© 2003 American Thoracic Society
Computed tomography is reasonably reliable for screening lung cancerTo the Editor:We read with interest the single-arm prospective cohort study of Swensen and colleagues (1) about the use of low-dose spiral computed tomography (CT) for screening lung cancer. The authors rightly underscored the high rate of benign nodule detection but they failed to point out that CT is reasonably reliable for screening lung cancer. We found two reasons. First, different endpoints were used to address the issues of false-positive and false-negative rates. Under the subsection "false-positive rates", it reads, "We estimate that approximately 98% of these are falsely positive findings." Lung cancer, a clinically significant event, was appropriately used as the endpoint. A high false-positive rate often goes with a high sensitivity and a low false-negative rate. We were puzzled when we read under the same subsection, "Twenty-six percent of participants had nodules that were missed on the baseline scan. This is a high false-negative rate." Here, the authors apparently referred to a different endpoint, i.e., uncalcified lung nodules of unknown clinical relevance, instead of lung cancer. Most of these missed nodules were probably clinically irrelevant because only six (2%) of them were 8 mm or bigger, which was the modal size range of the 25 detected lung cancer nodules. If lung cancer was used as the endpoint, there were two false-negative cases. Second, the authors did not examine the effect of different cut-off values on false-positive rates. Let us illustrate this point with a cut-off value of 8 mm. Twenty-two (88%) out of the 25 detected lung cancer nodules were 8 mm or bigger. The majority of the 782 participants did not have lung cancer and only 84 (10.7%) of their prevalence lung nodules were 8 mm or bigger. Therefore, the false-positive rate might be greatly reduced, whereas the sensitivity might be reasonably preserved when a different cut-off value is adopted. Newer imaging modalities have allowed the detection of tiny lung nodules but most of them are of unknown significance. Inter-observer variability also gets more significant near the lower limit of detection. Although lung cancer must start off tiny, and there are exceptional cases of early metastases, it does not necessarily follow that we should always go after the smallest nodules. A more critical analysis of specificity and sensitivity is called for in harnessing the developing technology for the benefit of mankind. Resource implication is important but the costs are bound to change. Traditional wisdom will also change. The question is when and how. Swensen and colleagues were given an opportunity to respond this letter, but declined to do so.
Department of Health Hong Kong, China REFERENCES
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
