Prepare a Precise Request for Missing Study Information

Define the unresolved item

Before contacting an investigator, check the main report, supplement, registration and linked reports. Record exactly what remains unavailable. “Please send more data” creates unnecessary work and may invite information that is unsuitable for the planned synthesis. A useful request identifies the outcome, intervention arms, assessment time, analysis population and missing quantity.

Cochrane’s collection guidance includes correspondence with investigators as a source of study information. Treat the response as a documented source, not an informal replacement for the published extraction record.

Prepare the information package

List the article citation, study identifier and table or passage being clarified. Explain the minimum aggregate information needed and how it will be used. Distinguish a request for clarification from a request to conduct a new analysis. Where participant-level data are unnecessary, do not ask for identifiable records.

For example, a missing change-score SD is different from a missing baseline SD. A request should state the change direction, paired-analysis count and time window. If an alternative such as a standard error or compatible confidence interval would suffice, say so clearly and preserve the method needed to interpret it.

Work through a hypothetical reconciliation

Imagine a report gives 80 randomized participants and a mean change for 72 participants but no SD. The extraction request asks for the SD of individual changes among those same 72 people. An investigator replies with an SD of 12 for 70 complete pairs, explaining that two participants lack baseline measurements.

Do not paste 12 beside the count 72. Record the response’s actual denominator of 70, preserve the discrepancy, and clarify whether the mean change also refers to those 70 pairs. The response may solve the missing-SD problem while revealing that another field needs correction.

Track what was received

Keep request and response dates, recipient role, source identity, permission or use restrictions, and the exact supplied values. Preserve an unanswered status without converting missing information to zero. Set a reasonable, prespecified follow-up policy and a cutoff for the current review version.

When a response changes an extracted result, retain the previous value and explain the replacement. If the author cannot resolve the discrepancy, reflect that uncertainty in the analysis choice and limitations. Report the extent of unsuccessful retrieval without implying that nonresponse proves selective reporting. The practical output is an auditable information transaction that another reviewer can interpret without access to your memory of the conversation.

Source verification: 6 November 2026. Examples are hypothetical.