical methods for IPD-MA are more complex than those for conventional meta-analysis. Analysts need expertise in advanced statistical modeling, including methods for handling missing data, non-linear relationships, and correlated outcomes. The 2024 methods paper on non-linear treatment-covariate interactions illustrates the complexity of these methods (Individual participant data meta-analysis to examine linear or non-linear treatment-covariate interactions at multiple time-points for a continuous outcome).

Risk of Bias in Included Trials

IPD-MA cannot overcome the limitations of the included trials. If the original trials have high risk of bias, the IPD-MA will inherit that bias. The 2021 review found that only 43% of IPD-MAs used a satisfactory technique to assess risk of bias (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Risk of bias should be assessed and accounted for in the interpretation of results.

Publication Bias

IPD-MA may be subject to publication bias if studies with null or negative results are less likely to be published and therefore less likely to be identified and included. Only 31% of IPD-MAs assessed and considered the potential of publication bias (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Methods for assessing publication bias should be prespecified and reported.

Professional Escalation Criteria

Researchers should seek additional expertise or escalate concerns in the following situations:

• When the proportion of eligible studies for which IPD can be obtained falls below a level that would allow meaningful analysis. The 2021 review used a threshold of 90% of eligible participants or trials to define adequate data acquisition (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Researchers should consider whether the available data are sufficient to answer the research question.
• When discrepancies between IPD and published reports cannot be resolved. These discrepancies should be documented and discussed with the data contributors.
• When statistical methods are required that exceed the expertise of the research team. Consultation with a statistician with experience in IPD-MA should be sought.
• When ethics or governance questions arise that cannot be resolved within the research team. Consultation with the institutional review board or research ethics committee should be sought.
• When data sharing agreements cannot be reached with study investigators. Alternative strategies, such as seeking data from repositories or using summary data for unavailable studies, should be considered.

Frequently Asked Questions

What is the difference between one-stage and two-stage IPD meta-analysis?

In a one-stage approach, all participant data are analyzed in a single statistical model that accounts for clustering within studies. In a two-stage approach, each study is analyzed separately and the study-specific estimates are then combined using meta-analytic methods. The blood-pressure-lowering meta-analysis used a one-stage approach with Cox models stratified by trial (Age-stratified and blood-pressure-stratified effects of blood-pressure-lowering pharmacotherapy for the prevention of cardiovascular disease and death: an individual participant-level data meta-analysis). The 2024 methods paper on non-linear interactions demonstrates a two-stage multivariate approach (Individual participant data meta-analysis to examine linear or non-linear treatment-covariate interactions at multiple time-points for a continuous outcome). The choice depends on the research question and the complexity of the data.

How long does an IPD meta-analysis take to complete?

The timeline varies depending on the number of studies, the availability of data, and the complexity of the analyses. The process of contacting investigators, negotiating data sharing agreements, receiving and validating data, and harmonizing variables can take many months. Researchers should plan for a longer timeline than conventional meta-analysis and secure adequate funding.

What should be included in a data sharing agreement for IPD meta-analysis?

A data sharing agreement should specify the purpose of the data use, data security requirements, publication rights, conditions for data destruction or return, and the responsibilities of both the data provider and the data recipient. The Research Data Framework from the National Institute of Standards and Technology provides a structure for describing data management practices relevant to data sharing.

How should missing data be handled in IPD meta-analysis?

Missing data can occur at the participant level, the variable level, or the study level. Complete case analysis may introduce bias if data are not missing completely at random. Imputation methods require assumptions about the missingness mechanism. The analysis plan should prespecify the approach to missing data. The 2024 methods paper notes that the multivariate approach can include participants or trials with missing outcomes at some time-points (Individual participant data meta-analysis to examine linear or non-linear treatment-covariate interactions at multiple time-points for a continuous outcome).

What is the role of risk of bias assessment in IPD meta-analysis?

Risk of bias assessment evaluates the internal validity of the included trials. The 2021 systematic review found that only 43% of IPD-MAs used a satisfactory technique to assess risk of bias and only 40% accounted for risk of bias when interpreting results (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Risk of bias should be assessed by two independent reviewers and accounted for in the interpretation of results.

How should unavailable IPD be handled?

When IPD cannot be obtained for some eligible studies, researchers should document the reasons and consider strategies to account for the unavailable data. The 2021 review found that among IPD-MAs that failed to obtain IPD from 90% or more of eligible participants or trials, only 48% provided reasons and only 17% undertook strategies to account for the unavailable IPD (The methodological quality of individual participant data meta-analysis on intervention effects: systematic review). Possible strategies include using published summary data for unavailable studies or conducting sensitivity analyses to assess the potential impact of the missing data.

What ethics approval is needed for IPD meta-analysis?

