Copyright R. Zamenhof, 2013.
Introduction
In 2009, the journal Archives of Internal Medicine carried a report that caused quite a stir in the media and among the medical community. The main conclusion of the paper was that the 57 million CT (computer assisted tomography) scans done in the U.S., in 2007, were expected to produce on a statistically predictive basis 14,500 future cancer deaths in the patient population scanned. How reliable are the data presented in this report and how logical are its conclusions? Radiation Epidemiology of CT Scanning The statement that 14,500 future cancer deaths may result from 57 million CT scans performed in one year in the U.S. at first seems terribly frightening, unless you recognize that diagnostic uses of radiation aren’t the only cause of cancer. There is a “baseline” cancer rate, due to various other factors, such as environmental carcinogens, man-made carcinogens (food, drugs, etc.), and natural background radiation. This average baseline fatal cancer-rate in the U.S. and Europe is approximately 20%; i.e., 20% of the population will eventually die of cancer. Now the fatal cancers supposedly caused by CT scans, using the Archives of Internal Medicine data, constitutes an individual risk of about 0.025%; i.e., if exposed to such CT scans, one's average likelihood of contracting fatal cancer rises from 20% to 20.025%. Not quite so frightening any more. Another way of looking at the predicted risk of death from cancer due to CT scans is to compare it with actuarial risks of death from other common human activities. For example: the risk of death from one typical CT scan—if we accept the Archives of Internal Medicine Journal’s conclusions for now and their estimate of 1 rem (10 msV) effective dose per CT scan-–is actuarially equivalent to the risk of death from smoking 220 cigarettes, drinking 360 bottles of wine, being exposed to air pollution by living in New York or Boston for 4 years, living for 14 years in Denver, or traveling 40,000 miles by car. From that perspective, the risk of death from fatal cancer contracted from a typical CT scan–-which is usually associated with a benefit in medical outcome–-again, doesn’t appear quite as threatening. So, why the overplayed response by the media and efforts at damage control by the medical and medical physics communities? An important point of concern to radiologists and physicists—both experts on radiation effects associated with medical procedures-–is the statistical interpretation of the cancer deaths ostensibly caused by CT. Most of the data linking radiation dose to fatal cancer come from observations of the effects of the Hiroshima and Nagasaki nuclear detonations in 1945, with additional data coming from studies of cancer in nuclear power plant workers in the U.K., patients exposed to x-ray fluoroscopy in U.S. and Canadian tuberculosis sanatoria between 1925 and 1954, and x-ray treatments of patients with ankylosing spondylitis (congenital fusion of the spine) in the United Kingdom. Typical radiation doses produced by CT scanning, however, are roughly 20-500 times lower than the lowest of the dose levels in the historical data referred to above. Using the historical data produced at substantially higher dose levels to extrapolate backwards to CT’s much lower dose levels, requires assumptions of the linearity of the dose-effect relationship over the very wide range of dose levels, so only shaky statistical estimates can be inferred for the corresponding associated cancer deaths at typical dose levels for CT. The BEIR-VII Committee, the dominant U.S. scientific body dealing with human effects of radiation, recently stated that “at typical CT dose levels, statistical limitations make it difficult to evaluate cancer risk in humans”. This is committee-speak for “any estimates of cancer deaths caused by CT must be considered very shaky!” To be fair, the Archives of Internal Medicine paper did include some discussion of inherent uncertainties in the data that were presented, but it failed to emphasize strongly enough the statistically shaky properties of those back-extrapolated data—exactly the point made by the BEIR-VII Committee. However, the media in reporting the paper’s conclusions omitted to mention the BEIR-VII report’s low confidence in the putative cancer deaths caused at the relatively low dose levels produced by CT. Finally, the paper’s authors underplayed, and the media reports failed to emphasize, the far larger number of patients who medically benefit from CT scans compared to those that would be statistically expected to die from cancer ostensibly caused by CT scans; that is, the very high benefit vs. risk ratio of CT was largely ignored in favor of focusing on the admittedly newsworthy aspects of the cancer deaths produced by CT. So what is the takeaway message? Do not be frightened by the reported ostensible risks of CT, but make sure that you have CT scans done in a facility certified by the American College of Radiology, and one that follows its official and rigorous recommendations for minimizing CT dose. In addition, make sure your doctor explains to you clearly the potential benefit of a CT scan, and that a medically useful outcome will result. Summary According to a report in the Archives of Internal Medicine, 57 million CT scans done in the U.S. are expected to produce 14,500 future cancer deaths in the patient population scanned. The epidemiology of radiation effects is complicated and is subject to frequent misrepresentation. Although the projected number of deaths from the CT scans seems very frightening, they should be put into proper perspective. Given that the baseline fatal cancer rate in the U.S. and Europe is roughly 20%, the fatal cancer rate due to CT scans (if the report’s conclusions are to be believed) raises this number from 20% to 20.025%. If the conventional epidemiological radiation risk/benefit model is invoked, one typical CT scan is actuarially equivalent to dying from cancer by living in Boston or New York for 4 years (due to air pollution). Finally, epidemiologists agree that at the relatively low doses typical of CT, the corresponding risk estimates are fraught with extremely large errors. In fact, the BEIR-VII Committee, the dominant U.S. scientific body dealing with human effects of radiation, recently stated that “at typical CT dose levels, statistical limitations make it difficult to evaluate cancer risk in humans”. In conclusion, the report of 14,500 projected cancer deaths due to CT scans done in the U.S in one year are based on extremely unreliable data and should be interpreted accordingly. Even if the reported cancer death rate from CT were true, percentage-wise this represents for a single individual an increase in risk from 20% to 20.025%; an increase that would most likely be buried in statistical "noise" and have very little meaning. So do not be concerned by the reported statistical fatal cancer risks of CT, but make sure that you have CT scans done in a facility certified by the American College of Radiology, and one that procedurally follows the ACR’s official and rigorous recommendations for minimizing CT dose to patients. In addition, make sure your doctor clearly explains to you the potential benefit of the CT scan you are about to have, and what medically useful outcome might result. The paper referred to in this article can be found in: Arch Intern Med. 2009 Dec 14;169(22):2071-7 |
About the Author Archives |