Welcome jayant

Correspondence

Volume 352, Number 9123 18 July 1998

		Low-dose spiral computed tomography for lung-cancer screening

Sir--Shusuke Sone and colleagues (April 25, p 1242)¹ conclude that their study of mass screening for lung cancer, "clearly showed the superiority of low dose spiral CT [computed tomography] for detection of small peripheral lung cancers". This may be so but the results (and any future data from this study) tell us little about the effectiveness of low-dose CT as a screening procedure.

In evaluating screening tests, sensitivity and specificity need to be shown. In calculating these, we have assumed that all 223 patients undergoing further diagnostic work-up on the basis of the results of the low-dose CT had a positive test result. This includes patients with "non-cancerous but suspicious lesions", "lesions suspicious of lung cancer", and "indeterminate small nodules less than 3 cm". 19 had histologically and surgically confirmed lung cancer. This gives a provisional sensitivity of 95% (though this will fall as missed cancers become apparent between screening scans). The specificity is 95%. The positive predictive value is 8·5%, in this population. Thus, 204 of 223 patients underwent unnecessary, extensive, and often invasive investigation for lung cancer.

Until the false-negative rate and the outcome of treatment in screen-detected cancer are known, we can say little about the effectiveness of CT as a screening test. Sone et al propose comparing outcomes after 1 year, but this analysis is unlikely to advance knowledge greatly because the length of follow-up is so short and the allocation has not been random (the screened groups are volunteers). Only well designed, randomised, controlled trials can show whether low-dose CT is an effective screening procedure for lung cancer.

*Sarah Conolly, Sarah Hearnshaw, Sarah Low, Richard Edwards

1 Sone S, Takashima S, Li F, et al. Mass screening for lung cancer with a mobile spiral computed tomograph scan. Lancet 1998; 351: 1242-45.

Sir--Shusuke Sone and colleagues ¹ report that spiral computed tomography (CT) was more accurate in mass screening for lung cancer, and led to early detection and an accurate diagnosis of lung cancer, and should be considered in the future health plans. They also claim that CT identified almost ten times as many cancers (0·48%) as standard mass screening (0·03-0·05%) in the same area. Clinically the positive predictive value (PPV) is more important than the cancer detection rate, and the PPV was only 8·5% (19 cancer cases from 223 with suspicious lesions, indeterminate nodules, and suspicion of lung cancer). 91·5% of patients referred for work-up by chest radiography and high resolution CT (some with transbronchial biopsy), proved not to have the disorder. A previous study, with chest radiography and sputum cytology found a greater PPV (19%)² and fewer false positives than Sone et al did.

Randomised trials at a population level and looking at survival or quality of life should be done before spiral CT screening is introduced. Widespread implementation of unproven screening methods makes subsequent rigorous evaluation much more difficult--indeed, it may be impossible to correct the original mistake.

Yasuharu Tokuda

1 Sone S, Takashima S, Feng L, et al. Mass screening for lung cancer with mobile computed tomography scanner. Lancet 1998; 351: 1242-45.

2 Fontana RS, Sanderson DR, Taylor WF, et al. Early lung cancer detection. Am Rev Respir Dis 1984; 130: 561-65 [PubMed].

Sir--Shusuke Sone and colleagues ¹ attribute to the small sample size of smokers the fact that 0·52% of smokers versus 0·48% of non-smokers had cancers detected by spiral CT scanner. We calculate that this study had at least 95% power to detect even a 5-fold increased risk among smokers; since the risk among smokers is expected to be in the range 10-20-fold, this study should have certainly detected the difference, assuming there is no selection bias and that non-smokers are neither ex-smokers nor heavy passive smokers. The discrepancy between the similar frequency among smokers and non-smokers in this study and the fact that 95% of lung cancers normally diagnosed are in smokers, suggests that these subclinical cancers are not clinically relevant. Even the fact that only two of the 19 cancers found were squamous-cell cannot explain this equivalence of risk.

The incidence reported for non-smokers is too high. Let us conservatively assume that the findings of one in every 200 of non-smokers having a lung cancer is a 2-year incidence. The life-time (40 year) risk, would be 1 death from lung cancer in every 20 non-smokers; in the UK one in every 200 non-smokers dies from lung cancer.² This further suggests that most cancers detected by spiral CT are not clinically relevant.

The finding is very interesting, however, biologically. Perhaps lung cancer (especially adenocarcinoma) is similar to breast and prostate cancer--ie, normal lung harbours multiple subclinical cancers, many of which will never surface in life. This might mean that the critical inhibition of angiogenesis goes on continuously in all of us and is more important in homoeostasis than previously supposed.³

*Jayant S Vaidya, Michael Baum

1 Sone S, Takashima S, Feng L, et al. Mass screening for lung cancer with mobile spiral computed tomography scanner. Lancet 1998; 351: 1242-45.

