Link Multiple Reports Without Merging Different Studies

Separate a record duplicate from a study family

Two database records describing the same article can usually be deduplicated at the citation level. Two different articles describing the same participants need a different operation: retain both reports and link them to one study. Deleting the second article can discard longer follow-up, additional outcomes or methods needed for appraisal.

Cochrane’s study-selection guidance distinguishes reports from studies. Turn that distinction into separate report and study identifiers in the extraction database. A report identifier should never change merely because its study-family assignment changes.

Assemble matching evidence

Compare registration identifiers, investigators, recruitment sites and dates, intervention descriptions, eligibility criteria, randomized sample sizes and baseline characteristics. Store the exact passage or table supporting each match. A common author or similar title is weak evidence; an explicit statement that the article is a secondary analysis of a named trial is much stronger.

Use three practical statuses: confirmed link, plausible link requiring investigation, and distinct study. Require a second reviewer to examine uncertain or consequential merges. Preserve the reason for changing a status so later reviewers do not repeat the same investigation.

Reconcile a hypothetical family

Suppose reports A and B both describe a trial of 240 randomized participants. Report A gives the primary 12-week result; B gives a secondary biomarker analysis in 180 of those participants. Report C comes from the same investigators but recruits 90 participants two years later.

Link A and B to one trial and keep C separate unless additional evidence shows overlap. The family does not contain 240 + 180 = 420 independent participants. The biomarker denominator is a subset whose selection and missingness need explanation. If the main review uses the 12-week outcome, B may contribute methodological information without adding another effect estimate.

Maintain outcome-specific independence

A confirmed study family does not mean every result can be collapsed into one row. Different outcomes and times still need separate extraction fields. Conversely, two articles can report partially overlapping cohorts without being identical studies. Record the overlap and choose an analysis that avoids treating shared participants as independent.

Audit the final synthesis by tracing each effect estimate back to its study identifier, report, outcome and population. Where overlap cannot be resolved, state the uncertainty and examine a justified alternative inclusion choice. The goal is neither maximum deletion nor maximum retention: it is an explicit map showing which observations can legitimately contribute together.

Source verification: 6 November 2026. Examples are hypothetical.