How to Update a Clinical Evidence Brief with New Studies

An evidence brief ages the moment it is finalized. New trials appear, systematic reviews are refreshed, and the search that produced the original document becomes stale. Updating a brief is not a matter of pasting in the newest headline study; it is a controlled process that starts from a defined question, runs a documented search, screens and links records, judges what the new evidence changes, and records a clear decision about whether the brief is ready to circulate.

This article gives clinicians and medical researchers a practical sequence for that work. It uses one focused example throughout — an adult population and a single clinical outcome — and separates evidence level, applicability, indirectness, and limitations so that each judgment is made on its own terms. It describes tools only at the steps where they genuinely help, and it ends with checkpoints, a reusable checklist, and stop rules.

Most failed updates begin with an unclear starting point. Before opening any database, establish four things about the brief you are updating.

The question and its scope. Write the population, intervention or exposure, comparator, and outcome as a single sentence. If the existing brief does not state them, reconstruct them from its methods section rather than inventing a broader question.

The date of the last search. An update is defined by what has appeared since the previous search date, not by what looks new. Record the exact date and the sources that were searched.

The eligibility criteria. Cochrane guidance on searching and selecting studies recommends planning searches directly from eligibility criteria. If the criteria are missing or vague, the update will drift.

The decision the brief supports. A brief that informs a team discussion has different stopping rules from one that feeds a formal recommendation. Knowing the decision tells you how much new evidence is enough to trigger a rewrite.

A quick diagnostic: if you cannot state the last search date, the eligibility criteria, and the outcome of interest in three sentences, pause and reconstruct them first. Searching before this step produces records you cannot interpret.

For the worked example, take a narrow question: in adults with a specified chronic condition, does a named class of intervention, compared with usual care, change a single defined outcome measured at twelve months?

Keeping the outcome singular matters. A brief that tries to update mortality, function, and quality of life simultaneously multiplies the screening and appraisal work and makes the "what changed" judgment ambiguous. Update one outcome rigorously, then repeat the process for the others if the brief requires them.

Write the question into a short protocol note that records:

• the population and any exclusions;
• the intervention and comparator;
• the single outcome and time horizon;
• study designs eligible for the update;
• the previous search date and sources;
• who will screen and who will arbitrate disagreements.

This note is the artifact that makes the rest of the update auditable. It also becomes the methods paragraph of the revised brief.

Cochrane's guidance is explicit that searches should be planned from eligibility criteria, draw on multiple sources, combine free-text terms with subject headings, and be documented. PubMed supports field tags, phrase searching, filters, Advanced Search, and saved searches, and its search details show how terms were translated and combined — a useful way to verify that a strategy did what you intended.

MeSH, NLM's controlled and hierarchical vocabulary, lets you expand or focus a strategy using preferred headings, entry terms, scope notes, qualifiers, and tree positions. Use it deliberately: a tree explosion broadens recall, while a qualifier narrows focus.

A workable sequence for the update search:

1. Draft free-text terms for each concept, including synonyms and spelling variants.
2. Map each concept to MeSH headings and decide whether to explode them.
3. Combine concepts with Boolean operators and apply field tags where precision matters.
4. Apply date limits set to the previous search date, not to an arbitrary recent window.
5. Run the strategy, then inspect the search details to confirm how terms were combined.
6. Save the strategy so the next update can reuse it.
7. Record the date, sources, full strategy, and result counts in the protocol note.

If you need to find academic papers through a scholarly search interface, WisPaper search is one place to run and collect candidate records; the public WisPaper Scholar Search page explains the entry point for finding academic papers. Whatever interface you use, the strategy itself — not the interface — is what you document.

Cochrane recommends deduplicating records and screening in stages, and it recommends linking multiple reports of one study. Both steps change the arithmetic of an update.

Deduplicate first. The same trial may appear as a conference abstract, a primary report, and a secondary analysis. Counting each as a separate study inflates the apparent weight of new evidence.

Screen in two stages. A title-and-abstract pass removes clearly ineligible records; a full-text pass confirms eligibility against the protocol. Record reasons for exclusion at full text so the revised brief can report them.

Link reports. Group every report belonging to one study before appraisal. A single study with three reports is still one study when you judge evidence level.

Track counts. Keep a simple flow: records identified, duplicates removed, screened, full texts assessed, studies included, reports linked. These counts belong in the updated brief.

PubMed's Similar Articles and citation lookup features can help you find related reports once you have a seed record, and links to full text appear when available. Use them to chase linked reports, not to replace the systematic strategy.

The most common error in updating a brief is collapsing several judgments into one impression. Keep them apart.

Evidence level. What design produced this result, and how does that design sit within the hierarchy the brief already uses? A randomized trial and an observational cohort answer the question with different strength, and the brief should say which it is relying on.

