A Guideline Recommendation Does Not Match the Studies You Found

You are preparing a short evidence brief for a team discussion. The guideline says one thing; the studies you pulled say something narrower, older, or aimed at a different population. The instinct is to force a match — either by trusting the guideline and dropping the studies, or by trusting the studies and quietly setting the guideline aside. Neither produces a defensible brief.

The practical resolution is to treat the mismatch as a finding, not a failure. Your job is to show why the recommendation and the retrieved studies diverge, then deliver a brief that states what the evidence supports for your specific population and outcome. This article walks through that repair using a focused example, then shows where search, screening, and library tools fit into the workflow.

A guideline recommendation and a set of retrieved studies answer related but not identical questions. The guideline synthesizes across populations, comparators, and time horizons, and it carries a strength rating that reflects more than the raw study results. The studies you found are a sample — shaped by your search terms, your filters, and the databases you used.

So a mismatch can come from several places at once:

• Evidence level. The studies may be lower-level designs than the evidence underpinning the recommendation.
• Applicability. The studies may enroll a population, setting, or baseline risk that differs from the one your team cares about.
• Indirectness. The outcome measured in the studies may not be the outcome the recommendation addresses.
• Limitations. The studies may be small, short, or at risk of bias in ways the guideline already accounted for.

Naming which of these is operating is the core analytical move. It converts a confusing contradiction into a structured statement your colleagues can evaluate.

Before touching the guideline again, audit how you got your study set. Cochrane guidance on searching and selecting studies is a useful checklist here: plan searches from explicit eligibility criteria, use multiple sources, combine free-text terms with subject headings, document the process, deduplicate, and screen in stages. If your current set came from a single quick query, you have a search problem before you have an evidence problem.

Ask yourself:

1. Did I define eligibility before searching? Population, outcome, study design, date range.
2. Did I use more than one source? A single database narrows what you can see.
3. Did I combine free-text and controlled vocabulary? Free-text alone misses indexed synonyms; headings alone miss unindexed terms.
4. Did I document the strategy? Without a record, you cannot tell whether the mismatch is real or an artifact of your query.
5. Did I link multiple reports of the same study? Duplicate reports can inflate your sense of how much evidence exists.

If several answers are "no," rebuild the search before interpreting the guideline. If the answers are "yes" and the mismatch persists, you have a genuine evidence-appraisal task.

A focused example keeps this concrete. Suppose your team is discussing a guideline recommendation about a preventive intervention in older adults, and the outcome of interest is a functional endpoint rather than a laboratory or surrogate measure. Your retrieved studies report the surrogate outcome, enroll a younger or lower-risk group, or run for a shorter period than the recommendation's evidence base.

Write the question in four parts: population, intervention, comparator, outcome. Then write the guideline's implicit question the same way. The gap between the two statements is your analysis.

You are not declaring the guideline wrong or the studies irrelevant. You are showing that they answer different questions, and stating which question your brief needs to answer.

A frequent cause of apparent mismatch is a search that never reached the studies the guideline relied on. The NLM Medical Subject Headings vocabulary is hierarchical and includes preferred headings, entry terms, scope notes, qualifiers, and tree positions. Those features let you deliberately expand or focus a strategy rather than guessing at synonyms.

Practical moves:

• Look up the preferred heading for your population and outcome, then check its entry terms for synonyms you missed.
• Use tree position to broaden to a parent heading or narrow to a child heading.
• Apply qualifiers to separate therapy, diagnosis, or prognosis literature.
• Read the scope note to confirm the heading means what you think it means.

PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, Similar Articles, and citation lookup, and its search details show how terms were translated and combined. That translation view is the fastest way to see whether your query actually captured the concept you intended. If your strategy was built on free text only, rebuild it with headings and rerun before concluding the guideline and the literature disagree.

When you want to run that rebuilt strategy in a scholarly search interface, WisPaper search is the place to find academic papers, and the WisPaper Scholar Search page explains the entry point for finding papers. Keep your eligibility criteria beside you as you search so the new results stay tied to the question you defined.

Cochrane's selection guidance recommends staged screening and linking multiple reports of one study. In practice this means:

1. Deduplicate before screening, so the same record does not consume attention twice.
2. Screen titles and abstracts against your eligibility criteria, then screen full texts.
3. Group reports by study, so a trial reported in three papers counts as one study, not three.
4. Record exclusion reasons, so your brief can explain what the retrieved set does and does not contain.

