As health care organizations in California and across the country work to develop the infrastructure necessary to track patients’ cumulative radiation dose, the question remains: how accurate are our methods of predicting cancer risk from medical radiation? “The real question, as clinicians, is how do we make that information useful?” says Daniel Durand, MD, a radiologist and adjunct faculty member at the Johns Hopkins University School of Medicine. “In an emergency department situation, where someone has a high enough risk of dying to be there in the first place, how do you measure how much radiation impacts your decision-making?”
Durand and colleagues set out to answer this question in a July article in the Journal of the American College of Radiology, “Utilization Strategies for Cumulative Dose Estimates: A Review and Rational Assessment.” “It was an exercise in logical thought,” Durand says of the piece. “We wanted to review the information we have and examine the consensus on how radiation results in cancer, and how that information would be useful clinically. To weigh the risk of a one-in-a-thousand chance of developing cancer against the potential impact of getting useful diagnostic information—it’s hard for human beings to do that calculus.”
Estimating Risk
Central to that calculus is a framework for estimating cancer risk based on cumulative dose. Durand notes that in the absence of prospective data, most risk predictions have been based on a large public health study of residents of Japan. “More recently,” he says, “The Lancet published a retrospective study on risk of brain cancer in pediatric patients who have received head CT, but to date, this has never been studied in a prospective way with uniform measures.”
Unfortunately, the Japanese data have some obvious flaws. “The attractive thing about it is the sheer number of people and degree of follow-up that are available,” Durand says. “But the imperfections are numerous—the estimates of dose are based on how far people were from the blast site, and are also based on looking at several different types of radiation and saying they are roughly equal to a certain amount of high-energy photons. It’s not apples-to-apples.”
Currently, the most common way of conceptualizing cancer risk from radiation exposure is linear, Durand says, and the available evidence from the Japanese study backs this approach. “The way we understand the current model is that every time you get exposed, you slightly increase your risk of cancer,” he observes. “In the clinical context, that means it doesn’t matter where the patient is on the line—you are still taking him or her slightly upward in terms of cancer risk with every scan.”
Prospective data, however, could challenge the linear approach. “If it’s a non-linear relationship, a U-shaped curve, then a patient’s hundredth scan is riskier than his or her first scan,” Durand explains. “It’s an unproven idea from the standpoint of public health, but if a prospective study could prove it, cumulative dose information would be very helpful.”
Looking Ahead
Developing a model in which cancer risk can be accurately weighed against clinical utility will require extensive further study, Durand says. “Right now, there is limited clinical use for cumulative dose estimates,” he says. “Unfortunately, the cost and logistics associated with finding out how many scans a patient has had and tracking that between different centers are challenging.”
He adds, however, that if the information could be tracked, it would provide valuable data when it comes to stratifying cancer risk. “A lot of people are saying we need to track individual dose, and I agree that it is important to gather this data prospectively and mine it in the future to see what the real relationships are between cumulative dose and cancer risk,” he notes. “We need a smart way to do this that is low-cost.”
With such an infrastructure in place, researchers could work toward developing a more effective model for understanding patients’ cancer risk—although Durand predicts that this model will be more applicable to screening applications of imaging than, say, the emergency department scenario referenced above. “It’s hard to imagine that this information would have much impact on the more pressing clinical issues—the applications would be more for things like CT colonography,” he says. “Modest cancer risk increases could be considered, and the analysis would still be risk versus benefit.”
Cat Vasko is editor of HealthIT Executive Forum.