Airborne radiation and congenital hypothyroidism of California newborns.

Mangano, Sherman, and Busby have recently released an article, “Changes in confirmed plus borderline cases of congenital hypothyroidism in California as a function of environmental fallout from the Fukushima nuclear meltdown.” They argued that rates of congenital hypothyroidism (CH) in California newborns were elevated after the Fukushima nuclear catastrophe in 2011, and this was associated with higher levels of gross beta radiation during this period. This disease is associated with exposure to radioactive iodine, and the beta radiation from iodine is a component of gross beta radiation. Further, they assert that radioactive iodine levels were not significant in 2012, when the levels of CH dropped again.

Dr. Paul Yarnold and myself have commented on the statistical methodology of this paper in two articles in Optimal Data Analysis (link, link).

The original article used an obsolete statistical methodology, clouding the validity of the study findings. Developers of the most current statistical paradigm therefore re-examined the data using rigorous state-of-the-art research methodology. The hypothesis that was tested in the original research — that newborns who were exposed to airborne radiation had a greater risk for CG — was statistically supported for California newborns with confirmed CH. Although the finding was statistically significant (i.e., “not a fluke, chance event”), the effect was very weak (i.e., not very many newborns were affected). Therefore a second study was undertaken to determine if the best alternative hypothesis might do a better job. The second study showed that the exposure hypothesis was best.

Unfortunately, the data for all of this research are hopelessly contaminated by a statistical anomaly called Simpson’s paradox, a type of ecological fallacy, which casts all of these findings into question. Although there is no way to prevent this confounding with the current data, the authors point out this may be fixed if radiation data and CH data are collected on a weekly or monthly basis in the future.

The CH data had different underlying statistical structures, and combining them may generate spurious results due to paradoxical confounding. And 2012 radiation data was combined with pre-Fukushima data, without explanation, even though a global radiation catastrophe occurred in between these disjoint time periods.

The authors conclude that the present findings require confirmation using non-contaminated data, but suggest that this avenue of investigation is warranted — reminding all to be mindful that the nuclear disaster in Japan is still unfolding, and to recall that the first articles on AIDS consisted of collections of a handful of case reports.