The relationship between emphysema on CT scan and lung cancer.

I commend Maldonado and colleagues 1 for their extended reanalysis of lung cancer screening data to explore the association between emphysema on CT scan and lung cancer in a recent issue of CHEST (December 2010). Attempts to validate previously published fi ndings are all too rare. The authors report the absence of a clear relationship between emphysema (quantitative or dichotomous) and the odds of lung cancer. As mentioned in the discussion, these fi ndings differ from the more than threefold risk associated with emphysema found in two independent prospective cohort studies. 2 , 3 Explor ing why such different observations were made will help advance our understanding of the seemingly complex relationship between lung cancer, COPD, and emphysema on chest CT scans. Maldonado et al 1 acknowledge that Wilson et al 3 observed an all-or-none effect, with even trace ( , 10%) or mild (10%-20%) emphysema increasing the risk of lung cancer. Yet their dichotomous analysis uses a threshold for emphysema at  15%. I would be interested to know the crude and adjusted impact of any emphysema (ie, . 0%), as this was the threshold used by de Torres et al 2 and Wilson et al. 3 Indeed, from the data provided in Table 2 of the article by Maldonado et al 1 with a threshold of 5%, the crude OR of lung cancer in the presence of emphysema is 1.7 (1.0-2.9). It may be that automated techniques or the threshold of 2 900 Hounsfi eld units generate too many false-positive diagnoses of minimal emphysema, a disease severity clinically important with respect to lung cancer risk. Conversely, it is possible that the automated quantifi cation technique used by Maldonado et al 1 eliminated a potential observer bias in the studies of de Torres et al 2 and Wilson et al. 3 Lung cancer cases that were detected on the initial screening scan were presumably apparent to the CT scan readers (eg, masses, adenopathy) and may have infl uenced their assessment of emphysema. Such a bias could spuriously create a false association with lung cancer. In summary, the work by Maldonado et al 1 forces us to refi ne our understanding of the relationship between emphysema on CT scan and lung cancer and raises new research questions. Perhaps the application of their quantitative algorithm to the datasets of de Torres et al 2 or Wilson et al 3 or the manual grading of emphysema within their dataset would advance our understanding.