Do leukemia effects at low doses look convincing?

An article based on a study of more than 110 000 Chornobyl clean-up workers from Ukraine was published in November 2012 in the journal of Environmental Health Perspectives. The authors tie 14% of leukemia cases in this cohort to radiation exposure.

Risks of most types of leukemia as a result of acute exposure to higher doses of ionizing radiation are well known. But the risk associated with long-term (chronic) exposure, and the relationship between exposure and chronic lymphocytic leukemia (CLL) is not clear yet. Therefore, researchers was set the following target: assessing the relative risk of CLL and non-CLL under exposure to chronic low dose.

The authors finally reached the following conclusions: irradiation to low doses and low dose rate of workers participated in liquidation of consequences of the Chornobyl accident led to a notably increased risk of leukemia, which is statistically consistent with effects obtained for people experienced the atomic bombing in Japan in 1945. Based on initial analysis, the authors concluded that both CLL and non-CLL are radiosensitive.

After reviewing the aforementioned article some general impressions have arisen: the conclusions do not look convincing. So here are some comments on this subject:

The first that catches the eye, the standard deviation (SD) of the average exposure for both the case and control groups is approximately 2.5 times higher than the average dose (M) (see Table 1). Keeping in mind that the dose is a positive value, this ratio SD / M means that the dose distribution is very unbalanced and use a normal distribution for its fit is not correct. It would be naturally to expect a logarithmic like shape of the individual dose distribution. For such a distribution its central statistical moment is calculated based on the median rather than the mean. In addition, for the distribution with a lengthy right tail the mean is very unreliable and usually much higher than the median. Just look table 1, where the maximum dose almost 30 times higher than the average dose for both groups, and the minimum dose is zero!

Each individual dose estimate in this article is performed by Realistic Analytical Dose Reconstruction with Uncertainty Estimation (RADRUE) also has its own error. But these errors were not even mentioned. Thus, according to Figure 1 the dose errors (horizontal bars) are not reflected there. One can only assume how big these errors may be for a method a sort of RADRUE.

And the last but not the least comment. Obviously, the final conclusions in the article made under considerable uncertainty, i.e. in conditions of (1) large statistical errors of the estimated value of the relative risk, (2) incorrect shape of  statistical fitting of dose distribution and (3) large hidden dose errors. At the same time the main conclusion of the authors is the manifestation of the additional effect of 26% at 1 Gy (in the article it is written as: ERR / Gy = 1.26). But given that the difference between the average dose of the case and control groups is just 50 mGy (i.e. 50 = 132 - 82, see. Table 1) the actual additional effect, which was observed by the authors, was 1.3% only (i.e. 26% * 50 / 1000 = 1.3%).

So the effect of 1.3% for the case cohort of 137 cases when the purely statistical error (i.e., the square root of the variance of a Poisson distribution, e.g., see Poisson distribution) is approximately 12 (i.e. SQRT [137]), and also in conditions of other significant aforementioned errors, can hardly be considered as convincing.

Source:

• Lydia B. Zablotska, et. al. Radiation and the Risk of Chronic Lymphocytic and Other Leukemias among Chornobyl Cleanup Workers. Environmental Health Perspectives, November 8, 2012, http://ehp.niehs.nih.gov/2012/11/1204996/

Ukrainian original:

• Чи виглядають ефекти лейкемії при малих дозах опромінення переконливими?