Hype Over Risk of CT

In a study publicized today, alarms are being sounded over the risks of repeated imaging with CT scans. Draw such a conclusion cautiously. I read the actual medical journal article rather than the news story to get some perspective.

Dr Aaron Sodickson and his colleagues at Harvard published their study in the April 2009 issue of the journal Radiology. The study concludes that patients undergoing CT scans have an increase in the risk of cancer as a consequence. It also concludes that the more scans you undergo, the greater your risk. Lastly, it concludes that the lifetime attributable risk, the risk of the CT inducing a cancer, is up to 12%.

My concerns about the way this study is being presented relate to its methodology. The investigators never actually study how many cancers occur in a population undergoing repeated CT versus a population not undergoing CT. They did analyze a cohort of patients: those undergoing CT scan at their hospital over the course of one year, 2007. Their analysis was limited to determining how many lifetime CT scans these patients underwent and for what reason.

Sodickson et al did not determine the risk of developing cancer from actual cancer cases in this cohort. Instead they modeled how many cancers they would expect based on a mathematical model called the Biological Effects of Ionizing Radiation VII methodology to estimate these risks. I decided to read much of that document too, to get a look at their methodology.

A report was prepared by the Board on Radiation Effects Research of the US National Research Council titled Health Risks from Exposure to Low Levels of Ionizing Radiation: BEIR VII Phase 2. The Council extrapolated data from Hiroshima and Nagasaki atomic bomb survivors, from studies of cancer patients receiving therapeutic doses of targeted radiation, and from studies of occupational and environmental exposures. The consortium also analyzes the effect of diagnostic radiation, tests that use x-rays, the only class of radiation exposure that applies to Sodickson's CT study as far as I'm concerned. For example:

"A cohort study of 64,172 tuberculosis patients was carried out in Canada to assess the risk of cancer associated with multiple fluoroscopies (Miller and others 1989). In this cohort, 25,007 patients were exposed to highly fractionated radiation from repeated fluoroscopic examinations used to monitor lung collapse from pneumothorax treatment. Howe (1995) studied the risk of lung cancer in this cohort...The average lung dose per fluoroscopy session was estimated to be 11 mGy. The mean total dose to the lung was 1.02 Gy (range 0–24.2 Gy), and the mean number of fractions was 92. During the study period (1950–1987), 1178 lung cancer deaths occurred. There was no evidence of an association between risk of lung cancer and dose: the ERR [excess relative risk] at 1 Gy was 0.00 (95% CI −0.06, 0.07)."

That's right, the excess relative risk was 0.00. Only a handful of additional studies are described, using chest x-ray not CT scan, and citing excess attributable risks on the order of about 10 per 10,000 person years per Gy (a unit of radiation dose). In other words, out of 100 people undergoing a CT scan, and receiving a dose of less than one Gy, on average it would take more than 10 years for one of them to develop a cancer caused by the 100 CT scans, assuming you could legitimately extrapolate the CXR data.

In the Sodickson study, only 1 patient of the 31,462 patients studied had a predicted lifetime attributable risk of cancer that was 12%. They also estimate that 99% of the patients scanned had a lifetime risk of dying from cancer that was less than 1.6%.

CTs aren't done for nothing. They are done to detail injuries, detect strokes or lung clots, monitor for cancer relapse, etc. In such circumstances, the longshot of suffering a CT induced cancer is not foremost in one's mind. Dropping dead of a pulmonary embolus or a ruptured spleen is.

The problem, the risk we need to be worried about, is when CT scans are done in healthy subjects to detect disease states early, like lung or colon cancer. In populations that would be screened for lung cancer, the annual risk of developing a lung cancer is on the order of 0.5% per year. That dwarfs the 0.1% risk in ten years posed by CT scanning.

I am still waiting for the National Lung Screening Trial to mature, comparing CXR to CT scan screening of lung cancer. Initial data may be reported in 2010.

Until then, accept the findings of Sodickson et al for what they are, conjecture and food for thought supported by mathematical models, not direct observation. Even taken at face value, the risks are simply not that high. Their findings should not be regarded as a legitimate warning against CT scanning.