Response to "CT screening for lung cancer: update 2007".

UNDER “KEY CONCEPTS AND DEFINITIONS” 1. The authors tell readers that the staging system for lung cancer “is based on differences in lung cancer survival.” Their reference is to the update on lung cancer staging by Mountain, which states that the lung cancer staging system is based on identifying the intersection of similar treatment options and similar outcome expectations [2]. 2. The authors present the term “curability rate,” although it has no concrete meaning in epidemiology. Their reference is to Martini and colleagues, where the concept of “curability rate” does not appear [3]. 3. The authors present, in Figure 1, estimates of “cure rates,” a concept that is not found in conventional epidemiology. They cite as their source for the cure rate estimates the “American Cancer Society Facts and Figures: 2006.” Yet, no cure rates appear in that report. 4. The authors claim that there is a distinction between “baseline” and “subsequent” rounds of screening. They cite Morrison [4]. In that article, no such distinction appears. In fact, Morrison focuses on a hypothetical scenario in which only a single round of screening (i.e., only the baseline round) has been performed. 5. The authors assert that “length bias” only affects the first round of screening, again referencing Morrison [4]. Morrison’s article explicitly contradicts the authors on this point, noting that length bias cannot be sequestered to certain patients in a single-arm noncomparative design (such as Early Lung Cancer Action Project [ELCAP]). The Morrison article more generally critiques the type of design pursued by ELCAP, while implicitly endorsing randomized trials such as the National Lung Screening Trial (NLST). He states that “the observed case-fatality (the ELCAP endpoint) is not an appropriate measure of the beneficial effect of screening. Outcome evaluation of a screening programme is best carried out by comparison of mortality rates in the screened population with those in another otherwise comparable unscreened population (i.e., the NLST design).”