Recruiting at-risk relatives to this behavioral randomized controlled trial was logistically challenging, with large numbers of cancer cases required to enroll one eligible relative. High contact intensity registries (Utah and Idaho) required fewer cases to recruit one eligible relative, whereas the medium contact intensity registry (Colorado) required moderate numbers of cases to enroll one relative and the lowest contact intensity registries (New Mexico and California) required the most cases to recruit one relative.
Although California initially yielded the largest number of cases with the lowest level of effort, individuals contacted through this registry were less likely to participate than individuals contacted using other registries. It is noteworthy that the California registry was the only participating registry that utilized passive consent for study contact, whereby individuals were informed of study contact via a single mailing with no follow-up; and case names were forwarded to Family CARE staff unless they notified the registry that they wished to decline study contact. Other state registries with higher contact intensity (phone follow-up, phone or mail consent required, additional mailings) were found to yield higher participation rates. In other words, the more contact intensity, the more likely cases were to participate. A qualitative study of participant recruitment for familial cancer research by Kreiger et al.  found that cancer cases approached for familial research expressed concerns about providing information about at-risk relatives for research purposes. It is possible that more intensive cancer registry contact and information about the study alleviated some concerns that cases felt, particularly in registries that performed phone follow-up and could answer specific questions about the study.
Subsequently, both Stages 2 and 3 of the recruitment process required intensive contact measures, including notification calls, mailings, postcards, reminder calls, and second or third mailings. Despite recruitment intensity, initial registry source continued to be predictive of case participation and relative enrollment in our study. Kreiger et al.  found that individuals were more willing to participate in research if the study were endorsed by a trusted and familiar source. Our results indicate that cases recruited from Utah and Idaho were most likely to return the FCF and that at-risk relatives were more likely to participate if they were Utah residents; this may be result from participants’ familiarity with and recognition of Huntsman Cancer Institute at the University of Utah, the study coordinating site. These findings emphasize the potential impact of participant familiarity with research groups on study participation.
Cancer registry recruitment costs were lower in state registries that utilized low-intensity contact procedures. However, the cases provided by these registries required more intensive and costly recruitment efforts by study staff, and ultimately yielded fewer total eligible relatives than registries with higher contact intensity. For example, the California registry yielded high numbers of potential participants at a relatively low cost, yet engagement of these individuals by study staff proved difficult; and the effect extended into relative recruitment, as only 34 relatives (7% of the total study population) of California cases were recruited into the study. This is an important consideration for future studies, as low initial contact costs may not translate into low total recruitment costs.
Nonwhite race and Hispanic ethnicity were associated with nonparticipation among cancer cases in the Utah and Idaho registries at recruitment Stage 2. This finding is congruent with previous studies that have shown nonwhite race and Hispanic ethnicity to be significant factors of nonparticipation in screening intervention trials [23–25]. Lai et al.  found that interventions that utilize culturally targeted strategies to recruit underrepresented groups had more effective recruitment outcomes. Our study utilized cancer registries as our recruitment source; more research is needed to determine what, if any, targeted recruitment strategies aimed at increasing racial and ethnic participation would be effective through a cancer registry .
In Stage 2, rural cancer cases were more likely to return the form than their urban counterparts. This finding is novel and further investigations are needed to determine what mechanisms influence participation at this level. Our study was designed to reach individuals in geographically underserved settings for remote intervention, which might have appealed to rural cancer cases, making them more likely to participate. This increase in participation by rural cases was not mirrored by an increase in rural relative enrollment, which suggests that recruitment of cases and at-risk relatives might be motivated by different factors. Additional research examining the differences in motivation to participate between cancer cases and their relatives would be valuable, if indeed different recruitment strategies would be more effective in recruiting each population.
More than half of our study participants (52.2%) had another family member enrolled in the study. Family participation was found to be a significant predictor of relative enrollment with individuals more likely to enroll when other family members also participated. This supports the findings of Glanz et al. , who reported that 63% of participants in their intervention study examining CRC-risk counseling had at least one other first-degree relative enrolled. Familial support has been found to be an important component of adherence to colorectal screening recommendations; our outcomes build on the idea that such support may also influence relative decision to enroll in CRC screening trials .
Our intervention targeted both those who had never undergone colonoscopy and those who were overdue for colonoscopy. Individuals were more likely to enroll in our study if they had already undergone colonoscopy before. Known barriers in obtaining a colonoscopy include fear about the procedure and concerns about the preparation; it is possible that individuals who had been through the process had fewer psychological barriers than those who had never experienced the procedure, and were thus more likely to enroll . Future studies with interest in enrolling only participants who have never undergone screening may need to adjust target recruitment goals to reflect these barriers.
One study limitation was our inability to analyze inter-registry predictors of recruitment, owing to differences in registry protocols and in the data provided. Although intra-registry comparisons provide insight into the factors potentially affecting recruitment within a registry, it is impossible to determine whether true similarities or differences exist between all-registry outcomes. This inhibited our ability to determine the true extent to which registry protocols, contact intensity, and diagnosis years influenced the recruitment process.
Another limitation of our study was the restricted amount of information we were able to obtain on nonconsenting cancer cases and on relatives whom study staff were not able to contact directly. Cancer surveillance data collected by registries do not include information concerning education level, income, socioeconomic status, or other potentially influencing factors. Consequently, we could not assess the effect these variables might have on recruitment outcomes. Furthermore, privacy concerns for nonconsenting registry cases restricted the types of data available for analysis. Data limitations were also present in at-risk relatives, as information about eligible relatives was provided by cancer cases; this limited our ability to obtain certain demographic data on nonconsenting relatives.
Although we have provided information about direct costs related to recruitment, a complete cost analysis that accounted for indirect costs, such as institutional fees, administrative salaries, fringe benefits and other items was not feasible within the scope of this paper. However, as our limited data indicate, costs of identification, contact, and recruitment for behavioral interventions depend on many factors. Further research could provide an important perspective on potential ways to minimize recruitment costs while maximizing outcomes in these populations. As indirect costs can vary dramatically by state and institution, including direct cost information might represent more comparable data for design of future trials.