Ethics requirements vary by jurisdiction. The INFORM protocol reports that ethics approval was granted by the Monash University Human Research Ethics Committee (Individual participant data meta-analysis of trials comparing frozen versus fresh embryo transfer strategy (INFORM): a protocol). A 2026 analysis identified five ethics items that should be reported: ethics approval for included trials, informed consent of trial participants, ethics approval for the IPD-MA itself, de-identification status of the data, and consent for secondary use and data sharing (Inconsistent ethics reporting in individual participant data meta-analyses: a call for updating PRISMA and PRISMA-IPD). Researchers should check with their institutional review board about requirements.

What reporting guidelines apply to IPD meta-analysis?

PRISMA-IPD provides reporting guidance specific to IPD-MA. The EQUATOR Network provides access to this guideline and other reporting resources. A 2026 analysis called for updating PRISMA and PRISMA-IPD to include five ethics items that are not currently included (Inconsistent ethics reporting in individual participant data meta-analyses: a call for updating PRISMA and PRISMA-IPD). Researchers should report these items even before the guidelines are updated.

Related Articles

• Statistical Data Analysis
• Statistical Data Analysis
• Statistical Data Analysis
• Genomics Services: How to Evaluate Sequencing, Analysis, and Reporting Options
• Genomics Services: How to Evaluate Sequencing, Analysis, and Reporting Options

References and Further Reading

• Research Data Framework. National Institute of Standards and Technology.
• EQUATOR Network. EQUATOR Network.
• Experimental Design Assistant. NC3Rs.
• NCBI Literature Resources. National Center for Biotechnology Information.
• PubMed. National Library of Medicine.
• Age-stratified and blood-pressure-stratified effects of blood-pressure-lowering pharmacotherapy for the prevention of cardiovascular disease and death: an individual participant-level data meta-analysis.. Lancet (London, England), 2021.
• Individual participant data meta-analysis to examine linear or non-linear treatment-covariate interactions at multiple time-points for a continuous outcome.. Research synthesis methods, 2024.
• Protocol for individual participant data meta-analysis of interventions for post-traumatic stress.. BMJ open, 2022.
• Meta-analysis for individual participant data with a continuous exposure: A case study.. Journal of clinical epidemiology, 2021.
• The methodological quality of individual participant data meta-analysis on intervention effects: systematic review.. BMJ (Clinical research ed.), 2021.
• Individual Participant Data Meta-Analysis: Individual Differences in Mediators of Parenting Program Effects on Disruptive Behavior.. Journal of the American Academy of Child and Adolescent Psychiatry, 2025.
• Individual participant data meta-analysis of trials comparing frozen versus fresh embryo transfer strategy (INFORM): a protocol.. BMJ open, 2022.
• Multivariate meta-analysis using individual participant data.. Research synthesis methods, 2015.
• Inconsistent ethics reporting in individual participant data meta-analyses: a call for updating PRISMA and PRISMA-IPD.. 2026.
• The Interpupillary Distance Differs Between Ethnicities and Associates with Horizontal Strabismus Patterns: Evidence from a Systematic Review and Meta-Analysis.. 2026.
• Comprehensive Safety and Efficacy Evaluation of Immunotherapy Combination Approaches Versus Tyrosine Kinase Inhibitor Monotherapy as First-Line Treatment of Hepatocellular Carcinoma: A Network and Individual Patient Data (IPD) Meta-Analysis.. 2026.
• A Possible Unmet Need: Pneumococcal Vaccination in the Workplaces-A Systematic Review of Invasive Pneumococcal Disease Among Shipyard Workers.. 2026.
• Body-mass index and all-cause mortality: individual-participant-data meta-analysis of 239 prospective studies in four continents. The Lancet, 2016.
• The relationship between obesity and obstructive sleep apnea in four community-based cohorts: an individual participant data meta-analysis of 12,860 adults. EClinicalMedicine, 2025.
• Effects of statin therapy on diagnoses of new-onset diabetes and worsening glycaemia in large-scale randomised blinded statin trials: an individual participant data meta-analysis. The Lancet Diabetes and Endocrinology, 2024.
• Performance of non-invasive tests and histology for the prediction of clinical outcomes in patients with non-alcoholic fatty liver disease: an individual participant data meta-analysis.. The Lancet Gastroenterology and Hepatology, 2023.
• Effect of statin therapy on muscle symptoms: an individual participant data meta-analysis of large-scale, randomised, double-blind trials. The Lancet, 2022.
• Prevalence of subclinical pulmonary tuberculosis in adults in community settings: an individual participant data meta-analysis.. Lancet. Infectious Diseases (Print), 2024.
• Rates and causes of death after release from incarceration among 1 471 526 people in eight high-income and middle-income countries: an individual participant data meta-analysis.. The Lancet, 2024.
• Value of intravenous thrombolysis in endovascular treatment for large-vessel anterior circulation stroke: individual participant data meta-analysis of six randomised trials.. The Lancet, 2023.

This article is educational and does not replace institutional policy, professional advice, or applicable safety and regulatory requirements.