1. Is it true that fundamental to all screening for lung cancer is the way in which curability is a function of the size of the cancer at diagnosis. Is the effectiveness of any particular regimen of screening implicit in the size-specific rates of curability together with the size-distribution of cancers diagnosed under this regimen?
2. Related to this, is it true that the scientifically ideal RCT would be conducted within a screened cohort, would focus on the diagnosed cases of LC, and would contrast immediate resection to leaving the case "alone" until, if ever, symptoms prompt diagnosis. Would such a trial indeed address both the fundamental issue of curability and the issue specific to the particular regimen at screening? Subissue: to what extent would such a study be distorted by diagnostic biopsies causing early spread?
3. Is it true that when the RCT contrast is between screening and no screening, the study no longer addresses the fundamental issue, and that insofar as the latter remains unsettled, each new regimen requires a new RCT?
4. Getting back to #2: a) Would the scientifically ideal study actually focus on radiologically suspected but unbiopsied cases of possible lung cancer, with the study arm involving resection without biopsy? In other words, as for actual practice, does the radiologic finding per se actually represent the proper "decision node" for resection, bypassing biopsy? b) Given that an RCT in the ideal domain is unethical, and given the concern for "overdiagnosis" in respect to the smallest tumors in particular, is not the quasi-experimental counterpart of that scientifically ideal RCT actually required? With informed consent, most patients would opt for immediate resection but some would not, especially when the tumor is small - with the choice random conditionally on the size of the tumor.
5. What credibility can be accorded to the ELCAP results so long as they address curability on the premise of freedom from "overdiagnosis."

Worshop Expanded Summary Statement

• At a minimum, the concept that funding for multiple designs should be considered rather than relying only on a single strategy to evaluate lung cancer screening. Further, the urgency of the public health problem warrants an immediate response by health agencies and professional organizations to support organized data collection and evaluation of the potential benefit and costs of all study designs to answer important questions.
• A RCT with a death endpoint offers the most unbiased answer, but is not embraced with enthusiasm due to costs (mostly time) and the realistic appraisal that precision erodes over time. In some countries, a decision to offer lung cancer screening may require results from a RCT with a mortality endpoint.
• A RCT with surrogate endpoints is more attractive because of its inherent economy. Workgroup members were troubled by uncertainty that the intermediate endpoints accurately indicated mortality. It was also recognized that if one is willing to accept a study with surrogate endpoints, one should be able to accept a non-comparative study.
• A non-comparative design could be used, following similar selection criteria. Data should be accumulated from the international sites currently performing helical CT studies and molecular analysis of airway cell markers. Common data collection procedures for all centers should be organized. This effort should be multi-disciplinary in order to measure all end-results (detection data and follow-up) and address harms as well as benefits, including psycho social issues. Models for efficacy and cost-effectiveness of surrogate endpoint and non-comparative designs should be developed. Efforts should be made to seek to answer questions about selection effect, lead time bias, length bias sampling (and overdiagnosis). Realistic estimates of the influence of potential biasing factors may result in greater applicability of alternative designs using comparative data (e.g., National Cancer Institute trial data such as PLCO, SEER registry, etc.)