Screening for Lung Cancer With Chest Radiographs

Abstract

WITHIN THE SPACE OF SEVERAL MONTHS, 2 VERY large randomized trials of screening for lung cancer have reported their findings, which fortunately complement one another. The National Lung Screening Study (NLST) found that annual lowdose computed tomography (CT) reduced lung cancer mortality by 20% relative to annual chest radiography. In this issue of JAMA, investigators from the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Randomized Trial report that annual screening chest radiography does not reduce lung cancer mortality relative to no screening. Should clinicians infer that screening with low-dose CT reduces lung cancer mortality by 20% relative to no screening? This editorial addresses that question and several other aspects of the PLCO trial. Why would the National Cancer Institute sponsor a large trial of screening for lung cancer with chest radiography? Although 6 randomized trials, most of them published in the 1980s, found no evidence that screening radiography reduced lung cancer mortality, the control group received screening chest radiography in all but the Mayo Lung Project. This study had important protocol deviations and was relatively small (9211 participants randomized; 366 cancers detected). So the body of evidence was inconclusive according to the investigators who designed the PLCO trial. The PLCO trial measured the effect of a package of screening interventions aimed at preventing death from 4 cancers. Patients were individually randomized to a usual care group or to an intervention group that was screened periodically for prostate, lung, colorectal, and ovarian cancer for 3 years and then monitored for PLCO cancers (a stop-screen design). Recruiting targeted the US general population aged 55 through 74 years. The study was designed to have a 90% probability of detecting a 10% reduction in lung cancer mortality. Over 8 years, 77 445 participants were randomized to screening and 77 456 to usual care. Half of the participants were ever-smokers; 10% were current smokers. Lung cancer mortality, the primary end point, was 14.0 per 10 000 person-years of follow-up in the intervention group and 14.2 in the control group (rate ratio, 0.99; 95% CI, 0.87 to 1.22). In high-risk patients who met the NLST eligibility criteria, the outcome was similar except for higher lung cancer mortality rates. The PLCO trial shows that a short-term chest radiography screening program has no effect on lung cancer mortality. The only potential concern about the validity of this conclusion is the reporting of follow-up contact with the trial participants. The authors’ diagram of the flow of participants through the trial (Figure 1) does not state the number of participants in the usual care group that the authors were unable to contact during follow-up. Differential ascertainment of lung cancer mortality and, especially, incidence could occur if follow-up rates were unequal in the screening and usual care groups for reasons linked to lung cancer incidence and mortality. Most of the 1696 cancers were interval cancers (n=198), arose in patients who were never screened (n=193), or arose in patients who had completed 3 rounds of screening (n=998). These far outnumbered the screen-detected cancers (n=307). The large number of deaths from cancers diagnosed after screening is a reminder of the transitory benefit from a short-term program of screening for lung cancer. The best test of lung cancer screening in high-risk individuals would be a trial of lifelong screening. With a stop-screen design, the true effect of screening is unclear because of uncertainty about how long to monitor patients after screening. At one extreme, stopping right after the last screen will miss cancers that screening did not detect but that would have been diagnosed if monitoring had been extended into the postscreening period. Stopping monitoring too soon, therefore, may overestimate the effect of screening by covering up some of its failures. At the other extreme, extending the period of monitoring long past the time when the last cancer missed by screening would have been diagnosed will underestimate the