Wednesday, August 15, 2012

Lung cancer screening: understanding "relative risk"

Enthusiasm for lung cancer screening has never been higher, following last year's report from the National Lung Screening Trial (NLST) that heavy smokers who underwent 3 rounds of low-dose computed tomography (CT) screening were 20 percent less likely to die from lung cancer compared to a similar group that was screened with chest x-rays (a "placebo" intervention that another study found to be ineffective). My own institution, Georgetown University Hospital, was one of the study sites for the NLST, and has been advertising its lung cancer screening services to physicians and the general public.

In a previous post, I offered a few reasons for eligible patients to consider not getting screened for lung cancer: 1) the risk of developing cancer from the CT scan itself; 2) high rates of false positive tests; 3) overdiagnosis and overtreatment; 4) incidental findings that could lead to additional, likely unnecessary, diagnostic procedures. I concede, however, that reasonable people might conclude that these potential harms are outweighed by the benefit of reducing one's risk of dying from lung cancer by one-fifth, especially if the patient in question has already quit smoking.

The next critical question that needs to be asked is: one-fifth of what?

The more you smoke, the longer you smoke, and the older you are, the greater your risk of dying from lung cancer. Men are also slightly more likely to die from lung cancer than women. The "average" participant in the NLST was a 62 year-old man who had smoked one and a half packs of cigarettes per day for 35 years. In the absence of screening, 20 out of 1000 men like this would die from lung cancer over the next 6 years. Taking away 20 percent of that risk with screening (20 minus 4) reduces his risk of lung cancer death to 16 out of 1000. Stated another way, about 250 people would need to be screened to prevent 1 death from lung cancer over 6 years, which, compared to other screening tests, is a pretty impressive figure. (In contrast, nearly 2000 women in their 40s need to be screened with mammograms to prevent one breast cancer death over a similar time period.)

As Drs. Peter Bach and Michael Gould point out this week in the Annals of Internal Medicine, however, few people currently being offered lung cancer screening have a risk of lung cancer death that resembles the average NLST participant. A 55 year-old woman who just kicked her 1 pack-per-day, 30-year smoking habit has only a 4 in 1000 risk of death from lung cancer in the next 6 years. Taking away 20 percent of that risk (4 - 0.8) drops it to about 3 in 1000. More than 1200 patients like this woman would need to be screened to prevent 1 death from lung cancer. And some hospitals across the U.S. are offering lung cancer screening to younger smokers at even lower risk, translating to numbers needed to screen of 35,000 or more to prevent one lung cancer death - a benefit that could easily be outweighed by all of the potential harms of screening.

Their conclusion: "The underlying chance that a person will benefit from CT screening should be considered when counseling patients about the potential benefits of being screening." Too often when physicians offer a screening test or other intervention, we tend to quote the impressive-sounding reduction in relative risk without providing patients with information about their baseline risk, or what their risk would be without it. This isn't likely to change any time soon. To make a fully informed medical decision, whenever a doctor says, "Test X will reduce your risk of disease X (or death from disease X) by 20 percent," patients should always ask, "20 percent of what"?