Key takeaways
- A large trial found vitamin D cut autoimmune disease by 22%, which by our math works out to about 2.5 fewer cases per 1,000 people over five years.
- A Cochrane review of 35 studies found that relative numbers make benefits look bigger and more persuasive, for doctors as much as for patients.
- The number needed to treat tells you how many people must take something for one to benefit; for statins and heart attacks in one review, it was 118.
- A result can be statistically significant and still too small to feel: paracetamol beat placebo for arthritis pain by about 3 points out of 100.
- Ask any seller or clinician for the risk with and without treatment, the time frame and whether the change is big enough for patients to notice.
In January 2022, researchers at Brigham and Women’s Hospital in Boston published a result that sounded like a clear win for supplements. In a trial of 25,871 older adults, those who took 2,000 IU of vitamin D a day were 22% less likely to develop a confirmed autoimmune disease, such as rheumatoid arthritis or psoriasis1.
The trial, called VITAL, followed people for about five years1. Here are the counts behind the percentage. Among 12,944 people on placebo, 155 developed a confirmed autoimmune disease. Among 12,927 people on vitamin D, 123 did1.
By our own arithmetic, that’s about 12 cases per 1,000 people without vitamin D and about 9.5 per 1,000 with it. The difference is about 2.5 cases per 1,000 people over five years. The paper reported the relative figure and hazard ratios, not this absolute difference1.
Is that a meaningful benefit? The authors argued that it is, because the supplements are well tolerated and nontoxic, and because nothing else is proven to prevent these diseases1.
You may well agree. But you can only decide once you see the absolute numbers next to the relative ones. Here is how to find them, read them and judge whether a benefit is worth having, whether it comes from a pharmacy or a health food store.
The same result can sound huge or tiny
Relative risk compares two groups as a ratio: “22% less likely” or “half the risk.” Absolute risk counts how many people out of 100 or 1,000 had the outcome.
Picture two made-up treatments, each advertised as cutting a risk in half. The first lowers the risk from 2 in 1,000 to 1 in 1,000. The second lowers it from 40 in 100 to 20 in 100.
Both are 50% relative reductions. But the first helps 1 person in every 1,000 who take it. The second helps 20 in every 100. Same headline, wildly different value.
Does the format change what people think? A 2011 Cochrane review led by Elie Akl pooled 35 studies with 83 comparisons of ways to present health statistics2. Relative risk reductions made benefits seem larger and were more persuasive than absolute numbers2.
Health professionals were swayed just as much as members of the public2. None of the studies tracked real-world behavior, and the reviewers rated the evidence as moderate in quality2. Their warning is the practical point: a relative number alone can mislead, and it makes benefits and harms hard to compare fairly2.
Number needed to treat turns a percentage into people
The number needed to treat, or NNT, answers a plain question. How many people must take a treatment, for a set time, for one of them to avoid the bad outcome?
You get it by dividing 1 by the absolute risk difference. If a treatment lowers risk by 1 percentage point, the NNT is 100. If it lowers risk by 10 points, the NNT is 10.
Statins, the cholesterol-lowering drugs, offer a well-documented example. In 2022, reviewers at Oregon Health & Science University pooled trials for the U.S. Preventive Services Task Force, an independent panel that advises on prevention3. The trials enrolled adults at higher-than-average risk who had never had a heart attack or stroke, and most lasted one to six years3.
Statins lowered heart attacks by about a third in relative terms3. In absolute terms, that was 0.85 percentage points, or about 8 or 9 fewer heart attacks per 1,000 people. The reviewers put the NNT at 1183.
For death from any cause, the absolute drop was 0.35 percentage points, an NNT of 2863. For a combined count of heart attacks, strokes and related events, the NNT was 783.
Two facts make an NNT useful. It always comes with a time frame and a specific outcome. And it depends on your starting risk.
The same review found that statins lowered risk by a similar share across groups, but people at higher starting risk gained more in absolute terms3. A person with a high risk of heart attack gets more from the same pill than a person with a low risk. That’s why your own risk estimate matters more than any headline.
Treatments also have a flip side, sometimes called the number needed to harm. Weighing one against the other is the subject of our companion article on benefits and risks.
Run the numbers yourself: a worked example
The table puts four claims through the same math. The first two rows are made up to show the arithmetic. The last two use published numbers, with our calculations labeled.
| Claim you might hear | Risk without | Risk with | Absolute difference | Number needed to treat | Over how long |
|---|---|---|---|---|---|
| “Cuts your risk in half” (made-up example A) | 2 in 1,000 | 1 in 1,000 | 1 in 1,000 | 1,000 | Not stated |
| “Cuts your risk in half” (made-up example B) | 40 in 100 | 20 in 100 | 20 in 100 | 5 | Not stated |
| Vitamin D and autoimmune disease, VITAL trial | About 12 in 1,000 (our math) | About 9.5 in 1,000 (our math) | About 2.5 in 1,000 (our math) | About 400 (our math) | About 5 years1 |
| Statins and heart attack, Task Force review | Varies by trial | Varies by trial | 8.5 in 1,000 | 118 | 2 to 6 years3 |
A significant result is not the same as an important one
Researchers call a result statistically significant when it would be unlikely if the treatment did nothing at all. The usual cutoff is a p-value below 0.05.
