Resolve shared-control contributions before synthesis

Draw the comparison before entering the result

Cochrane Chapter 23, checked 7 November 2026, explains that shared-control comparisons are correlated and require appropriate handling. The analyst owns the contribution method.

List the study’s actual arms and identify which pair each proposed estimate compares. Give the arms stable labels that remain recognizable across reports. Keep this map separate from the effect-size table so an analyst can see the design even if the reported numbers are incomplete. A comparison name such as new design versus usual practice may conceal which particular usual-practice group supplied the observations.

Limit the initial map to published or authorized aggregate information. This exercise does not require collecting participant identities. Mark suspected reuse where the reports are unclear, and identify the missing design detail that would resolve it. Do not assume that similarly labeled controls are identical, or that differently labeled controls are separate, without examining the study description.

Separate relevance from contribution structure

Consider a fictional business-document trial with guided examples, searchable help and static instructions. The review includes both enhanced approaches. Draw its design before discussing numerical results; the diagram should let the analyst ask a precise question about the proposed input. Both contrasts use the static-instruction arm.

Now ask what the proposed synthesis is intended to answer. Does it concern enhanced instruction generally, or does the business decision distinguish guided examples from search? Keep that distinction explicit. Define the substantive question before choosing a convenient arithmetic repair. Keep alternative contribution plans visible until the review owner and analyst agree on the target.

Choose a justified treatment of the relationship

Ask the analyst to explain the proposed contribution plan in words as well as in its implementation. Which comparisons remain, which information is combined and what uncertainty assumptions are used? Do not independently repair the table merely because the software rejects it. A mechanically tidy input is not evidence that the design relationship has been handled appropriately.

Use a small synthetic input to verify that the approved plan is implemented as described. Trace each original arm into the contribution table and account for any excluded comparison. Keep a record of why a choice was made before examining which method produces the most favorable headline. If relevant aggregate information is missing, preserve the limitation rather than inventing a correlation or pretending that one software default supplies it.

Retain the design behind the reported summary

Include the dependency map and the approved contribution rule in the analysis handoff. Report the number of studies separately from the number of comparisons where both help explain the evidence base. A reader should be able to identify the role of a multi-arm study without reconstructing the source publications or assuming that every plotted row is a separate experiment.

If the review later adds another result from the same investigation, revisit the map before appending it to the model input. The new result may answer a different question, but its relationship to the existing evidence still needs consideration. Close with an explicit scope: the approved plan handles the documented relationships in this synthesis. It does not certify every study’s quality or replace the separate decisions about eligibility, effect definition and meaningful grouping.

Make the reasoning concrete.

A fictional editor receives a comparison table without its study-design diagram. The analyst cannot establish how its proposed rows relate.

The team resolves the contribution plan before the synthesis is interpreted. It does not manufacture new records or choose a method because it strengthens the result.

Take the questions into your next brief.

Use these prompts in your own notes. Nothing is submitted from this page.

1. Arms

   Which actual groups produced each contrast?

2. Relationship

   What relationship must the analyst handle?

3. Question

   Would combining treatments preserve the decision being studied?

4. Plan

   What analyst-approved contribution rule will be used?

5. Trace

   Can every arm be followed into the final input?

Watch for the wrong conclusion.

• Treating a spreadsheet row as a complete design description.
• Repairing sample sizes without an approved method.
• Combining treatments that answer different business questions.

Sources & scope

Original adult business decision scenarios are explicitly fictional. Methodological claims and their applications remain source-derived. Qualified analysts own formal calculations. Cochrane-specific guidance is not a universal mandate. No actual research operations or publication occur.

• Cochrane Handbook, Chapter 23: Including variants on randomized trials →Primary technical documentation · Supports: Supports shared-control dependence.