How to Separate Clinical Outcomes from Surrogate Outcomes

A surrogate outcome is a measurable substitute that stands in for a clinical outcome you actually care about. Separating the two is a labeling and reasoning task, not a vocabulary exercise: you decide what kind of outcome a study reports, state the causal assumption that lets a surrogate stand in for a clinical endpoint, then test whether the follow-up and the population make that substitution credible for your question. This article gives a repeatable sequence for clinicians and researchers answering a focused evidence question, with diagnostic checkpoints, a worked example, a reusable checklist, and stop rules.

Label the outcome first, before you judge study quality. Ask what the reported measure is and what the reader is meant to conclude from it.

• Clinical outcome: a health state the patient experiences directly, such as a symptom, function, or a clinical event.
• Surrogate outcome: a substitute measure used in place of that health state.

A single study can report both. Label each reported outcome separately rather than assigning one label to the whole paper. This prevents a strong surrogate result from silently borrowing the authority of a clinical claim.

Checkpoint 1: Write one sentence per reported outcome: "This study measures ___, which is [clinical / surrogate] for the question of ___ ." If you cannot finish the sentence, you have not yet labeled the outcome.

A surrogate only carries clinical meaning through an assumption: that change in the surrogate tracks change in the clinical outcome for this population. Make that assumption explicit and write it down.

Ask three questions:

1. What is the proposed causal path? State how the surrogate is supposed to connect to the clinical outcome.
2. Is the path asserted or examined? Note whether the study treats the link as established background or tests it.
3. Does the path hold for this population? A link demonstrated in one group may not transfer to another.

Checkpoint 2: If the causal path is only assumed, mark every downstream clinical claim as conditional on that assumption. Do not let the assumption disappear when you summarize the finding.

Surrogate and clinical outcomes often operate on different time horizons. Compare the study's follow-up with the horizon your question implies.

• Does the follow-up span the period over which the clinical outcome would occur?
• Does the surrogate change earlier than the clinical outcome would?
• Would a clinical claim require observation beyond the reported window?

Checkpoint 3: If the follow-up is shorter than the clinical horizon, restrict your conclusion to the observed window. Do not extend a surrogate finding into a longer-term clinical claim.

Patient relevance asks whether the outcome reflects something patients and clinicians would recognize as important. A measure can be precise and still sit far from what a patient experiences.

Use these prompts:

• Would a patient recognize this outcome as a change in how they feel or function?
• Is the outcome a step toward a clinical endpoint, or the endpoint itself?
• If the surrogate improved but the clinical outcome did not, what would that mean for your conclusion?

The last prompt is the most useful. It forces you to state what the surrogate is standing in for.

These four judgments are often collapsed into a single verdict. Keep them apart.

Indirectness is the judgment most tied to surrogate outcomes: an outcome can be indirect even when the study itself is well conducted. Record indirectness at the outcome level so it does not get absorbed into a general quality score.

Suppose your question is: in adults with a chronic condition, does an intervention reduce a hard clinical event over several years?

1. Label outcomes. The study reports a biomarker (surrogate) and a clinical event rate (clinical). Label them separately.
2. Trace the assumption. The implied path is that the biomarker change predicts the clinical event reduction. Write that path down.
3. Compare follow-up. If the biomarker is measured early and the clinical event window is longer, note the mismatch.
4. Test relevance. Ask whether patients would recognize the biomarker change as meaningful on its own.
5. Separate judgments. Record the design (evidence level), the population match (applicability), the biomarker-as-proxy issue (indirectness), and any study weaknesses (limitations) in separate fields.
6. Write the claim. State the clinical conclusion only to the extent the clinical outcome and follow-up support it; otherwise state the surrogate finding as a surrogate finding.

This example is deliberately generic so the same sequence can be reused across different focused evidence questions.

Your ability to separate outcomes depends on retrieving studies that report them clearly. Plan the search from your eligibility criteria, combine free-text terms with subject headings, and document the process. The Cochrane Handbook, Chapter 4, recommends planning searches from eligibility criteria, using multiple sources, combining free-text terms with subject headings, documenting the process, deduplicating records, screening in stages, linking multiple reports of one study, and rerunning searches before publication.

PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, Similar Articles, citation lookup, and links to full text when available; its search details show how terms were translated and combined. MeSH, NLM's controlled and hierarchical vocabulary, offers preferred headings, entry terms, scope notes, qualifiers, and tree positions that you can use to expand or focus a PubMed strategy.

Include outcome terms explicitly so surrogate and clinical reports are both captured. When you need to find academic papers through a scholarly search interface, WisPaper search is one entry point; for a public explanation of finding academic papers, see WisPaper Scholar Search.

Checkpoint 4: Before screening, confirm your strategy includes outcome terms and that you can explain how terms were combined.

Once records are retrieved, keep outcome labels attached to each study as you screen. Link multiple reports of one study so a surrogate report and a clinical report from the same study are not treated as independent findings.

Save or upload the papers you retain and organize them so outcome labels stay with each record using My Library. For questions based on papers you have uploaded to My Library or added from search results, Library QA lets you ask questions grounded in that set.

Checkpoint 5: For each retained study, confirm you have recorded the outcome type, the causal assumption, the follow-up window, and the population.

Use a claim ladder. Each rung requires the evidence of the rung below it.

1. Surrogate finding: "The surrogate changed." Supported by the surrogate result alone.
2. Assumed link: "The surrogate is proposed to track the clinical outcome." Supported only if you state the assumption.
3. Clinical claim: "The clinical outcome improved." Requires the clinical outcome to be reported.
4. Population claim: "This applies to my population." Requires applicability to be assessed.

Most errors happen when a rung-1 finding is written as a rung-3 claim. Keep the rung number in your notes and in your draft.

Use this checklist for each study you appraise:

• Each reported outcome is labeled clinical or surrogate.
• The causal assumption linking surrogate to clinical outcome is written down.
• The assumption is marked as asserted or examined.
• Follow-up is compared with the clinical horizon.
• Patient relevance of the outcome is stated.
• Evidence level, applicability, indirectness, and limitations are recorded separately.
• The claim ladder rung is recorded for each conclusion.
• Multiple reports of the same study are linked.
• The search strategy and screening steps are documented.

Stop and revise your claim when any of the following is true:

• No clinical outcome is reported. Do not write a clinical claim; report the surrogate finding as a surrogate finding.
• The causal assumption is unstated. Stop and state it before proceeding.
• Follow-up is shorter than the clinical horizon. Restrict the conclusion to the observed window.
• Applicability is unassessed. Do not extend the finding to your population.
• Indirectness is unresolved. Keep it visible in the outcome-level appraisal rather than folding it into a general quality judgment.

Escalate to a fuller review when a surrogate finding is being used to support a guideline or scholarly claim that depends on the clinical endpoint. At that point, the outcome label, the causal assumption, the follow-up comparison, and the applicability judgment all need to be documented before the claim is written.

For longer research tasks that fit a literature review workflow, Scholar Agent is available. If you are assembling a project around this question, Research Projects provides a place to create and organize it.