How to read a radon study without being misled by the headline

What kind of study is it?

Ecological studies compare aggregates — average radon in a county against disease rates in that county. They are cheap, they cover large populations, and they cannot tell you about individuals, because the average exposure of a county is not the exposure of anyone living in it. Conclusions drawn about individuals from aggregate data have a name in epidemiology: the ecological fallacy.

Case-control studies compare people with a condition against people without it, reconstructing individual exposure. They are much more informative about individual risk and much harder to run, because reconstructing decades of household exposure is genuinely difficult.

Pooled analyses combine multiple case-control studies to gain statistical power. When a radon finding is described as well established, it is usually pooled residential case-control evidence being referenced rather than any single study.

How was exposure actually measured?

This is where radon studies live or die. Ask whether exposure was measured in the participants’ homes or estimated from geology, whether the measurement was long-term or short-term, how many years of residence it covered, and how movement between houses was handled.

A study using long-term alpha-track measurements in the actual homes over years of residence is doing something quite different from one assigning exposure by postcode. Both may be published; only one supports strong statements about individual risk.

Measurement error in exposure assessment tends to pull an observed association toward nothing rather than exaggerating it — which is why studies with better exposure data often find stronger relationships, not weaker ones.

• Measured in the home, or estimated from location?
• Long-term detector, or short-term snapshot?
• How many years of residence were covered?
• How were house moves handled?
• Was the lowest lived-in level used consistently?

What was controlled for?

Smoking dominates lung cancer epidemiology, and EPA describes radon and smoking as compounding risks. Any radon study that does not carefully account for smoking history is difficult to interpret, and the strongest work handles it in detail rather than as a checkbox covariate.

Other confounders worth looking for include occupational exposures, socioeconomic factors that correlate with both housing type and health outcomes, and the age and construction of the housing stock itself.

A finding that survives careful adjustment is far more interesting than a larger raw association that has not been adjusted at all.

What does it actually claim, and about whom?

Read the effect size and the confidence interval, not the adjective in the abstract. A statistically significant association can be small, and a large point estimate with an interval spanning no effect is weak evidence however dramatic the number looks.

Then check the population. A study of miners, a study of one country’s housing stock, or a study restricted to a specific age band may not transfer to a household in the American Midwest, and the authors normally say so in a limitations section that the headline omits.

Finally, note whether the study is about risk or about behaviour. Research on testing rates, mitigation uptake, or disclosure compliance is useful and is not a finding about health.

What this means for your house

Almost nothing in the research literature changes what you should do, and that is a feature rather than a disappointment. The established position is that radon exposure carries risk that rises with concentration, that risk compounds with smoking, and that the only way to know your own exposure is to measure your own address.

Zone maps, county averages, and neighbourhood reputation are planning tools for public-health agencies. They describe populations. Two houses on the same street land on opposite sides of the action level often enough that treating an area figure as your own is a guess.

A new study, however good, will not tell you your number. A ten-dollar kit will.