That speaks to whether an effect is likely there, not to how big it is. In 2016, the American Statistical Association put it bluntly in a formal statement: a p-value does not measure how big an effect is or how much it matters4. It also warned against making decisions based only on whether a result crosses that line4.
VITAL shows why. Its vitamin D result had a p-value of 0.05, right at the edge1. The confidence interval, the range of results the data fit, ran from a hazard ratio of 0.61 to 0.991. In plain terms, the true effect could be a large drop or almost no drop at all.
There is a hopeful detail, too. When the researchers counted only cases diagnosed after the first two years, the relative drop grew: 54 cases on vitamin D versus 87 on placebo, a hazard ratio of 0.611. That hints that the benefit may build over time, though only further trials can confirm it. By our count, the raw gap in cases barely changed: 33 after year two, versus 32 for the whole trial.
The Federal Trade Commission (FTC), which polices health advertising, draws the same line. Its December 2022 guidance for health products says a statistically significant result is not enough on its own; it has to add up to a benefit that matters clinically for shoppers5. In the FTC’s view, a significant change can still be too trivial to affect anyone’s health5.
The smallest change you can feel has a name
Researchers call it the minimal clinically important difference, or MCID. It is the smallest improvement that patients notice and value.
In 2001, a team led by John Farrar analyzed 10 trials of the pain drug pregabalin, with 2,724 people living with chronic pain6. The conditions ranged from back pain to fibromyalgia and osteoarthritis. The team compared changes on a 0-to-10 pain scale with how much better people said they felt6.
On average, a drop of about 2 points, or about 30%, matched people saying they were much or very much improved6. People who started with more pain needed a bigger drop in points to feel the same relief6. The pattern held across studies and diseases6.
That yardstick can humble a familiar medicine. A 2019 Cochrane review of 10 trials with 3,541 people tested paracetamol, called acetaminophen in the United States, for knee and hip osteoarthritis7. On a 0-to-100 pain scale, it beat placebo by about 3 points7.
The reviewers treated 9% of the scale, or about 9 points, as the smallest change that matters. So paracetamol’s benefit was real but too small to matter, and they rated that conclusion high certainty7.
The same yardstick applies to lifestyle remedies. A 2024 Cochrane review pooled 139 trials with 12,468 people on exercise for knee osteoarthritis8. Compared with no treatment or usual care, exercise eased pain by about 13 points out of 100. Against a placebo-style attention program, the gain was about 9 points8.
This review set its threshold at 12 points and called the benefits of uncertain clinical importance8. Many trials were small, and people knew which group they were in, which can inflate results8. Physical function improved by about 11 to 13 points, with moderate certainty8.
Notice that the two reviews used different thresholds, 9 points and 12. There is no single magic number, so it’s fair to ask which one a study used and why.
Averages also hide people. An average gain of 3 points could mean most people gained nothing while a few gained a lot. That’s why good trials also report responders, the share of people who reached a meaningful improvement. Farrar’s team proposed their 30% benchmark as a standard way to judge future pain trials6.
What drug and supplement ads are supposed to show you
Prescription drug ads aimed at consumers have their own rulebook on numbers. In December 2023, the Food and Drug Administration (FDA) finalized guidance on how these ads should present results9.
It says relative figures are harder for consumers to grasp and can make a drug look better than it is. Ads that use them should give the matching absolute numbers prominently, right alongside9. Ads should also show results for both the drug group and the comparison group, in the same format with the same denominator9. Graphs should show the whole picture, not a squeezed axis that exaggerates a gap9.
The guidance is nonbinding: it states the agency’s current thinking rather than a legal requirement9.
In September 2025, the FDA announced a crackdown on misleading drug ads. It said it was sending thousands of letters, including about 100 cease-and-desist letters, and would start rulemaking to close a 1997 “adequate provision” loophole10. The agency said companies had used that loophole to leave critical safety information out of broadcast ads10. It also cited a 2024 review finding that every drug company social media post studied highlighted benefits, while only a third mentioned harms10.
Supplement ads fall mainly under the FTC. Its guidance calls randomized controlled trials, in which people are assigned by chance to a treatment or a comparison group, the most reliable evidence for health claims5.
One of its hypothetical examples fits this article neatly. A careful 12-week study found no difference in body fat, even though two shorter six-week studies had found a small but significant loss. Weighing all three, the FTC said a promise of major fat loss was not backed up5.
Supplement labels usually carry phrases like “supports joint health” instead of numbers. If a seller does give you a number, use the four-step check above.
Honest numbers serve natural-health values
Many people drawn to natural health think in years, not days. They take a supplement or follow a routine for a long time, hoping to prevent something rather than fix it.
That makes the time frame part of the decision. An NNT of 400 over five years means a lot of people taking a daily pill, and paying for it, so that one avoids an illness.