2 Wald NJ, Hackshaw A. Cigarette smoking: an epidemiological overview. Br Med Bull 1996; 52: 3-11 [PubMed].

3 Folkman J. New prospectives in angiogenesis research. Eur J Cancer 1996; 32A: 2534-39.

Author's reply

Sir--Sarah Conolly and her colleagues and Yasuharu Tokuda argue that our study of spiral CT reveals little about the method's effectiveness as a screening procedure. It is argued that the follow-up is too short and the lack of randomisation is noted. We are conducting this Nagano project on lung-cancer screening with a mobile spiral CT scanner in a 5-year programme to investigate the medical applications of satellite communications, from 1995 to 1999, and we cannot, unfortunately, extend the follow-up. Although the follow-up will be too short to reveal the outcome for patients treated in the project we will be able to establish the radiological diagnoses for almost all those receiving CT screening.

To improve outcome in lung cancer we must detect and treat it much earlier--eg, Sagawa et al¹ reported 5-year-survival rate of 83% for patients with lung cancer measuring 2 cm or less and no lymph-node metastasis. Because most of the cancers found by CT alone were smaller than 2 cm with no lymph-node metastasis, yet showed no evidence in the chest radiograph, we would expect a better outcome than for those found in the chest radiograph.

We are now accumulating data on sensitivity and specificity; however, there is a question about how to define the presence of lung cancer before interpretations can be classed as false negative or false positive. There seems to be no consensus, from a clinical standpoint, about size of tumour;¹ 3 mm or 5 mm may be the clinically significant threshold because lung cancers less than 3 mm are very difficult to detect even on conventional CT and nodules less than 5 mm are difficult to establish radiographically or by biopsy-based histology. 5 mm seems an appropriate threshold and we allocated a case to false negative when a tumour of 5 mm or more was missed in the CT image--but other professionals may hold different opinions. We cannot pursue high specificity when interpreting screening CT images because that will increase missed cancers. We must maintain a high sensitivity while at the same time avoiding cursory interpretation of screening CT images, to keep the numbers needing further work-up exams within a reasonable range.

The X-ray dose used in the CT screening is relevant here. The low exposure dose means that the image quality reveals lung nodules but is inadequate for a precise differential diagnosis. We prefer a low dose in screening symptomless individuals, reserving higher-dose diagnostic examinations for those with suspicious or indeterminate nodules. In other words, the role of CT screening is to check the presence or absence of a lung nodule not to test for lung cancer. Therefore we do not feel it appropriate to talk of specificity for CT screening in detecting lung cancers (rather than lung nodules). By the way, our diagnostic work-up is not as expensive or invasive as your correspondents imply. We do further CT scans without contrast and rarely recommend bronchoscopy. Nor were all the work-ups on non-cancer patients unnecessary; some patients had non-cancer lung lesions that demanded medical consultation or treatment.

Jayant Vaidya and Michael Baum argue that the incidence of lung cancer (mainly adenocarcinoma) among non-smokers was too high and they suspect that most cancers detected by spiral CT are not clinically relevant. I agree that more needs to be known about the growth characteristics of this type of cancer if we are to manage patients on a sound scientific basis. In the meantime, however, we should look for early (preclinical) cancer and treat it;² most patients with lung cancer die because of delay in diagnosis, and we do not yet know how to discriminate the preclinical but life-threatening cancer from indolent one. The prevalence rate of CT-screening-detected cancer was nearly 5 per 1000 screened (males 10 cases in 2115 screenees, females, 11 in 1852). The age and sex-adjusted expected cancer incidence in the screened population was 4·57 (male 3·85, female 0·72) for 3967 screenees based on the data in the Cancer Registry of Japan. This means that we have detected nearly 4·6 times as many cancer patients as expected (2·6 times in males, 15·7 in females). I suspect that our high detection rate was due to the inclusion of lung cancers missed by the general health survey of the previous year, and adenocarcinoma of the lung in the female may lie undetected by conventional chest radiography for over 15 years on average.

Shusuke Sone

1 Sagawa M, Saito Y, Takahashi S, et al. Clinical and prognostic assessment of patients with resected small peripheral lung cancer lesions. Cancer 1990; 66: 2653-57 [PubMed].

2 Noguchi M, Morikawa A, Kawasakii M, et al. Small adenocarcinoma of the lung. Histologic characteristics and prognosis. Cancer 1995; 75: 2844-52 [PubMed].