A lot of people in the anti nuclear movement point to the INWORKS study as definitive proof that low levels of ionising radiation exposure from living near, or working at nuclear power plants is likely to increase the risk of getting cancer. Here’s why INWORKS doesn’t do that:
The data has a load of noise, as you can see from these wide error bars. They’re trying to affiliate risk with dose in a linear Non threshold manner. But at lower doses, the analysis has low statistical power and a high level of heterogeneity, despite studying 309,000 workers
This means that the workers greatly vary from each other in things like exposure, lifestyles, sensitivity and other risk factors. Therefore you cannot confidently rule out a true cause and effect of the cancer cases.
To control for such wide error bars you can use a large data pool, stratified data, sensitivity analysis etc, but INWORKS did all this and STILL has wide error bars. Which shouldn’t be surprising given that Nuclear workers are more heathy than the general public.
Which brings us to confidence intervals at doses lower than 100mGy of radiation. Such small doses are considered unlikely to cause cancer but INWORKS says there is an: Excess Relative Risk of +1.23 (123%) increase in risk compared to baseline.
Scary right? That’s just the upper bound of the 90% confidences interval. The lower bound is: -0.21 A 21% DECREASE relative risk compared to baseline. Meaning they’re 90% sure the risk of cancer is either 21% less likely or 123% more likely for very low radiation doses
INWORKS is garbage. It tries very hard to dance around an idea that low levels of radiation are a mortality risk, when the data is heterogenous and has a low number of events despite the large data pool.
Here is the full study if you wish to read it for yourself https://www.bmj.com/content/38...