A small absolute benefit can be worth it when a remedy is safe, cheap and easy. The VITAL authors made a version of that case, noting that the supplements are well tolerated and nontoxic1. It may not be worth it when the remedy is costly, burdensome or risky. That trade-off is a values call, and it belongs to you.
The key is to use one yardstick for everything. Ask the same questions of a statin, a vitamin, a painkiller and an exercise plan. Prevention-minded people deserve real numbers, not just big percentages.
What the evidence shows
| Claim | Evidence rating | What it rests on |
|---|---|---|
| Relative numbers alone make benefits seem bigger and more persuasive | Consistent findings across 35 studies, rated moderate quality2 | |
| Statins lower heart attacks in higher-risk adults without heart disease | Pooled randomized trials for the Task Force3 | |
| Daily vitamin D (2,000 IU) lowers autoimmune disease in older adults | One large trial with a borderline result1 | |
| Paracetamol gives meaningful pain relief for knee and hip osteoarthritis | Ten trials; benefit about 3 points out of 100, below the meaningful threshold7 | |
| Exercise eases knee osteoarthritis pain by an amount patients clearly notice | 139 trials, many small and unblinded; low certainty for pain, and average gains sit near the threshold8 | |
| A drop of about 2 points or 30% on a 0-to-10 scale is a meaningful pain improvement | Consistent across 10 trials and several conditions6 |
Questions readers ask
Is a small absolute benefit worthless?
No. A small benefit can be worth having when the harm, cost and effort are small too, and when the outcome is serious. The point of absolute numbers is to let you weigh the trade-off fairly, not to dismiss every modest gain.
Why do prevention treatments have such large NNTs?
Because most people in a prevention trial would not have had the illness during the study anyway. When the starting risk is low, even a good treatment helps only a few people out of each 1,000. Higher-risk people usually gain more in absolute terms, as the statin review found3.
Where can I find the absolute numbers?
Look in a study’s results table for how many people in each group had the outcome, then run the four-step check. With a clinician, ask: “Out of 100 people like me, how many avoid this problem over the next five or ten years if they take it, and how many if they don’t?”
Should I stop a medicine if its NNT looks high?
Not on your own. Your personal risk may be higher or lower than the trial average, and some medicines need a planned taper. Bring the numbers to your prescriber and decide together.
The bottom line
A percentage on its own tells you how a result was framed, not how much it helps. To judge a benefit, you need the risk with and without treatment, the time frame, and whether the change is large enough for real people to notice.
Those questions work the same way for vitamin D, statins, painkillers and exercise. Asking them protects you from big-sounding promises, and it lets you choose small, safe benefits with your eyes open.
Related reading
About this article
This article was researched using official U.S. government sources, Cochrane reviews and peer-reviewed studies, listed below, and was reviewed in October 2026. It is general education, not medical, legal or financial advice..
Sources
- Hahn J, Cook NR, Alexander EK, et al. “Vitamin D and Marine Omega 3 Fatty Acid Supplementation and Incident Autoimmune Disease: VITAL Randomized Controlled Trial.” BMJ, 2022;376:e066452. pubmed.ncbi.nlm.nih.gov
- Akl EA, Oxman AD, Herrin J, et al. “Using Alternative Statistical Formats for Presenting Risks and Risk Reductions.” Cochrane Database of Systematic Reviews, 2011 (plain-language summary). cochrane.org
- Chou R, et al. “Statin Use for the Primary Prevention of Cardiovascular Disease in Adults: A Systematic Review for the U.S. Preventive Services Task Force.” Agency for Healthcare Research and Quality, Evidence Synthesis No. 219, 2022. uspreventiveservicestaskforce.org
- American Statistical Association. “American Statistical Association Releases Statement on Statistical Significance and P-Values.” ASA, March 7, 2016. amstat.org
- Federal Trade Commission. “Health Products Compliance Guidance.” FTC, December 2022. ftc.gov
- Farrar JT, Young JP Jr, LaMoreaux L, Werth JL, Poole RM. “Clinical Importance of Changes in Chronic Pain Intensity Measured on an 11-Point Numerical Pain Rating Scale.” Pain, 2001;94(2):149-158. pubmed.ncbi.nlm.nih.gov
- Leopoldino AO, Machado GC, Ferreira PH, et al. “Paracetamol Versus Placebo for Knee and Hip Osteoarthritis.” Cochrane Database of Systematic Reviews, 2019 (plain-language summary). cochrane.org
- Lawford BJ, Hall M, Hinman RS, et al. “Exercise for Osteoarthritis of the Knee.” Cochrane Database of Systematic Reviews, 2024 (plain-language summary). cochrane.org
- U.S. Food and Drug Administration. “Presenting Quantitative Efficacy and Risk Information in Direct-to-Consumer (DTC) Promotional Labeling and Advertisements.” Guidance for Industry, Revision 1, December 2023. fda.gov
- U.S. Food and Drug Administration. “FDA Launches Crackdown on Deceptive Drug Advertising.” Press release, September 9, 2025. fda.gov