Computed tomography screening for lung cancer

Abstract

Results of randomised trials are needed before recommending its adoption

It seems logical—and appealing—that early diagnosis of lung cancer is beneficial. But simple logic can be misleading when interpreting studies on cancer screening. The central issue is that the longer survival of patients with screen detected cancers results from a combination of lead time bias, length bias, overdiagnosis bias, and true effectiveness of screening.¹

By design, screening detects cancers earlier (lead time), but earlier detection may not change the time until death from cancer. Periodic screening will detect a large proportion of slower growing cancers because they persist longer in an asymptomatic state (length bias), and it may detect slow growing cancers that do not need treating (overdiagnosis). Without a control group, it is difficult if not impossible to distinguish between these effects, or even to be sure that screening has any true effect at all.

Despite this, early detection promises the best hope for reducing mortality due to lung cancer. The recently published results of the International Early Lung Cancer Action Program (I-ELCAP)² have revived hopes that effective screening measures can be developed. However, the findings of that study—that patients with lung cancers detected by computed tomography have longer survival—must be put into the context of the evidence base for lung cancer screening.³ We should also remind ourselves of the apparent counterintuitive truth that longer survival is not equivalent to reduced mortality.

A case in point is the Mayo lung project, one of several trials of chest x ray screening performed in the 1970s. Patients in the intervention arm had higher survival rates than those in the control arm, but screening had no effect on mortality, even after nearly three decades of follow-up.⁴ If this study had not included a control arm, the higher survival in patients with cancers detected by screening rather than via usual care might have prompted the inappropriate adoption of widespread chest x ray screening. Given this hindsight, will we still adopt computed tomography screening on a similar grade of evidence from the I-ELCAP study?

The publicity surrounding the I-ELCAP article² and the widespread encouragement to undergo screening for breast and colorectal cancers may make the public think that the issue is clear. But the 80% reduction in deaths from lung cancer predicted by the authors is open to question, for the reasons just discussed. The contrast between the 92% five year survival of patients whose lung cancers were resected and the death from lung cancer of eight untreated patients with stage I disease by five years is cited by the authors as evidence that resection offers a cure. However, this interpretation is problematic for several reasons. The sample size was small, some patients chose not to take up treatment, and case mix may have been a problem. The non-resected cancers may have been of a more aggressive histological type than the resected ones, most of which were slow growing adenocarcinomas.

Our aim is not categorically to dismiss the use of computed tomography screening for lung cancer. Currently, we are evaluating screening with simulation modelling, and we expect that screening with computed tomography will reduce mortality from lung cancer by a modest amount. But we must balance any potential (as yet unproved) benefits with the real risks of morbidity from invasive follow-up procedures. Between 13% and 50% of participants in recent computed tomography screening trials had positive (variously defined) baseline results, of which 88-97% were false positives (benign pulmonary lesions).^{2 5 6 7 8} Such screening will therefore subject otherwise healthy people to follow-up examinations, including needle aspirates with risks of complications and morbidity. In a recent pilot study, 13% of people with positive computed tomography results had at least one biopsy or invasive test; more than 40% of these had benign disease.⁹

Resources devoted to lung cancer screening will not be available for other, possibly more useful, interventions to reduce mortality from lung cancer. Of course, smoking cessation alone is not a panacea; half of lung cancers are now diagnosed in former smokers, so some kind of screening for lung cancer is needed. In the future, biomarkers (such as proteomic or metabonomic patterns) could be used alone or in conjunction with an imaging based technology, such as computed tomography, to identify early stage lung cancers. In addition, genetic markers could be used to identify people at high risk for lung cancer who would benefit from more intensive screening. However, we currently have no evidence that any of these technologies help with early detection.

Where does all of this leave practising doctors and their patients? Lung cancer screening may soon be shown to be beneficial, but it would be prudent to await results of ongoing randomised studies^{10 11} before recommending its adoption. We must be careful what we promise the public; evidence based science should inform policy. Recall the hasty about face after hormone replacement therapy was shown in randomised trials (as opposed to observational studies) to offer no overall benefit on mortality.¹² Such abrupt reversals degrade public trust in biomedical research at a time when scientific knowledge grows increasingly complex.