Abstract
Background:
Thoughtful approaches to study recruitment are a critical step in designing and implementing randomized controlled trials. Delays and challenges in recruitment can be costly and can result in smaller than proposed sample sizes which have downstream effects, such as underpowered studies.
Purpose:
The current study evaluated recruitment methods (e.g., targeted mailings, brochures/flyers, social media) and their relationship to eligibility, randomization, participant characteristics and retention at end of a randomized controlled trial of physical activity adoption and maintenance among breast cancer survivors.
Methods:
Screening data from 874 women was analyzed for recruitment method, study eligibility, randomization and retention through end of treatment. Costs per randomized participant were calculated by recruitment method. Baseline participant characteristics were compared across recruitment methods and between randomized and retained participants.
Results:
Rates of participant accrual from eligibility screening through randomization differed statistically significantly depending on recruitment method (p <.05). The highest randomization rates were obtained via targeted mailings (88.2%) and lowest via brochure/flyer (0.4%). Among the randomized sample, there were no differences in demographic characteristics between recruitment methods, however, there were differences within the targeted mailings category (registries vs. organizational recruitment). There were statistically significant differences in costs between recruitment methods (p<.001) with lowest (non-negligible) cost being targeted mailings.
Conclusions:
The current RCT of breast cancer survivors successfully recruited and retained participants, highlighting the need to combine recruitment methods to achieve accrual goals. Recruitment methods differed substantially in their cost and their ability to attract individuals who would ultimately be randomized.
Keywords: Recruitment, randomized controlled trial, breast cancer survivors, recruitment costs
Introduction
Recruitment of participants for interventions tested using randomized controlled trial (RCT) designs can be a lengthy and expensive process. Stringent inclusive/exclusive criteria result in much effort expended to secure samples of sufficient size and statistical power to detect effects. Other factors such as overestimation of eligible candidates, participants not wanting to participate in a control arm, etc. can add to the challenges and researchers may have to change their recruitment approaches, extend the recruitment phase, and likely expend more funds. Retention of participants raises other challenges but both achievement of accrual goals and retention of participants are required for robust statistical power. Difficulties in either or both can adversely affect the ability to detect intervention effects. It has been reported that only 55% of clinical trials in the U.K. accrue their targeted sample size and 78% achieve 80% of the project sample size [1]. For these reasons, a recent Cochrane review recommended that trialists should include evaluations of recruitment strategies in their trials [2].
As others have pointed out, recruiting breast cancer survivors is a time-intensive, expensive and challenging endeavor 2016[3]. Our own research team has used various recruitment methods for a RCT to test the effects of a peer mentoring approach to physical activity adoption and maintenance among sedentary breast cancer survivors. Prior research had showed the feasibility and acceptability of the intervention [4] followed by a RCT that demonstrated the efficacy of the intervention on moderate-to-vigorous physical activity at post-intervention (3 months) and at 6 month-follow up [5]. These studies were conducted in New England and at sites that were not designated NCI Cancer Centers. The current RCT that built upon our previous studies and was powered to examine the efficacy of the intervention on physical activity adoption and maintenance over a 12 month period and was conducted in the SouthEastern U.S. (also not at NCI designated Cancer Centers). All assessments and intervention contacts were distance-based, in-person visits were not required and no biological samples were required: features which we anticipated would reduce the challenges of participant recruitment and retention.
Randomized controlled trials of behavioral interventions have used various recruitment approaches, commonly cancer registries, in-person recruitment and more recently, social media [7,8, 9,10,11,12) with varying degrees of success (e.g., yield of eligible participants, numbers randomized, numbers retrained). The demographic characteristics of participants recruited also vary across the approaches so that investigators may have to carefully select the approaches that will yield participants with the characteristics required for trial participation (e.g., age, race and ethnicity, etc.). In many instances, the costs of these approaches have not been tracked making it difficult for investigators to be strategic about their choice of recruitment approaches within time and costs restraints. The challenges of recruitment are particularly acute in countries like the U.S. that do not have a national healthcare system making it difficult to contact participants and requiring researchers to approach each state cancer registry and/or hospital-based registries separately.
