In this article, Samartzis et al. examine the association between ionizing radiation and the development of bone sarcoma. One strength of this study is that it was conducted with use of a large database. Another strength is that the study employed long follow-up times. Indeed, it is one of the largest and longest studies ever conducted on this topic.
As the authors imply, the results of the study are likely to be of particular interest to individuals undergoing certain medical diagnostic procedures and those exposed to certain occupational hazards. In addition, the lay public often takes an interest in the topic of "radiation." In this context, it is important to note that bone sarcoma is relatively rare. This rarity poses certain methodological challenges that, in turn, suggest that the findings of the study need to be viewed with some degree of caution. The need for this caution is not the result of poor study design, improper data collection, or some other flaw in the conduct of this research. It is the consequence of studying a rare disease. In their study, the authors analyzed the nineteen cases of bone sarcoma that occurred among the 80,181 subjects who met the study's inclusion criteria. As a consequence of this small number of cases, the 95% confidence interval for the threshold radiation dose is wide—0.12 to 1.85 Gy—suggesting that the 0.85 Gy point estimate of the threshold might not be all that accurate. The same uncertainty can be seen with regard to the estimate of the excess relative risk of bone sarcoma; the best estimate of the excess risk is 7.5-fold per gray, but the 95% confidence interval extends from 1.34 to 23.14-fold. The lower bound represents a relatively small increase in risk, whereas the upper one represents a tremendous increase.
The rarity of bone sarcoma may be the reason why the literature regarding the risk of radiation-induced bone sarcoma is apparently contradictory. The wide confidence intervals in this study tell us that we should expect any future replications of the study to yield somewhat different results. Also, the small number of individuals with bone sarcoma tells us that we may expect some results that are not statistically significant.
So, what should one make of all of this? The principles of evidence-based medicine tell us to use the best available data, as long as the data meet certain standards of reliability. In this study, the fact that the results are unlikely to be due to chance (i.e., they are statistically significant) satisfies one of those standards. More important, however, is the reported relationship between the radiation dose and the likelihood of developing bone sarcoma. A dose-response curve goes a long way toward showing that radiation was, indeed, the cause of the excess cancers. This means that, despite the wide confidence intervals, the estimated threshold dose and the estimated excess relative risk are our current "best guess" regarding the effects of radiation.
The case regarding the effects of low radiation doses is not closed. The current "best guess" given in this paper is far from perfect. Hopefully, "the best possible data" will never be collected.