Why the Numbers Trip People Up
Study headlines love big percentages. But the same result can sound dramatic or trivial depending on how it's expressed. Knowing three terms — relative risk, absolute risk, and the confidence interval — lets you see what a finding really means.
Relative vs Absolute Risk
Suppose a supplement lowers the yearly chance of some outcome from 2% to 1.6%.
- Relative risk reduction: the drop from 2% to 1.6% is a 20% *relative* reduction. Impressive-sounding.
- Absolute risk reduction: the actual change is 2% minus 1.6% = 0.4 percentage points. Out of 1,000 people, that's 4 fewer affected.
Both describe the same result. Marketing tends to quote the bigger relative number; the absolute number tells you what it means for one person.
Number Needed to Treat
A related idea: if 0.4% of people benefit, you'd give the supplement to about 250 people for one of them to avoid the outcome. That figure — the number needed to treat — turns statistics into something tangible.
Confidence Intervals
No study measures the 'true' effect exactly; it estimates it. A 95% confidence interval is the range in which the true value most plausibly sits.
- A narrow interval means a precise estimate (often from a large study).
- A wide interval means a lot of uncertainty (often a small study).
- If the interval crosses the point of 'no effect' (a relative risk of 1.0, or a difference of 0), the result is not statistically significant — the data can't rule out 'no real effect.'
P-values, Briefly
A p-value below 0.05 is the common threshold for calling a result 'statistically significant.' It does *not* tell you the effect is large or important — only that it's unlikely to be pure chance. A statistically significant result with a tiny absolute benefit may not matter in real life.
The Takeaway for Reading Studies
Ask for the absolute numbers, check whether the confidence interval crosses 'no effect,' and treat any single small study cautiously. NCCIH and MedlinePlus both publish plain-language primers on these statistics [1][2].