The current study examines the efficacy of recruitment methods in terms of return as measured by the proportion of screened participants who were eligible, the proportion of eligible participants who were randomized, and the proportion of randomized participants retained at the end of the study (12 months). We were also interested in assessing the costs involved with the various recruitment methods (direct financial costs excluding manpower). Finally, we wanted to compare the socio-demographic characteristics of individuals recruited through the various methods to inform future research with this population and to explore whether or not these characteristics predicted participant retention at 12 months. We anticipated that these data would be useful for other investigators in making decisions about which recruitment approaches to invest in to achieve accrual targets, particularly those working in countries that do not have population-based cancer registries.
Method
The original recruitment approach was to collaborate with the American Cancer Society (ACS), the community organization that was the partner in the funded trial. As originally proposed, the ACS sent informational mailings to their breast cancer constituents in SC, GA, and NC. Due to low yield, mailings were then provided by ACS offices in FL, MD, DE, VA, WV, NY and NJ (a total of 8,293 mailings). Informational mailings were also provided by the SEER registry in GA, and tumor registries in NC, hospital/clinics in SC and NC (n=1,194) and the Army of Women which is sponsored by the Dr. Susan G. Love Foundation (n=116). In addition, study information was provided on social medial (FB and Instagram), print advertisements and radio, mailings from physician offices, in-person recruitment at community events and snowball sampling. All recruitment information specified that only participants in the Eastern Time Zone (U.S) were eligible because the intervention was delivered by phone calls from the coaches to the participants at mutually convenient times. Recruitment was conducted from February 2016 to December 2018.
Patients were provided brief information about the study via the various recruitment approaches and asked to call the study staff using a local or toll-free number. The staff explained the study to them and obtained verbal informed consent to conduct a telephone screen for eligibility. If eligible on the phone screen, the staff sent the participant an informed consent document in the mail, along with a physician consent form. After the consent forms were obtained, participants were mailed a packet of questionnaires and an Actigraph (Model GT3X) accelerometer with instructions to wear the Actigraph for 7 days, and envelopes for returning the questionnaires and Actigraph. After the completed forms and Actigraph were received, the Research Assistant contacted the participant to conduct a physical activity interview (7 Day Physical Activity Interview, [6]). Finally, after the baseline assessments were completed, the participant was randomized to one of the three study groups.
There were three study groups: Reach Plus, Reach Plus Phone and Reach Plus Messages. The peer coaches for this study as in prior work [4,5] were volunteers recruited from the American Cancer Society’s Reach to Recovery program. Reach to Recovery is a peer-based program of breast cancer survivors who provide informational and emotional support to breast cancer patients, generally over the telephone. All three study groups received a 3-month home-based intervention delivered by phone by the Reach to Recovery peer coaches. These coaches were recruited for the study and were trained to deliver the home-based program by the research as we had done in prior work[4,5]. The 3-month intervention that promoted the adoption of least 150 minutes of moderate-intensity physical activity/week was the same theoretically grounded intervention that had previously been tested in a randomized controlled trial [5]. After the 3-month intervention and assessments, all study participants received a 6-month program aimed at promoting maintenance. Reach Plus participants were asked to continue to self-monitor their physical activity using a pedometer and received monthly physical activity feedback reports from the study team. Reach Plus Phone participants received monthly phone calls from their coaches aimed at encouraging physical activity maintenance; they were asked to self-monitor their physical activity using a pedometer and received monthly feedback reports. The Reach Plus Messages group received a text or email message each week for 6 months aimed at supporting physical activity, in addition to self-monitoring their physical activity and receiving physical activity feedback reports. Assessments of physical activity and related outcomes occurred at baseline, at 3, 6, 9 and 12-month follow ups. Intervention contact with all participants ended at 9 months. Hence, all study participants received a 3-month physical activity adoption intervention, followed by 6-months of interventions for physical activity maintenance.
