Screening for lung cancer: do we need randomized trials?

Abstract

In a scientifically ideal randomized controlled trial (RCT) on the efficacy of screening for lung cancer, screening-detected cases would be allocated to immediate intervention or to no action until symptoms lead to diagnostics. The study would provide for learning about the extent to which earlier interventions (defined by disease stage and stage-conditional tumor size) enhance curability, and also about the distributions of disease stage and stage-conditional tumor size at the time of diagnosis under the particular regimen of screening and its associated diagnostics. Because ethics call for randomization to screening or no screening, this contrast no longer provides for studying the curability function of shared concern for all regimens of screening; it addresses only the overall curability advantage specific to the regimen deployed. The same information that is provided by the scientifically ideal RCT is obtainable from a noncomparative study in which each member of the study cohort is subject to both screening and early intervention, so long as the problem of "overdiagnosis" is avoided by documenting growth before biopsy for cytologic or histologic criteria of malignancy and so long as the outcome of intervention is documented by follow-up. The ethically feasible RCT, in addition to compromising the objects of study, involves validity problems of its own and is less efficient by an order of magnitude.