1. National Cancer Control Initiative, Melbourne, Vic
2. Clinical Epidemiology and Health Service Evaluation Unit, Royal Melbourne Hospital, Vic
A technique which can detect a growing pulmonary malignancy when it is small should allow more effective therapy and a better outcome for patients. Henschke and her colleagues in the Early Lung Cancer Action Project (ELCAP) screened 1000 asymptomatic smokers or former smokers aged over 60 with low-dose helical CT and conventional chest X-rays1-3. On screening, 23% had non-calcified pulmonary nodules detected by CT, while only 7% had abnormalities detected by conventional chest X-ray. Of those with nodules on CT (27), 12% were eventually diagnosed as malignancy, of which 85% were stage I tumours.
There are now other published studies, from Japan4-7, the US Mayo clinic study8, Finland9 and Germany10,11. In all of these except Finland, the proportion of cancers detected by CT that are early stage is extremely high – 80% or more in the American and Japanese series. This experience compares very favourably to normal clinical practice – of lung cancers presenting in Victoria in 1995, only 20% were localised12.
The cancer detection rate varies from 27 per 1000 screened in the ELCAP series to around five per 1000 in one of the Japanese series. To detect this number of cancers, a substantial proportion of the individuals screened have initially positive results: over 50% in the Mayo clinic, 35% in the German, and 23% in the ELCAP studies. Most of these can be managed by further non-invasive diagnostic tests without needing surgical exploration, but clearly in some of these series the numbers of people with a false-positive CT result and the necessity for work up is excessive. The ratio of the number of cancers detected to the number of individuals who test positive is the predictive value of screening (PPV). In other screening contexts such as breast and colorectal cancer, this is around 10% for cancers detected. The ELCAP study comes close to this at 12%, while the Mayo clinic series shows a predictive value of only 2%. There are substantial differences in the interpretation of CT scans in terms of what constitutes a positive test.
This evidence alone is enough to convince some people that the technique is valuable13,14. Others are not convinced15-18. Surely detection of previously unrecognised disease, with a stage distribution much better than is normally seen, is sufficient proof of benefit? However, past experience of cancer screening shows that even an improved stage distribution in the patients detected with cancer is not a guarantee that those patients have benefited from the procedure. The worry about why screening may not work is based largely on the results of earlier randomised trials, which assessed chest X-ray and sputum cytology. For example, in a randomised trial of chest X-ray in smokers in Czechoslovakia19, in the screening group 53% of tumours were early staged, 25% were operable, and five-year survival was 23%, compared to figures of 21%, 16% and zero survival in the unscreened group. However the rate of lung cancer deaths over the subsequent 15 years was actually higher in the screened group.
In the other major randomised trial of chest X-ray and sputum cytology screening, the Mayo clinic randomised trial20, in the intervention group the five-year survival was over 30%, compared to around 15% in controls. Yet again, the death rate showed from about four years onwards an excess of mortality in the intervention group compared to the control group. So both these randomised trials show higher death rates from lung cancer in the screened than in the unscreened group.
A meta-analysis of five randomised trials of lung cancer screening gives an overall mortality ratio in screened versus unscreened populations of 1.07, with 95% limits of 0.95 up to 1.2021. However, in Australia, 25% of general practitioners say they recommend regular chest X-rays for older men and heavy smokers22.
There are a number of recognised biases in screening studies23. The first is lead-time bias – that is, moving the diagnosis forward without necessarily moving death later. However, the survival rate of lung cancer is so poor that it seems difficult to see how lead-time bias could affect long-term survival. Several authors have concluded that the main effect causing the increase in mortality in the Mayo lung project and other trials is overdiagnosis bias, that is the detection and treatment of cancers detected by screening that never would have progressed to clinical disease during a person’s lifetime20,24-26.
If conventional chest X-ray and cytology pick up some such lesions, spiral CT may well identify them more frequently. Overdiagnosis is an accepted issue in screening for breast cancer, prostate cancer, cervical cancer, and melanoma amongst others. Indeed, an inevitable consequence of introducing new screening or diagnostic techniques is that clinical or pathological entities are identified whose natural history is unknown, and the natural history may be more indolent than expected24,25. In the Mayo clinic study the excess of incident lung cancer cases in the screened group was restricted to early stage cases, whereas the incidence rate of late stage cases was similar in both groups20. But while the diagnosis of biologically indolent conditions can easily be seen to lead to an increase in the incidence rate of detected tumours in the screened group, it should not lead to an excess of mortality. Black has suggested27 that some such deaths may have been due to the interventions following diagnosis, or to what he calls “sticky diagnosis” bias, that is death occurs after a clinical diagnosis, so death is attributed to the disease that was diagnosed. In that sense any excess mortality from lung cancer after diagnosis of pseudo-disease would itself be an artefact. Black elsewhere argues that the only true way to deal with this is to analyse randomised trials of screening in terms of their impact on total mortality, but unfortunately the sample size requirements of such an approach are extreme28,29.
