Indeed, it seems quite difficult for this methodology to control for "nuclear plants are near other nasty industrial things":

> While a broad set of confounding factors with distant relationship to cancer are included in the model, sources of carcinogenic pollutants are inexplicably missing. NPPs tend to be located in industrialized and densely populated regions to minimize cost of transmission. Other industries, particular chemical and petrochemical plants are well established sources of carcinogens, yet they are not considered as a confounding factor. Occupational hazards are also excluded from the model.