Halpern emphasizes that both qualitative and quantitative information on the enrollment process is needed to help clinicians apply trial results to individual patients. We strongly agree. Detailed clinical information about patients who were eligible but elected not to participate could yield important information about the degree of selective enrollment and the impact of this selectivity on applicability. As we mentioned in our article, however, it is not always possible for investigators in large multicenter trials to gather this information. Further developments that can facilitate these activities will be of great benefit.

Friedman raises the important issue of the lack of consensus about the definition of eligible versus screened patients. Although the original CONSORT (Consolidated Standards of Reporting Trials) statement specifically included the number of "eligible" patients in the flow diagram, no specific definition was provided (1). The revised CONSORT statement clarified matters somewhat; this flow diagram begins one step earlier in the enrollment process, with the number of patients "assessed for eligibility" (2). We used a similar framework for our analysis, which presumes that if patients are never approached, they would not be included in the initial "assessed for eligibility" portion of the diagram. Of note, a recent report found that only half of published randomized trials included a flow diagram and that inclusion of such a diagram was associated with improved quality of reporting (3). Since a lack of clarity about which information is pertinent to include in the diagram may have been one of the factors that hinders reporting, further efforts to clarify and standardize the enrollment terminology may be warranted.

Abel raises the issue of the engagement fraction, which he defines as the proportion of the full target population that was engaged by investigators as potential participants. He uses the MEDLINE database as a sample of the full target population for our study. Although we agree that this quantitative information would be very useful, in the vast majority of clinical syndromes, the true size and characteristics of the full target population (the denominator) are unknown. Furthermore, collapsing the concept of recruitment, which begins at the point of investigator–patient contact, with engagement, which describes the way in which investigators contact potential participants, does a disservice to the distinct and important information that each term provides.