How to Compare Treatment Studies with Different Follow-Up Times

Comparing treatment studies with different follow-up times is a matching problem before it is a statistics problem. Two trials can report the same outcome label yet answer different questions because one measures events over 12 weeks and another over 5 years. The practical answer is to build a comparison frame first: define the population, outcome, and time window; retrieve studies with a documented, reproducible search; extract the timing of every outcome; then judge whether the studies are similar enough to compare directly or must be grouped and described separately.

This guide is written for clinicians and medical researchers preparing a focused evidence brief, a research update, or a team discussion. It uses one running example throughout: adults with a specified chronic condition and a single named clinical outcome, such as a composite event or a validated symptom score. The aim is a defensible comparison that supports your team's discussion of the evidence.

Follow-up time is not a formatting detail. It determines what an outcome means. An event rate observed over a short window and one observed over a long window describe different periods of risk, and a symptom score measured at one time point is not the same construct as the same score measured later. Before you compare any two studies, write down the time window each one actually reports, not the window you wish it reported.

A useful discipline is to treat time as part of the outcome definition. "Composite event" is incomplete; "composite event within the reported follow-up period" is a definition you can compare. When studies differ in follow-up, you have three honest options: compare them directly only if the windows are close enough to be clinically equivalent for your question; group them by time window and describe each group; or report them side by side without a pooled comparison. Deciding which option applies is the core judgment of this task.

The Cochrane Handbook's guidance on searching begins with eligibility criteria: plan the search from what you will include, not from whatever the database returns. Apply that here by writing a one-page comparison frame before you open any search tool.

Your frame should state:

• Population: the specific adults, condition, and setting you are briefing on.
• Intervention and comparator: the treatment classes or specific agents under discussion.
• Outcome: one primary outcome, defined precisely, plus any secondary outcomes you will report separately.
• Acceptable follow-up windows: the range of follow-up times you will treat as comparable, and the windows you will report as separate strata.
• Study designs: which designs you will include and how you will label their evidence level.

Checkpoint: if you cannot state the acceptable follow-up range in one sentence, you are not ready to search. A frame that says "any follow-up" will produce a pile of studies you cannot compare.

The Cochrane Handbook recommends combining free-text terms with subject headings, using multiple sources, documenting the process, deduplicating records, and screening in stages. PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, and a Search Details view that shows how terms were translated and combined. MeSH provides a controlled, hierarchical vocabulary with preferred headings, entry terms, scope notes, qualifiers, and tree positions that you can use to expand or focus a strategy.

In practice, build the strategy in three layers:

1. Population terms as free text plus the relevant MeSH heading.
2. Intervention terms as free text plus headings, including synonyms and entry terms.
3. Outcome and design terms, kept deliberately broad so you do not exclude studies that report your outcome under a different label.

Record the date, database, full strategy, and number of records at each stage. That record is what lets a colleague reproduce your brief.

When you need to find academic papers through a scholarly search interface, WisPaper search is the natural entry point for this retrieval step. Keep your written strategy alongside whatever tool you use, so the search remains documented.

The Cochrane Handbook recommends staged screening and linking multiple reports of one study. This matters enormously for follow-up comparisons, because the same trial often appears in several papers: an early report at one time point and a later report at an extended follow-up. If you treat them as two studies, you will double-count and misread the evidence.

Screen in two passes:

• Title and abstract pass: remove clearly irrelevant records and obvious duplicates.
• Full-text pass: confirm population, outcome definition, and reported follow-up, and group multiple reports under a single study identifier.

Checkpoint: every included record should map to exactly one study identifier, and every study identifier should list all its reports with their respective follow-up times. If a study has two reports at two time points, record both; do not pick one silently.

Most extraction templates bury timing in a notes column. Promote it. For each study, capture the planned follow-up, the actual follow-up achieved, the time point at which your primary outcome is reported, and whether that time point was prespecified or added later. Also capture whether follow-up was complete or truncated, because a study that reports a long window with substantial early attrition is describing a different population than one with complete short follow-up.

A reusable extraction table keeps the comparison honest:

Checkpoint: if the "reported time point" column is blank for any study, you cannot yet place it in a comparison group.

