Study-aware participant-level pooled analysis¶

Joint analysis of comparable participant records from multiple studies while accounting for their study of origin.

Version: v1 · 2026-10-17 · History
Domain-specific #: 14243
Domain group: Applied Sciences & Engineering
Origin domain: Medicine & Healthcare
Subdomain: Individual Participant Data Synthesis → Medicine & Healthcare

Core Idea¶

Study-aware participant-level pooled analysis jointly analyzes individual records from at least two source studies to answer a common question. The analyst establishes that the relevant variables can be interpreted together and keeps each participant's study of origin available to the estimate. A larger combined file alone does not establish this method.[^{ref-997f85c8ad71][}ref-65d06672c307]

The qualified title identifies a collective-data subtype, not every published use of pooled analysis. Authors may use the bare term for a two-stage method that first estimates effects within each study. Conversely, a one-stage individual participant data (IPD) meta-analysis can instantiate this subtype when it jointly estimates from comparable participant records and accounts for study origin.[^{ref-9eaeeb7419c3][}ref-65d06672c307]

Scope of Application¶

The method applies when clinical or epidemiologic studies supply person-level records for a shared question. Randomized trials can combine treatment and event records across trials; observational studies can combine exposure, outcome and covariate histories across contributing projects. The studies need not share every protocol or population, but question-relevant variables must be comparable and the joint analysis must represent source-study structure.[^{ref-21d1810510f1][}ref-afcf33be8749][^ref-997f85c8ad71]

Clarity¶

Ask four questions: Which independent studies supplied participant records? Which variables were made comparable for the common question? How were the records jointly analyzed? Where did study origin enter that analysis? Harmonizing an outcome does not remove source clustering; including a study label cannot repair variables with incompatible meanings.[^{ref-997f85c8ad71][}ref-65d06672c307]

The word pooled alone answers none of these questions. Published usage covers both joint participant-level estimation and combination of already estimated study results, so the analytic operation must be named.[^{ref-9eaeeb7419c3][}ref-997f85c8ad71]

Manages Complexity¶

Different studies may code events or exposures differently. A variable crosswalk can make a shared target workable while preserving documented differences. Ross and colleagues reconciled adverse-event terms reported under two vocabularies; Ge and colleagues applied a common job-exposure matrix to occupational histories from multiple case-control studies.[^{ref-21d1810510f1][}ref-afcf33be8749]

Joint estimation compresses many records into a common relation, while retaining study origin limits what is lost. A common coefficient, when reported, does not prove identical effects, populations or measurements across contributing studies.[^{ref-997f85c8ad71][}ref-65d06672c307]

Abstract Reasoning¶

Trace the full information chain from source studies through participant records and comparable variables to a joint estimate that handles source study. Replace the records with only published coefficients and this subtype is unavailable. Analyze the records separately and then combine study coefficients, and the procedure is a two-stage neighbor. Delete source identifiers from a joint file, and the computation may remain possible while losing a defense against clustering and hidden differences.[^{ref-9eaeeb7419c3][}ref-997f85c8ad71][^ref-65d06672c307]

Knowledge Transfer¶

The same four roles appear in unlike designs: randomized drug-safety trials and observational occupational case-control studies. The specific exposure, outcome, model and design assumptions change; the requirements for person-level data, question-relevant comparability, joint analysis and source-study handling remain.[^{ref-21d1810510f1][}ref-afcf33be8749] Neither a numerical effect nor its causal interpretation transfers between those settings.

Example¶

Randomized trials. Ross and colleagues jointly analyzed subject-level data from placebo-controlled rofecoxib trials. They reconciled cardiovascular event terms across CRISP and MedDRA reporting and used Cox models stratified by study, with exact methods for sparse arms.[^ref-21d1810510f1] The trials supply the source studies, subjects supply records, event reconciliation supports comparability, the model estimates jointly, and stratification retains trial origin.

Occupational case-control studies. Ge and colleagues combined individual occupational and smoking histories from 14 European and Canadian case-control studies, assigned silica exposure through a common job-exposure matrix, and estimated lung-cancer associations with logistic models adjusted for study.[^ref-afcf33be8749] The contributing projects supply source studies, their participants supply records, common exposure construction supports comparability, and study adjustment retains origin in the joint estimate.

Neighborhood in Abstraction Space¶

Study-aware participant-level pooled analysis sits in a sparse region of the domain-specific corpus (95^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2721 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-17

Not to Be Confused With¶

Every use of pooled analysis. The bare term also names methods that combine study-specific estimates.[^ref-9eaeeb7419c3]
Published-estimate meta-analysis. It uses study results rather than jointly reanalyzing participant records.[^ref-997f85c8ad71]
Every IPD meta-analysis. One-stage and two-stage routes differ in their collective operation.[^ref-65d06672c307]
An unstratified mega-study. Concatenating records without attending to source study does not establish the study-aware operation.[^ref-65d06672c307]

References¶

[^ref-21d1810510f1]: Joseph S. Ross, David Madigan, Kevin P. Hill, David S. Egilman, Yongfei Wang, and Harlan M. Krumholz (2009), “Pooled Analysis of Rofecoxib Placebo-Controlled Clinical Trial Data, Lessons for Post-Market Pharmaceutical Safety Surveillance,” Archives of Internal Medicine 169(21), 1976–1985, original author manuscript, pp. 2–4, “Data source,” “Main outcome measure,” and “Data analysis.” Published title uses a colon after “Data.” https://doi.org/10.1001/archinternmed.2009.394 ; full author manuscript https://pmc.ncbi.nlm.nih.gov/articles/PMC2830805/ [^ref-afcf33be8749]: Calvin Ge et al., “Respirable Crystalline Silica Exposure, Smoking, and Lung Cancer Subtype Risks, A Pooled Analysis of Case-Control Studies,” American Journal of Respiratory and Critical Care Medicine 202, 412–421 (2020), printed pp. 412–414, Abstract, “Study Population,” “Exposure Assessment,” and “Statistical Analysis.” Published title uses a colon after “Risks.” https://doi.org/10.1164/rccm.201910-1926OC ; original PDF https://dspace.library.uu.nl/bitstream/handle/1874/408764/rccm.201910_1926oc.pdf?sequence=1 [^ref-997f85c8ad71]: Isabel Fortier et al., “Maelstrom Research Guidelines for Rigorous Retrospective Data Harmonization,” International Journal of Epidemiology 46(1), 103–115 (2017), printed p. 104, Introduction, and pp. 108–110, harmonization steps and Table 2. https://doi.org/10.1093/ije/dyw075 ; version of record https://pure.tue.nl/ws/files/72437527/dyw075.pdf [^ref-65d06672c307]: Jayne F. Tierney, Lesley A. Stewart, and Mike Clarke (2024), “Individual participant data, Cochrane Handbook chapter 26 version 6.5,” in Cochrane Handbook for Systematic Reviews of Interventions, chapter 26, §§26.1.1, 26.3.1, 26.4.2; chapter last updated October 2019. Official chapter title is “Individual participant data.” https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-26 [^ref-9eaeeb7419c3]: Stephanie A. Smith-Warner et al., “Methods for Pooling Results of Epidemiologic Studies, The Pooling Project of Prospective Studies of Diet and Cancer,” American Journal of Epidemiology 163(11), 1053–1064 (2006), original abstract, Methods. Published title uses a colon after “Studies.” https://doi.org/10.1093/aje/kwj127