Perception, paradox, paradigm: Alice in the wonderland of lung cancer prevention and early detection

In less than a century, lung cancer has progressed from a medical curiosity to the most deadly of all malignant diseases on our planet. Because cigarette smoking is responsible for the existing global lung cancer epidemic, policy initiatives have focused almost exclusively on primary prevention. There is no question that smoking prevention is the most effective method of reducing future lung cancer mortality rates among children, adolescents, and young adults. However, smoking cessation has limited effectiveness as a lung cancer prevention strategy among long term smokers, particularly in the short term.

Conventional wisdom maintains that screening for lung cancer is ineffective. This is because no randomized trial has demonstrated a significant reduction in lung cancer mortality. Indeed, mortality was higher in two of four randomized trials focusing on chest X-ray (CXR) screening for lung cancer. Accordingly, the recommendation against CXR screening is believed to be based upon powerful and direct evidence from randomized trials that CXR screening is ineffective.

Because lung cancer is almost uniformly fatal, a plausible explanation for the ineffectiveness of lung cancer screening, at least with CXR, is readily apparent. Coventional widsom maintains that in lung cancer, the asymptomatic preclinical interval is so short that apparently localized cancers are already metastatic when they are detected at an apparently localized stage. Accordingly, "early" lesions are not truly amenable to cure through surgical resection. The problem with this interpretation, however, is that it pays no heed to what the data actually show. While mortality reductions have not been observed, significant stage and long term survival advantages have consistently been demonstrated in populations randomized to screening. Interpretation of existing trials within the strict constraints of our accepted paradigm lends support to the hypothesis that CXR screening detects and labels as cancer a substantial number of early stage lesions that are clinically unimportant in that they would never have become clinically evident during the life of the patient. The paradox is that this hypothesis, known as overdiagnosis, is biologically implausible and is not supported by any epidemiologic or clinical evidence.

Based upon our accepted paradigm, a reduction in cause specific mortality in a population-based randomized trial is accepted without question as an unbiased and definitive measure of screening effectiveness. The mortality paradigm is dependent upon two assumptions, which relate first, to the randomization process, and second, to the confounding influence of screening biases on other endpoints. The fundamental problem, however, is that these assumptions, which should always have been the focus of investigation rather than supposition, are invalid. Reconsideration of our assumptions is imperative to a proper understanding of the effect of interventions in population-based research. Indeed, reexamination of our paradigm is key to reducing the global burden of lung cancer mortality.