Others argue that overdiagnosis is unlikely in lung cancer, on the basis of the rapid change in survival rate with staging of clinical detected disease, or because most untreated lung cancers will progress and cause death within five years30. However these arguments are using observations on clinically detected and treated lung cancers to apply to screen detected lesions, which may be very different.
Does helical CT confer enough improvement in diagnosis that therapy will be effective, in contrast to screening by chest X-ray or cytology? Lead-time bias, over diagnosis of pseudo-disease, failure of treatment, and prevalence-duration bias are all reasons why it is possible that patients diagnosed with small, early stage tumours after helical CT may not be better off than if they had had no such screening.
Screening is always controversial. Helical CT screening for lung cancer is at the stage breast cancer screening was at 40 years ago, with the demonstration that it can pick up small lesions and lead to the diagnosis of small lung cancers. We have no randomised trial evidence of benefit, only evidence based on clinical series compared with historical controls, which have been shown in other screening techniques to be not only of little evidential value, but positively misleading. We have the conventional view, as expressed frequently, that it is only on the basis of randomised trials that we can establish the effectiveness of a screening technology17,31,32. But we have the alternative view that just as one would say that a dramatic cure for mesothelioma would not need a randomised trial to show its benefit, because the normal natural history is so poor, it may be that CT screening for lung cancer produces such a dramatic difference in survival that evidence based on historical or concurrent comparisons in a non-randomised setting may be sufficient33-35.
Others have felt that a sceptical approach to lung cancer screening is yet another way in which patients with lung cancer are poorly dealt with. The advocacy group ALCASE (the Alliance for Lung Cancer Advocacy Support and Education) have asked “why has much of the world turned its back on people with lung cancer?” and stated that “people with lung cancer deserve the same respect, sympathy, empathy, and care that all other people who are diagnosed with cancer receive”36.
An NCCI working group is addressing several questions on helical CT. The major work on CT screening, in terms of large -scale studies, is in North America and Europe. With cervical, breast, and colorectal cancer, Australia was little involved until overseas evidence was sufficient to show that the screening techniques were beneficial. Then Australia had to decide if the results of the trials were applicable to Australia, address the logistic and cost questions, and decide if and how to set up such screening. Should we do the same for this new screening technique? Or should we be much more intimately involved with its development? Should we set up our own randomised trial in Australia to assess whether the screening technique reduces mortality at an acceptable cost? Should we join with one of the existing European or North American trials, in order both to contribute to it and to learn from the process, and so that our clinical experience stays abreast of the developing technical and scientific issues? If randomised trials carried out wholly or largely overseas do show benefits, will the results be applicable to Australia? Is the Australian clinical experience of lung diseases different? Are the prevalence rates and the clinical and pathological features of pulmonary nodules different in Australia, and do they vary within Australia in terms of smoking, asbestos exposure, or other characteristics of the patient? What will be the cost, logistic, and cost-benefit issues around CT screening? The NCCI group will not fully answer all these questions. Its objective is to provide a reasoned discussion paper and some recommendations for further work. This report should be complete by the end of 2002.
1. CI Henschke, DI McCauley, DF Yankelevitz, DP Naidich, G McGuinness, OS Miettinen et al. “Early Lung Cancer Action Project: overall design and findings from baseline screening.” Lancet, 354 (1999): 99-105.
5. S Sone, F Li, ZG Yang, T Honda, Y Maruyama, S Takashima et al. “Results of three-year mass screening programme for lung cancer using mobile low-dose spiral computed tomography scanner.” Br J Cancer, 84 (2001): 25-32.
6. S Sone, F Li, ZG Yang, S Takashima, Y Maruyama, M Hasegawa et al. “Characteristics of small lung cancers invisible on conventional chest radiography and detected by population based screening using spiral CT.” Br J Radiol, 73 (2000): 137-45.
7. T Sobue, N Moriyama, M Kaneko, M Kusumoto, T Kobayashi, R Tsuchiya et al. “Screening for lung cancer with low-dose helical computed tomography: anti-lung cancer association project.” J Clin Oncol, 20 (2002): 911-20.
11. S Diederich, D Wormanns, M Semik, M Thomas, H Lenzen, N Roos et al. “Screening for Early Lung Cancer with Low-Dose Spiral CT: Prevalence in 817 Asymptomatic Smokers.” Radiology, 222 (2002): 773-81.
12. GE Richardson, V Thursfield, GG Giles. “Reported management of lung cancer in Victoria in 1993: comparison with best practice. Anti-Cancer Council of Victoria Lung Cancer Study Group.” Med J Aust, 172 (2000): 321-4.
17. DR Aberle, G Gamsu, CI Henschke, DP Naidich, SJ Swensen. “A consensus statement of the Society of Thoracic Radiology: screening for lung cancer with helical computed tomography.” J Thorac Imaging, 16 (2001): 65-8.
20. PM Marcus, EJ Bergstralh, RM Fagerstrom, DE Williams, R Fontana, WF Taylor et al. “Lung cancer mortality in the Mayo Lung Project: impact of extended follow-up.” J Natl Cancer Inst, 92 (2000): 1308-16.