Route stop 1
The Detail: Turn the Headline Into a Checkable Sentence
Illustrative data, not patient results.A headline can compress several claims into a few words. Copy the exact wording, publication date, publisher, and link before it changes. Circle the verb. Lowered, prevented, caused, was linked with, and may help do not carry the same level of certainty. Identify the outcome: A1C, a measured glucose statistic, self-reported behavior, diagnosis, or something else. Identify the population and time period if they are stated.
Rewrite the claim as five blanks: Among whom? What exposure or intervention? Compared with what? Which outcome? Over how long? A sixth blank asks how large the difference was. If the headline does not answer a blank, mark it missing rather than borrowing an answer from the image or caption. This produces a neutral search target.
For this route, the fictional headline is Weekly messages cut high glucose by 40%. It names neither the study design nor the outcome definition. It could refer to a relative difference in one event count, an A1C change, time above a threshold, or a subgroup analysis. The number cannot be checked until its denominator and outcome are found. Illustrative claim, not evidence or medical advice.
Route stop 2
Trace the Link to the Primary Study
A news story, press release, advertisement, or social post is a secondary layer. Follow its citation to the journal article, government report, conference abstract, or trial record that contains the methods and results. Confirm the title, authors, publication, date, and digital object identifier when available. A link to an organization's home page or an unrelated paper does not support the specific claim.
Read more than the abstract. Abstracts are useful maps, but they compress eligibility rules, missing data, analysis changes, adverse events, and limitations. Locate the methods, results tables, figure legends, disclosures, and supplementary material. For a clinical trial, compare the publication with the registration and prespecified outcomes when available. For an observational analysis, find how exposures, confounders, and outcomes were defined.
If the primary source is inaccessible, state that limitation. Do not treat a reporter's summary as if you inspected the methods. Search the exact study title in authoritative indexes, check the publisher's page, and look for an official public record. A transparent conclusion can be This claim cannot yet be checked from the available material.
Route stop 3
Map Participants, Design, and Comparison
Start with who entered the study, who completed it, and who was analyzed. Record age range, relevant diagnoses or eligibility rules, setting, geographic area, sample size, recruitment method, and important exclusions. A study in a narrow specialist-clinic population may answer a valid question without representing every adult who reads the headline. Generalizability is a separate judgment from whether the study was conducted correctly.
Next name the design. In a randomized controlled trial, chance assignment can help balance known and unknown differences between groups when the trial is well designed and conducted. In an observational study, researchers observe existing exposures or choices; adjustment can address measured factors, but unmeasured confounding may remain. A cross-sectional snapshot, case series, laboratory experiment, animal study, and systematic review each support different inferences.
Describe the comparison in plain language. Usual care, placebo, another intervention, no exposure, and a person's own earlier period are not equivalent controls. Check whether groups received different attention, devices, education, or follow-up in addition to the named intervention.
Route stop 4
Define the Outcome Before Reading the Effect
Find the primary outcome, which is the result the study was principally designed to assess, and distinguish it from secondary, exploratory, and post hoc outcomes. Record how it was measured, at which time point, and whether the definition changed. A1C, average device glucose, time in a defined range, episodes above a threshold, medication use, and quality-of-life scores are different outcomes. One cannot silently replace another in a headline.
Ask whether the measure was laboratory based, device generated, self-reported, or calculated. Check the observation window and data-completeness rules. If a device outcome excludes periods with missing data, inspect whether missingness differed between groups. If a result is based on a subgroup, determine whether that subgroup and analysis were planned before researchers saw the data.
Timing changes meaning. A result measured at four weeks does not demonstrate persistence for a year. A study designed to test a short-term marker may not establish effects on symptoms, complications, safety, or everyday functioning. Keep the stated outcome and duration attached to every summary sentence.
Route stop 5
A Worked Example
Illustrative data, not patient results.Educational example. The fictional Weekly messages cut high glucose by 40% headline traces to a twelve-week observational report. Eighty adults chose whether to enroll in a messaging program. The article defines the outcome as the proportion of recorded days with at least one device reading above a stated study threshold. It reports 15% of days in the participant group and 25% in the comparison group. The relative difference is 40%, while the absolute difference is 10 percentage points. Illustrative data, not patient results.