This stage is where many apparent mismatches dissolve. A study that looked like it contradicted the guideline may turn out to be a secondary report of a trial already represented, or a subgroup analysis that does not stand alone.

As your screened set stabilizes, save the papers you are keeping so the brief has a stable evidence base. My Library lets you save or upload papers and organize them in the WisPaper library — useful when the same set needs to serve the brief, the team discussion, and a later update.

These four judgments are often collapsed into a single verdict. Keep them apart in your brief, because each points to a different conclusion.

Evidence level. What design produced the result? A synthesized body of evidence and a single small trial carry different weight, and the guideline's strength rating reflects that synthesis.

Applicability. Does the study population resemble the one your team is discussing? Differences in age, baseline risk, setting, or comorbidity change what a result means for your group.

Indirectness. Does the study measure the outcome the recommendation addresses? A surrogate or short-term measure is indirect relative to a functional endpoint.

Limitations. What constrains the individual studies — size, duration, or risk of bias? The guideline may already have weighed these; your brief should say so rather than re-litigating them.

A useful sentence pattern for the brief: "The recommendation addresses [outcome] in [population]; the retrieved studies report [outcome] in [population] over [horizon], so they inform but do not directly test the recommendation's question." That sentence is narrower than either source and is exactly what a team discussion needs.

Once the papers are organized, you can interrogate the set directly instead of re-reading each one. Library QA supports asking questions based on papers uploaded to My Library or added from search results. Use it to check consistency across your retained studies — for example, whether the enrolled populations and measured outcomes line up with the question you wrote.

Use the answers as a way to locate passages and compare studies, then verify each point against the paper itself before it enters the brief. The goal is a faster route to the specific sentences that establish population, outcome, and limitation, so you can read them directly.

Your finished brief should make the mismatch legible in a fixed structure. A workable outline:

1. Question. Population, intervention, comparator, outcome, and time horizon.
2. What the guideline recommends. State the recommendation and the question it implies.
3. What the retrieved studies show. Summarize the screened set, grouped by study.
4. Why they diverge. One short paragraph each for evidence level, applicability, indirectness, and limitations — only where relevant.
5. What the evidence supports for your population and outcome. A narrow statement, not a treatment recommendation.
6. What would change the picture. Which study design, population, or outcome would close the gap.
7. Search documentation. Sources, terms, headings, filters, dates, and screening decisions.

That last section is what makes the brief reusable. Cochrane recommends rerunning searches before publication; documenting the strategy now means the next update starts from a record rather than from memory.

If the brief will feed a larger review or a longer synthesis, Scholar Agent is designed for longer research tasks in WisPaper's literature review, inspiration, or reproduction use cases. Keep the focused question and the screened set as the inputs so the longer task stays anchored to the mismatch you diagnosed.

A brief that misattributes a reference undermines the analysis around it. Before circulating, verify that every citation in your draft points to a real record. TrueCite screens BibTeX citations by checking them against real academic databases to flag hallucinated references. Run it on the bibliography, then confirm that each surviving reference is one you actually screened and linked to a study.

This step matters most when the brief was assembled quickly or when references were carried over from an earlier draft. A flagged citation is a prompt to return to the source, not a verdict on the analysis.

The result is ready when a colleague can read it and answer four questions without asking you:

• What question does the brief answer? Population, outcome, and horizon are explicit.
• What does the guideline say, and what question does it imply? Stated in the guideline's own terms.
• What do the retrieved studies actually show? Grouped by study, with the screened set documented.
• Why do they diverge? Evidence level, applicability, indirectness, and limitations are separated, and only the relevant ones are discussed.

If any of those is missing, the brief still reads as a contradiction rather than an analysis. If all four are present, the mismatch has become the brief's central finding — which is a stronger deliverable than a forced agreement.

• Diagnose the search before interpreting the evidence. Many mismatches are query artifacts; controlled vocabulary and documented strategies test that first.
• Write both questions. The guideline's implied question and yours, side by side, make the gap visible.
• Keep the four judgments separate. Evidence level, applicability, indirectness, and limitations lead to different conclusions and should not be merged.
• Group reports by study. Duplicate reports distort how much evidence appears to exist.
• Document as you go. Search strategy, screening decisions, and citation checks make the brief updatable.
• State narrower claims than your sources. A brief that says what the evidence supports for your population is more useful than one that declares a winner.