Applicability. Does the population, setting, and delivery of the intervention in the new study resemble the population and setting the brief addresses? Applicability is about fit to your context, not about study quality.

Indirectness. Does the study measure the outcome you specified, at the time horizon you specified, in the comparison you specified? A study that reports a related but different outcome is indirect for your question even if it is well conducted.

Limitations. What are the study's own weaknesses — its size, its conduct, its analysis? Limitations sit inside the study; the other three axes sit between the study and your question.

Record each axis in its own column. When a reviewer disagrees, the disagreement usually localizes to one axis, which makes arbitration faster and more transparent.

Once new studies are appraised, the update requires an explicit judgment. Three outcomes are possible, and each has a different consequence for the brief.

No change to the bottom line. New studies are consistent with the existing synthesis. The brief's conclusion stands, but the evidence base and search date must be updated, and the new studies cited.

Strengthened or weakened confidence. The direction holds, but the volume, consistency, or quality of evidence shifts. The brief's wording about confidence should change accordingly.

Changed bottom line. New evidence conflicts with the existing conclusion. This is the case that requires the most care: re-examine whether the new studies are direct, whether earlier studies were missed, and whether the change is driven by one study or by a pattern.

Write the judgment as a short paragraph with a stated reason. "The conclusion is unchanged because the two new trials are consistent in direction and comparable in population" is a defensible statement; "the new evidence looks similar" is not.

If the update involves a longer synthesis task — assembling a literature review around the refreshed question — the Scholar Agent area is designed for longer research tasks including literature review. Use it to organize the synthesis work, then apply your own appraisal judgments to what it surfaces.

An update generates more artifacts than the original brief: new PDFs, linked reports, screening decisions, and appraisal notes. Losing track of which record maps to which decision is the most common cause of a stalled update.

Save or upload the papers you are retaining and organize them in My Library so that the included studies for this update sit together and can be revisited at the next refresh. When you need to ask questions based on the papers you have uploaded or added from search results, Library QA supports that question-and-answer step against your own library.

If the brief is part of a larger ongoing program of work, the Research Projects area is where a project can be created to hold that work. Keeping the update inside a project makes the next refresh a continuation rather than a restart.

A revised brief carries new references, and every one of them must resolve to a real record. BibTeX entries assembled during drafting are a common failure point.

TrueCite is built for BibTeX citation screening: it checks BibTeX against real academic databases to flag hallucinated references. Run your reference file through that check before the brief goes to reviewers, and reconcile any flagged entry against the source record you screened.

This step belongs late in the workflow, after the included studies are fixed, so that you are verifying the references you will actually cite rather than a moving list.

Suppose the existing brief concluded that, in adults with a specified chronic condition, the named intervention class probably produces a modest improvement in the twelve-month outcome compared with usual care, based on a search completed two years ago.

The update proceeds as follows. The protocol note fixes the population, intervention, comparator, single outcome, and the two-year search window. The strategy combines free-text terms with MeSH headings, is limited to the previous search date, and is saved with its result count. Screening removes duplicates, a title-and-abstract pass excludes irrelevant records, and full-text assessment identifies, say, three new reports. Linking reveals that two of those reports describe the same trial, so the update contains two new studies, not three.

Appraisal separates the axes. Both new studies are randomized trials, so evidence level is comparable to the existing base. One enrolls a population and setting close to the brief's scope, so applicability is high; the other enrolls a narrower group in a different care setting, so applicability is lower. Both measure the specified outcome at twelve months, so indirectness is low. One has a small sample and wide uncertainty, which is a limitation of that study.

The judgment: the direction of effect is consistent with the existing conclusion, the evidence base is larger, and confidence is modestly strengthened — but the second study's narrower applicability means the brief should note that the finding may not extend to the broader population. The conclusion paragraph is revised to say exactly that, the search date is updated, the two new studies and their linked reports are cited, and the reference file is checked before circulation.

That example shows the shape of a defensible update: a fixed question, a documented search, linked reports, separated appraisal axes, an explicit judgment, and verified citations.

Use this table as the working record for each update. Each row is a checkpoint with a completion condition.

The brief is ready to circulate when every row is complete, the change statement is written, and the methods paragraph reflects the actual update rather than the original search.

Set these rules before screening begins so that decisions are not made under time pressure.

Stop and re-scope if the search returns a volume of records far beyond what the protocol anticipated — the question is probably too broad, and narrowing the outcome is the fix.

Stop and reconstruct the baseline if the previous search date or strategy cannot be recovered. An update without a baseline is a new review.

Escalate to a second reviewer when the two screening passes disagree on eligibility, or when appraisal judgments on applicability or indirectness conflict. Record the resolution.

Escalate before changing the bottom line when new evidence conflicts with the existing conclusion. A changed conclusion deserves a second appraisal pass and an explicit note in the brief about what drove the change.

Do not circulate while any reference remains unverified or any included study lacks a completed appraisal row.