The study did not randomly assign enrollment. Participants who chose messages also uploaded more complete device data and attended more education sessions. Those differences could influence the outcome. The headline's word cut implies causation, but the design shows an association after the reported adjustments. The analysis lasted twelve weeks and did not report A1C as its primary outcome. A faithful summary would describe the population, observed association, outcome definition, absolute figures, duration, and major limitations.
This example does not show that the program works or fails. It shows why a percentage needs its baseline and why a design needs an appropriately cautious verb.
| Headline element | What the fictional paper reports | Reader's revision |
|---|---|---|
| Cut | Nonrandomized association | Was associated with |
| High glucose | Days with at least one reading above a study threshold | Use the paper's exact outcome |
| 40% | 25% versus 15% of days | Add 10 percentage-point absolute difference |
| Implied broad audience | 80 self-selecting adults at one setting | Name the studied group |
| Implied lasting result | Twelve-week analysis | Keep the duration visible |
Route stop 6
Read the Effect Size, Denominator, and Uncertainty
Statistical significance does not tell you whether an effect is large, useful, durable, or applicable to a particular person. Find the actual group values, the absolute difference, the relative difference when reported, and the confidence interval or other uncertainty measure. A wide interval can include effects that would lead to very different practical conclusions. Check whether the authors tested many outcomes or subgroups, which increases opportunities to find a chance result unless the analysis accounts for multiplicity.
For percentages, distinguish percentage points from percent change. A move from 25% to 15% is a 10 percentage-point difference and a 40% relative decrease from the 25% starting value. Reporting only 40% makes the effect sound detached from its baseline. For continuous outcomes, inspect the measurement scale, units, starting values, and variation, not only a p-value.
Then ask whether the reported difference is clinically meaningful. That judgment requires subject knowledge and context; it cannot be inferred from statistical significance alone. The FTC's health-claims guidance emphasizes both reliable study methods and results meaningful to consumers when evaluating substantiation.
Route stop 7
Read Limitations, Funding, and the Surrounding Evidence
Read the authors' limitations, then look for issues not emphasized in the headline: attrition, missing data, protocol changes, measurement error, short follow-up, unblinded assessment, baseline imbalance, selective reporting, and analyses added after results were known. No single checklist score settles quality. The importance of each issue depends on the question and design.
Record who funded the work, who designed and analyzed it, whether authors report financial relationships, and whether data or code are available. Funding does not automatically invalidate a study, and an absence of commercial funding does not guarantee quality. Disclosure helps readers examine incentives and methods rather than dismissing or accepting evidence by label.
Finally, place the result within the total body of relevant evidence. Replication by independent teams, consistency across strong designs, and explanations for conflicting results matter. Publication alone is not proof, and a single favorable trial does not establish consensus. The FTC warns that the quality of evidence matters more than simply counting studies and that contrary evidence cannot be ignored.
Route stop 8
What It Does Not Tell You
A careful paper review does not determine what a reader should eat, which medicine or dose to use, how often to measure glucose, or what personal target to pursue. It does not turn a group average into an individual forecast. Even a well-conducted trial may answer a narrow question for a defined population and outcome. Safety, feasibility, cost, adherence, and long-term effects may require different evidence.
This route is a literacy method, not peer review or medical advice. When a claim could change care, bring the source and your questions to a qualified healthcare professional. When a commercial message makes an objective health claim, look for evidence that matches the advertised product, population, outcome, and wording. Do not let a disclaimer cancel a bold promise that the research itself does not support.
- Quote the exact claim and circle its verb.
- Find and read the primary source beyond the abstract.
- Map participants, design, comparison, outcome, and duration.
- Report absolute values alongside relative changes.
- Keep uncertainty, limitations, funding, and contrary evidence visible.
- Stop at the boundary between evidence description and personal care advice.