I should have found the BEIR VII report, BEIR standing for Biological Effects of Ionizing Radiation, some time ago. It’s been almost twenty years since BEIR V was a constant background presence at my job, the guide for understanding the radiation that my workers and I might be exposed to, one of the small number of studies used by governments to develop exposure standards. I happened on it quite accidentally, a week or so ago.
Over the past week, I’ve been reading that almost 400-page report. Well, not quite reading the whole thing. I’ve skimmed through sections, checking to see what kind of data it includes and how the data are treated, with ranges of error, reading the conclusions and summaries to check that I’ve gotten my skimming right. I read Chapters 11 and 12 carefully, on their risk assessment.
It’s important to understand how the BEIR VII committee did their work. I’ve done the kind of thing in my research that they did. It’s is some of the hardest work that scientists do, connecting the macro to the micro in order to derive a systematic understanding, in this case, of how radiation causes harm to the human body and how that risk can be quantified. In my case, the issue was how catalysts work. It requires building a framework of the best data, bridging between macro and micro.
That framework connects a wealth of epidemiological and laboratory data in order to predict, in the case of BEIR, the human health risks of radiation, and, in my case, how to design catalysts to facilitate particular chemical reactions. My project was much smaller than BEIR, and I won’t say more about it. Success in such an endeavor is not an exact fitting of every data point to an infallible prediction. Success is providing a simple and consistent model.
BEIR’s macro is a number of studies of humans who have been exposed to radiation. The micro is studies of cells in vitro and animals exposed to radiation. The ideal model would work from the kinds of radiation damage to cells as a function of type and energy of radiation, through the steps from that to cancer. The probabilities of the various steps would allow calculation of the risks to an individual or to a population. Unfortunately, the complete chain of information is a long way off. Radiation can result in other kinds of damage, but the data doesn’t support an analysis of those probabilities.
A framework is constructed with the information that is available. Some links in the framework are shorter and stronger than others, and multiple links substitute for certain knowledge. The objective is to form a network in which one bad link won’t collapse the whole structure, and I think that the BEIR VII committee has succeeded in doing this. That is not to say that things might not change in the future, as more information becomes available. But for now, BEIR VII and a few similar reports are the best that we have, and they’re pretty good.
BEIR VII winnows down the large number of studies that have been done.
Epidemiologic studies, in general, have limited ability to define the shape of the radiation dose-response curve and to provide quantitative estimates of risk in relation to radiation dose, especially for relatively low doses. To be informative in this regard a study should (1) be based on accurate, individual dose estimates, preferably to the organ of interest; (2) contain substantial numbers of people in the dose range of interest; (3) have long enough follow-up to include adequate numbers of cases of the disease under study; and (4) have complete and unbiased follow-up. Unfortunately, the published literature on environmental radiation exposures is not characterized by studies with such features. p. 235The in-vitro and animal studies are also subject to a number of criteria in order to be included.
BEIR VII continues to use the study of the survivors of the Hiroshima and Nagasaki atom bomb attacks as its primary source of epidemiological data. It is the source that comes closest to the criteria above. As more information becomes available from Chernobyl, it seems likely that it will play a bigger role in BEIR estimates. But large numbers of people are needed to study the small effects of low-level radiation.
For example, using the usual criterion for statistical testing in order to detect with probability .80 a 5% increase in risk when the baseline risk is 0.10, the number of individuals at risk in the exposed group would have to be approximately nj,E = 30,000. p. 261Several other studies, while not fully meeting the committee’s criteria, are examined for their support or lack thereof for the Japanese survivor data. There is support for the most part, but there are some departures. A blog post that appeared on March 30 lists a number of the studies covered in BEIR VII and a few more. It’s interesting, but not of the same reliability as BEIR VII.
The in-vitro and animal studies tend to support the assumption that radiation effects are linearly related to very low doses. This assumption is one of the places where critics of BEIR VII focus; both lower and higher effects have been promulgated.
Which brings me to the subject of cherry-picking. There are enough references in BEIR VII that almost anyone could select the ones that support their viewpoints. The importance of BEIR VII is in its conclusions and how those conclusions are supported; focusing on a single set of data or parts of a reference does not invalidate the BEIR VII findings. The value and meaning of individual references must be taken in context. So be wary of claims that BEIR VII supports particular viewpoints. The report itself is quite clear about its limitations.
What BEIR VII gives us is a working hypothesis, the best we have now.
I’ve given no numbers so far. I think that understanding what BEIR VII does and how it does it is important. Numbers to come in a future post.
[Cross-posted at the BMJ Blog.]