A little-noticed risk model may help identify good candidates for lung cancer screening, adding to the arsenal of tools to help physicians target their ministrations at those most likely to benefit. Unlike risk-based treatment and prevention strategies, risk-based screening is still in its infancy, but “it’s becoming more important,” said David Levy, Ph.D., a professor at Georgetown University and an expert on modeling. Risk stratification “tells us who can be most effectively screened—which candidates are most likely to have true positives,” he said. “But at the same time you want to take into account the probability of false positives, because those impose heavy psychic as well as medical costs.” A misconception lingers that testing asymptomatic patients for cancer is an unmixed blessing because of the potential to detect disease earlier, when it’s often more treatable. But that is rarely the case for any given screening test. According to the U.S. Preventive Services Task Force, colonoscopy for adults aged 50–75, Pap tests for women aged 21–65, and mammograms for women aged 50–74 are the only cancer screening tests for the general population whose benefit (cancer deaths prevented) exceeds their harms (unnecessary treatment of those who don’t have cancer). Mammography for women in their 40s and prostate-specific antigen tests for men have been controversial, as have chest X-rays and computed tomography (CT) scans for those considered at high risk of developing lung cancer. In 2011, the National Lung Screening Trial showed a 20% decrease in lung cancer deaths and a 6% decrease in all-cause mortality when current or former smokers aged 55–74 with a history of at least 30 pack-years were screened with low-dose CT annually for 3 years. However, not all lung cancer screening trials have found a benefit. And in all the trials, CT scanning found suspicious nodules in an average of 20% of participants, more than 90% of whom did not have lung cancer. Expert panels, including the International Association for the Study of Lung Cancer, have called for better ways to flag individuals at high risk for the disease to maximize screening’s benefit–harm ratio. Researchers have developed several risk models to predict individual risk for lung cancer within a specified period. One of those models, the Liverpool Lung Project, has now been validated in three independent populations who underwent CT screening. John Field, Ph.D., of the University of Liverpool, and colleagues reported their results in the Aug. 20, 2012, issue of Annals of Internal Medicine . The Liverpool model was developed to identify smokers’ and nonsmokers’ absolute 5-year risk for lung cancer. In addition to smoking duration, it used information on history of pneumonia, history of cancer, family history of cancer, and asbestos exposure. Researchers evaluated the model for accuracy by using data from the European Early Lung Cancer and Harvard case–control studies and the Liverpool Lung Project prospective cohort study. The model was simple, could distinguish between persons who will and will not develop lung cancer, and performed better than smoking duration alone as a tool to decide who should be screened. Although agreeing with the authors that the Liverpool model requires more prospective evaluation, Barnett S. Kramer, M.D., M.P.H., director of the Division of Cancer Prevention at the National Cancer Institute, called it a step in the right direction. The model does not take into account actual health outcomes, Kramer said, and “the average clinician is not going to incorporate it into practice at this point, so it’s still a research tool. [But] it leads into new directions that come closer to precision medicine.” Pamela McMahon, Ph.D. Kramer was hard-pressed to think of many other examples of risk-based cancer screening. The most notable is breast cancer, where “we have a number of pretty good models for predicting risk of getting breast cancer, and some people have tried to personalize the approach to breast cancer screening for women in their 40s.” He was referring to the recent work of researchers from the Cancer Intervention and Surveillance Modeling Network (CISNET) and the Breast Cancer Surveillance Con sortium, published last May in the Annals of Inter nal Medicine . The researchers found that women aged 40–49 with very dense breasts or a first-degree relative with breast cancer had the same risk as an average woman in her 50s and, therefore, could expect to benefit from mammography screening. Another possible example, Kramer said, is cervical cancer, “where they’re refining screening decisions based on [human papillomavirus] status.” “We’re trying to figure out what questions to ask. What if we screened for more years? What if we screened people with a shorter smoking history or started screening at an earlier age?” Risk modeling for lung cancer is about to get much more sophisticated. CISNET teams have been working with the data from the National Lung Screening Trial and other clinical trials and are likely to report their results in the next few months. Pamela McMahon, Ph.D., associate director of the Institute for Technology Assessment at Massachusetts General Hospital and principal investigator for CISNET’s lung cancer group, said the group’s goal is the same as that of the Liverpool Lung Project: “to make screening efficient by screening those at highest risk.” But whereas Liverpool is a statistical model, CISNET simulates individual people. “What we’re doing is much more detailed regarding the natural history of the disease,” she said. “Like ‘The Sims’ [a video game] without the graphics, we simulate a population over time. We can impose screening or not, have people start or stop smoking. We’re trying to figure out what questions to ask. What if we screened for more years? What if we screened people with a shorter smoking history or started screening at an earlier age?” In addition to other questions, the CISNET modelers will look at the relationship between smoking behavior and screening. Although a false-positive screening test could scare people out of smoking, it’s equally possible that others could take a negative test as a license to continue smoking. “We need to inform them that a negative screen doesn’t mean they don’t need to worry about smoking,” said Levy. Another important variable for the CISNET teams is what clinicians do to follow up a positive screening test. “If you can reduce the bad effects of a false positive,” said McMahon, “that moves the threshold toward screening more people—you could get more benefit for the same level of harm.” “Screening always involves trading off benefits to those at high risk and harms to those who won’t get it,” she explained. “We’re looking at a huge population, most of whom won’t have the cancer. There’s a small risk overall, but it’s unequally distributed. Some people will get a big benefit [screening finds cancer early, extending their life], but many will get a tiny harm [screening flags something that proves not to be cancer].” Asked whether she saw risk-based screening as the wave of the future, McMahon hesitated. “I don’t see a lot of that,” she said. “I’d like to think we’re getting closer, but there’s a lot of pushback. When you say we’re screening only those at a certain risk, people accuse you of rationing care. So it’s not a foregone conclusion.”
No takes yet. Share an insight, caveat, or question.
João Peres (2012) studied this question.