How to Find Methods Details Hidden in Supplements and Registries

A trial's main paper rarely tells the whole methods story. Planned outcomes, analysis plans, and protocol deviations usually sit in supplements, registration records, and linked reports. For a time-sensitive decision memo or due-diligence brief, the job is to follow those links deliberately, record where each detail came from, and state how confident you are in it.

This guide gives a repeatable sequence for that work, with diagnostic checkpoints, a worked example, a reusable checklist, and clear rules for when to stop or escalate.

Cochrane's guidance on searching for studies begins with eligibility criteria: define what counts as a relevant study before you search, then build the search from those criteria. That discipline matters here too. If you start from a single paper and chase its links, you will find that paper's supplements but may miss sibling reports, earlier registrations, or companion publications.

Before opening any supplement, write down:

• The population, intervention, comparator, and outcomes your memo actually needs.
• Which study types count (for example, only registered trials, or also protocols and secondary analyses).
• The decision the memo supports, so you know when detail is sufficient.

This is the same logic NIH guidance applies to rigorous research practice: assess prior knowledge, strengths, weaknesses, and knowledge gaps before drawing conclusions. For a memo, that means knowing what you already have and what is still missing.

Cochrane recommends using multiple sources and combining free-text terms with subject headings, then documenting the process. In practice, treat each study as a cluster of documents rather than a single article.

A useful source map for one study includes:

1. The primary report — main results and the methods summary.
2. Supplements — protocol summaries, outcome definitions, analysis populations, sensitivity analyses.
3. Trial registration records — planned outcomes, planned sample size, status, and history of changes.
4. Protocols and statistical analysis plans — the planned comparisons and how missing data were to be handled.
5. Linked reports — companion papers, secondary analyses, and follow-up publications.

Cochrane specifically notes the need to link multiple reports of one study. Do that explicitly: give each study a single internal ID and attach every document to it. When you search, use WisPaper search to locate candidate papers, then keep the cluster together rather than treating each hit as a separate study.

Checkpoint 1: You can name every document in the cluster and say which are primary, which are supplementary, and which are linked reports. If you cannot, you are not ready to extract methods details.

Supplements are where planned outcomes and analysis details most often hide, but they are only useful if you read them against the registration record. Use a fixed order so you do not anchor on the main paper's framing.

Open the registry entry before the supplement. Record the registration date, the planned primary and secondary outcomes, the planned sample size, and any recorded history of changes. The registration is your baseline for what was planned.

Where a protocol or statistical analysis plan is available, note the planned comparisons, the analysis populations, and the handling of missing data. These are the details that determine whether a reported result matches what was planned.

Now open the supplement. For each planned outcome, ask:

• Is it reported in the main paper, the supplement, or both?
• Is the definition the same as in the registration record?
• Is the analysis population the same as planned?
• Are there outcomes in the registration that do not appear anywhere in the report?

A deviation is any difference between what was planned and what was reported. Log it with the document, the section, and the exact wording. Do not paraphrase a deviation into a conclusion yet.

Checkpoint 2: Every planned outcome has a status — reported as planned, reported with a changed definition, or not reported. Every status has a document reference.

PRISMA 2020 emphasizes transparent reporting of search, selection, and synthesis methods, and provides flow-diagram templates for that purpose. Your memo needs the same transparency at a smaller scale. A provenance table is the simplest way to get it.

Two rules keep the table honest:

• One row per claim per source. If a claim rests on two documents, it gets two rows.
• Confidence reflects source quality, not your certainty about the result. A planned outcome confirmed in both the registration and the protocol is high confidence as a plan. Whether the trial delivered it is a separate question.

When you need to check that the references you are citing are real and correctly matched, TrueCite screens BibTeX against academic databases to flag hallucinated references. Use that citation-integrity result to prioritize which source records you open and verify first.

Suppose your team is preparing a client briefing on whether a trial's reported primary outcome matches its plan. The main paper reports a primary outcome and a sample size. The supplement contains an outcome definitions table. The registration record lists two primary outcomes.

What you do, in order:

1. Map the cluster. Primary report, supplement, registration record, and any linked reports. Assign one study ID.
2. Extract the plan. From the registration record, note both planned primary outcomes and the planned sample size. From the protocol, note the planned analysis population.
3. Extract the report. From the main paper and supplement, note the reported primary outcome, its definition, and the analyzed population.
4. Compare. One planned primary outcome is reported with a matching definition. The second planned primary outcome does not appear in the main paper or the supplement.
5. Log the deviation. Record the missing outcome with the registration section and the report sections you checked.
6. Assign confidence. The plan is high confidence (registration plus protocol agree). The deviation is high confidence (absence confirmed across both report documents). The interpretation of the deviation is lower confidence until you check linked reports for a secondary analysis.
7. Check linked reports. If a companion paper reports the second outcome, update the deviation row to note where it appears.

What the memo says: State the planned outcomes, the reported outcomes, the deviation, and the confidence level for each. Keep the interpretation separate from the extraction. A reader should be able to trace every sentence back to a row in the provenance table.

Checkpoint 3: A colleague who has not read the sources can follow each memo claim to a document and a location. If not, the table is incomplete.

Supplement chasing generates many files: registration exports, protocol PDFs, supplements, and linked reports. Losing track of which version you read is a common source of error.

Save or upload the papers in My Library so the cluster stays together and you can return to the same documents when the memo is reviewed. When you need to ask questions across the papers you have collected, Library QA lets you query papers uploaded to My Library or added from search results. Use it to locate a detail quickly, then confirm the detail in the source document before it enters the provenance table.

For longer research tasks that involve reviewing a body of literature around the trial, Scholar Agent supports literature review work within WisPaper's agent area. Keep the agent's output as a finding aid; the provenance table still points to the primary documents.

A time-sensitive memo needs a defined end point. Use these rules.

Stop and report when:

• Every planned outcome has a status and a document reference.
• Every deviation has been checked against the registration record and the protocol.
• Linked reports have been checked for outcomes missing from the primary report.
• Confidence levels are assigned and the basis for each is stated.

Escalate when:

• The registration record and the protocol disagree on a planned outcome, and you cannot determine which governed the analysis.
• A planned outcome is missing from every document in the cluster, including linked reports.
• The supplement and the main paper define the same outcome differently, and the difference could change the memo's conclusion.
• You cannot confirm that a cited reference is real or correctly matched.

Escalation means flagging the specific gap to the memo's owner with the documents you checked. It does not mean filling the gap with an assumption.

Cochrane recommends rerunning searches before publication. The same applies to a memo with a short shelf life. Before the briefing goes out, rerun the search for the study cluster and check for new linked reports, updated registration records, or newly posted supplements. Update the provenance table and note the rerun date.

PRISMA 2020's emphasis on transparent reporting of search and selection methods supports this: record what you searched, when, and what changed. A memo that states its search date and its rerun date is easier to defend than one that does not.

• Eligibility criteria and decision purpose written down before searching.
• Study cluster mapped: primary report, supplements, registration, protocol, linked reports.
• One internal study ID assigned; all documents attached.
• Registration record read before the supplement.
• Protocol or analysis plan read for planned comparisons and populations.
• Supplement read against registration and protocol, outcome by outcome.
• Every planned outcome given a status and a document reference.
• Every deviation logged with document, section, and exact wording.
• Provenance table complete: one row per claim per source.
• Confidence level assigned to each claim, with its basis stated.
• Linked reports checked for outcomes missing from the primary report.
• Stop or escalate decision made using the rules above.
• Search rerun before publication; rerun date recorded.