These four judgments are often collapsed into a single "quality" score. Keep them apart, because they answer different questions.

• Evidence level describes the design and the strength of inference the design supports. A randomized trial and an observational cohort sit at different levels regardless of how similar their follow-up windows are.
• Applicability asks whether the study population, setting, and care context resemble the people you are briefing. A well-conducted trial in a different care setting may be high-level evidence with limited applicability to your audience.
• Indirectness asks whether the study's population, intervention, comparator, or outcome differs from your comparison frame. A study that measures your outcome at a much shorter window than your frame specifies is indirect for your question even if everything else matches.
• Limitations are the study-specific constraints the authors report or that your extraction reveals, such as incomplete follow-up or a non-prespecified time point.

A study can be high-level evidence, highly applicable, and still indirect because its follow-up window does not match your frame. Say so explicitly rather than averaging these judgments into one adjective.

With extraction complete, sort studies into three buckets and state your rule in the brief:

1. Directly comparable: follow-up windows fall within the range your frame defined as equivalent for this outcome.
2. Stratified: windows differ meaningfully, so you report results grouped by time window and describe each group separately.
3. Not comparable: the outcome construct or the time horizon is so different that side-by-side description is the most you can honestly offer.

Write the rule before you look at the results, so the grouping is not shaped by which studies favor a conclusion. If most studies land in the stratified bucket, that is a finding worth reporting, not a failure.

Suppose your frame covers adults with a specified chronic condition, a composite clinical event as the primary outcome, and an acceptable follow-up range of roughly one to two years. Your search returns five studies after linking reports.

• Study A reports the composite event at 12 months, randomized design, population closely matching your frame.
• Study B reports the same composite at 24 months, randomized design, similar population.
• Study C reports the composite at 6 months, randomized design, but the population is drawn from a different care setting.
• Study D reports the composite at 5 years, observational design.
• Study E is a second report of Study A at 36 months.

Applying the rule: Studies A and B fall inside the acceptable range and can be compared directly, with their different windows noted. Study E is linked to Study A, so it is not a sixth study; its 36-month data sit outside the frame and are reported as extended follow-up. Study C is inside a short window but outside the frame's range and has applicability concerns from its setting, so it is described separately. Study D is both outside the range and a different evidence level, so it is reported as contextual evidence, not pooled with the randomized studies.

The resulting brief has a direct comparison of A and B, a clearly labeled extended-follow-up note for E, a separate description of C with its applicability caveat, and a contextual paragraph for D. No study is discarded, and no time window is silently merged with another.

As the study set grows, keep the included papers and their linked reports in one place so the extraction table and the source PDFs stay together. My Library is where you can save or upload papers and organize them, which suits this consolidation step. If your team is coordinating the brief across several people, Research Projects provides the area for creating and managing a project around it.

Before circulating the brief, verify that every reference in your bibliography is real and correctly identified. TrueCite checks BibTeX against real academic databases to flag hallucinated references, which is the right check at this stage. A citation that cannot be matched to a real record should be removed or corrected before the brief leaves your hands.

The brief is ready when a reader can answer four questions from it without opening the underlying papers:

• What population and outcome does this comparison cover?
• Which studies were compared directly, and why were the others stratified or described separately?
• What follow-up window does each reported result actually reflect?
• What are the evidence level, applicability, indirectness, and limitations for each study?

If any of these requires the reader to guess, the brief is not finished. A short methods paragraph stating the search date, strategy, screening stages, and time-window rule completes the package.

Use these rules to decide when to pause rather than push forward:

• Stop and reframe if you cannot state an acceptable follow-up range for your primary outcome. Return to Step 1.
• Stop and re-screen if two included records turn out to be reports of the same study that you had counted separately.
• Escalate to a methodologist if the studies you want to compare use different outcome instruments or definitions that you cannot reconcile from the papers themselves.
• Escalate to a librarian or search specialist if your strategy returns too many irrelevant records to screen or too few to answer the question, and revising MeSH terms and free-text layers has not resolved it.
• Do not pool studies whose follow-up windows fall outside your frame; stratify or describe them instead.
• Do not circulate a brief with unmatched citations.