Write the exact claim before judging it
A claim becomes easier to judge when you can say exactly what was measured, in whom, and what someone proposes to do with the result. Start with that sentence. It gives the conversation a shared target.
Before you start: The three validity questions separate correct measurement, medical meaning and useful action. Prognostic versus predictive biomarkers separates risk information from evidence about a treatment's effect.
Where this step sits
This is the first step in judging a test or company claim. You will define the claim, inspect its validation, then compare the decisions it could change.
Turn the brochure into an answerable sentence
Suppose a brochure says, “Our personalized platform identifies the best treatment.” That sounds helpful. Yet it leaves the reader guessing what “best” means. The platform might measure a molecule, rank drugs in a dish, or estimate outcomes from past patients. Those are different claims.
Try this sentence:
In this population and treatment setting, using this specimen and method, the result measures or predicts this outcome, and is proposed to support this decision.
You can leave blanks. A blank is a good question for the next conversation, rather than a reason to fill the gap with an assumption.
For a risk model, name the time horizon and the care context. Risk under usual treatment is different from risk without treatment. For a drug-ranking service, name the actual treatment, dose or combination being proposed. A drug's name alone cannot tell you whether the tested conditions resemble patient care.
Separate the claims inside the claim
A report often carries a chain of reasoning:
| Part of the report | Question to ask |
|---|---|
| Measurement | What material was tested, and did it meet quality requirements? |
| Biological interpretation | What mechanism does the finding support, and what alternatives remain? |
| Prediction | Which clinical outcome, population and time period does the model address? |
| Proposed action | What treatment or testing decision would change? |
| Patient benefit | What evidence shows that this action helps compared with the alternative? |
Mark each part as measured, inferred, proposed, or unknown. An observation can be sound while later steps remain hypotheses. That preserves the value of the measurement without silently extending it.
The status label deserves its own line. Expand research use only (RUO) when that label appears. Identify the laboratory that produced the result and the exact test service. A company's clinical offering and research offering may have different intended uses.
Worked example: a fictional “Drug A responder” report
A fictional laboratory measures a protein in a stored tumor sample. Its report says, “Drug A responder.” The underlying study related high protein levels to tumor shrinkage among patients who all received Drug A for advanced Cancer X.
A caregiver is asking whether this predicts benefit after surgery for early Cancer X. The measurement, response association and proposed postoperative use should receive separate lines in the claim worksheet.
The study gives a reason to investigate the marker. Because everyone received Drug A, it may not establish whether the marker identifies greater benefit from Drug A compared with an alternative. The postoperative setting also differs from the studied setting. Neither gap is closed by the phrase “personalized.”
A helpful question is: “What comparison shows that choosing Drug A with this result improves outcomes in the proposed setting?” Ask for the study and its limitations. That invites a concrete answer.
What can go wrong at this step
- A broad title hides a narrow result. “Predicts benefit” may mean an association with response among treated patients. Write the outcome and comparison.
- A gene or target name becomes a drug recommendation. Actionability requires a more complete evidence chain.
- A research measurement is mistaken for a clinical report. Ask which service, version and intended use produced it.
- A negative test becomes “nothing to worry about.” Detection limits, timing and the base rate affect interpretation.
Try it
A fictional brochure says, “Our blood test detects recurrence earlier, so patients live longer.” What are the two claims?
Answer: First, the test may detect a signal before another method. Second, a test-guided care strategy may improve survival. Earlier detection alone does not establish the second claim; lead-time bias can make survival measured from detection look longer without postponing death.
Explain it back
Complete: “The report measures ____. The proposed decision is ____. The missing evidence is ____.”
One possible answer: “It measures a protein. The proposed decision is adding Drug A after surgery. The missing evidence is whether that strategy improves outcomes in this setting.”
Takeaway
Write the measurement, interpretation and proposed action as separate claims before you decide how persuasive the report is.
Next: Read the validation.
Sources and scope
Source check: October 9, 2026. All reports and drugs in the exercises are fictional. Expert and learner review remain pending.
- MedlinePlus Genetics: valid and useful testing — separate validity and utility questions.
- FDA–NIH BEST: predictive biomarker — treatment-effect claims require the appropriate comparison.
- FDA: RUO and investigational-use diagnostic guidance — intended-use labeling.