How to Read a Health Headline
Three questions that catch most misleading health headlines before you fall for them, walked through with real, current examples.
In-house · synthesized from the cited primary sources
At a glance
Strong A 2007 review found that across six treatments tested in both animals and humans, the animal result matched the human result in only three of six, a coin flip.
What the evidence says
Check whether the result came from people or animals
A 2007 BMJ review found that across six treatments tested in both animals and humans, the animal result matched the human result in only three of six. Worked example: a 2026 rat study on aspartame (70 rats, no PMID, not MEDLINE-indexed) found a 100 mg/kg/day dose is 2 to 2.5 times the FDA/EFSA ADI in raw mg/kg, but the FDA's own rat-to-human conversion (divide by 6.2) puts that same dose below both agencies' ADI. The two conventions differ roughly sixfold, so a trustworthy source states the full range and names which conversion it used, rather than picking a single number.
Why ↓
Only about half of animal results replicate in humans, and unit-conversion choices can flip a headline's framing entirely.
Perel et al., BMJ 2007; FDA CDER 2005 dose-conversion table
Check whether the study assigned the exposure or just observed it
Observational data linked beta carotene to less lung cancer. Two randomized trials in smokers then found the opposite: 18% more lung cancer (ATBC, 1994) and a rate ratio of 1.28 (CARET, 1996). In mostly non-smoking physicians, the same supplement did nothing (Hennekens, 1996, rate ratio 0.98). Population and study design both mattered.
Why ↓
An observational link and a randomized trial can point in opposite directions, and the beta carotene story is the cleanest documented case of it.
ATBC NEJM 1994; CARET NEJM 1996; Hennekens NEJM 1996
Check whether the dose used maps onto a real human exposure
The FDA's own conversion table divides a rat's mg/kg dose by 6.2, and a mouse's by 12.3, to estimate a human-equivalent dose. A headline dose that sounds alarming in raw animal mg/kg can look far more modest, or far more concerning, once converted, depending on which convention is used.
Why ↓
Species convert dose differently, and skipping that conversion is a common way a headline exaggerates or understates real-world relevance.
FDA CDER, 'Estimating the Maximum Safe Starting Dose,' July 2005
Check whether the causal claim came from the study or the press release
A 2014 analysis of 462 university press releases found that when a release turned a correlation into a cause, 81% of the resulting news stories did too; when it did not, only 18% did. A 2019 replication confirmed the causal-claim finding.
Why ↓
This is the single strongest fact on the page: most distortion happens in the press release, not the paper, and not because anyone lied.
Sumner et al., BMJ 2014;349:g7015
See the three questions applied to a live July 2026 headline
An 8-year observational study (12,772 people, diet measured once by questionnaire) found an association between artificial sweetener consumption and cognitive decline. The study has a published erratum, and the authors' own limitations section flags self-reported diet and residual confounding as issues. A ScienceDaily story dated 18 July 2026 covering it turned out to be a re-issue of a press release embargoed since September 2025, with no DOI, no author citation, and no mention of the erratum. This page reaches no conclusion about whether sweeteners are safe or harmful; that verdict is not what this kind of study can establish.
Why ↓
This is a live, current example of exactly the pattern the page teaches: the distortion happened between the paper and the reader, not in the paper itself.
Goncalves et al., Neurology 2025;105(7):e214023, erratum Neurology 2025;105(11):e214393
The evidence 6
Strong A 2007 review found that across six treatments tested in both animals and humans, the animal result matched the human result in only three of six, a coin flip.
Estimating the Maximum Safe Starting Dose in Initial Clinical Trials (FDA CDER, 2005)
Comparison of Treatment Effects Between Animal Experiments and Clinical Trials (Perel et al., BMJ, 2007)
The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC, NEJM, 1994)
Effects of a Combination of Beta Carotene and Vitamin A on Lung Cancer (CARET, NEJM, 1996)
Lack of Effect of Long-Term Supplementation With Beta Carotene (Hennekens et al., NEJM, 1996)
The Association Between Exaggeration in Health Related Science News and Press Releases (Sumner et al., BMJ, 2014)
Not medical advice. This page is for education only and is not a substitute for professional medical care. Consult a qualified clinician before changing your health routine.
Editorial disclosure. This protocol is written and fact-checked by the YourProtocol editorial team directly from the primary sources cited below; it is not written or reviewed by any outside expert.
Is this for you
- Anyone who wants a quick, repeatable way to check a viral health headline
- Anyone who has been confused by a study that seems to contradict an earlier one
- Anyone curious how a press release can distort a paper without anyone lying
Cautions
- This page reaches no conclusion about whether any specific substance, including artificial sweeteners, is safe or harmful; its point is only about what each study design can and cannot establish
- Worked examples are used to illustrate a method, not to render a verdict on any product or ingredient
- Educational only, not medical advice
Common questions
How do I know if a health study applies to humans?
Check whether the result came from people or animals. A 2007 BMJ review found that across six treatments tested in both, the animal result matched the human result in only three of six, a coin flip, not a guarantee.
What's the difference between a randomized trial and an observational study?
An observational study watches what happens to people who already made a choice; a randomized trial assigns the choice and compares groups. They can disagree: observational data linked beta carotene to less lung cancer, then randomized trials in smokers found 18 to 28% more lung cancer on beta carotene.
Why do animal-study doses sound so much higher than what I'd take?
Species convert dose differently. The FDA's own guidance divides a rat's mg/kg dose by 6.2, and a mouse's by 12.3, to estimate a human-equivalent dose, so a headline dose that sounds alarming in raw mg/kg can be far more modest once converted.
Why does a headline sound more certain than the actual study?
A 2014 BMJ analysis of 462 university press releases found that when a release turned a correlation into a cause, 81% of the resulting news stories did too; when it didn't, only 18% did. The distortion usually happens in the press release, not the paper.
Does this page conclude whether sweeteners are safe or harmful?
No, deliberately. This page is about how to read a study, not about rendering a verdict on any specific substance. The worked example on sweeteners and brain aging is used only to illustrate the method.