Eligibility Criteria.
Women aged >21 years were eligible if they: 1) were diagnosed in the past five years with Stage 0–3 breast cancer (patients on radiation or chemotherapy were eligible as were patients receiving hormone therapy), 2) were able to read and speak English, 3) were ambulatory, 4) were sedentary (i.e., <30 minutes/week of vigorous exercise or <90 minutes/week of moderate-intensity exercise for the past six months), 5) were able to walk unassisted, and 6) had access to a telephone. Women with more advanced disease (Stage 4), medical or psychiatric problems (e.g., substance abuse, and orthopedic problems) that may interfere with protocol adherence were not included. These inclusion/exclusion criteria are similar to those used in our prior work [4, 5]. Participants were asked to provide consent for medical chart review to extract disease and treatment history.
Analyses
We categorized the recruitment approaches into targeted mailings (e.g., ACS mailings to breast cancer constituents, cancer tumor registry mailings, Army of Women etc.), paid advertisements in newspapers and magazines (local and regional), non-print advertisements (e.g., radio spots, TV coverage), brochures and flyers distributed to cancer support groups, cancer-related charity walks and similar events; email blasts (e.g., local church groups, university listservs, Craig’s list) and social media platforms e.g., study’s Facebook page, postings on TV personalities’ FB pages, Instagram).
Among the screened sample, the distribution of recruitment methods was summarized, and chi-squared tests were used to compare study flow across recruitment methods from screening through randomization. The goal was to understand whether or not there were differences in the proportion of screened participants who were eligible for the study (for example) depending on type of recruitment.
To examine factors associated with participant retention, we compared baseline characteristics among the randomized sample, across recruitment methods using Analysis of Variance (ANOVA) for continuous variables, chi-squared tests for categorical variables and graphical methods as appropriate. Subsequently, study retention (from randomization through end of follow-up) was compared across recruitment methods using chi-squared tests. Recruitment methods were coded individually as a categorical variable (targeted mailings, paid advertisements, non-print advertisements, social media, brochure/flyer distribution and email blasts) and subsequently collapsed as targeted vs. non-targeted advertising.
Finally, cost per participant randomized was estimated for each recruitment method and compared using ANOVA. To determine cost per participant randomized, total costs for each recruitment method were tallied (excluding staff time) and divided by total randomized participants. All analyses were run in SAS 9.0 with alpha set a priori at .05.
Results
Study Flow
Pre-randomization Period
The full study consort is presented in Figure 1. Overall, 1,235 women contacted the study team and 874 were screened (70.8%). Of those 874, 441 (50.5%) were deemed ineligible, 279 were eligible at phone screen (31.9%), 114 (13%) became non-responsive to efforts to reach them, and 40 were no longer interested (4.6%). Next, we considered the sample of 279 participants deemed eligible via screening. Of those, 63 became ineligible (reasons such as became too active, see reasons in Figure 1), 15 were no longer interested, 32 did not respond to efforts to reach them and the remaining 169 were randomized: 5 of the 169 never began the trial and 3 became ineligible at post-randomization (because they had become too active) resulting in 161 randomized to one of the three study arms.
Figure 1.
Consort Diagram
Rates of participant accrual from eligibility screening through randomization differed significantly depending on recruitment method (p <.05). Of the 279 individuals who were deemed eligible to participate post screening. 87.5% of participants endorsed being recruited via targeted mailings vs. other all other methods (p<.001). Of those eligible, 57.7% of participants continued on to be randomized into the trial. The proportion of eligible participants who converted into randomized participants ranged from 40.0% (N=2) of those who were recruited via email to 100% for brochure/flyer-recruited participants(n=1). However, cell sizes were low in these cases. A similar pattern of findings (favoring targeted mailings) was seen when comparing proportion of screened participants ruled out for single and multiple reasons during recruitment and proportions of those ineligible post screening.
As the greatest proportion of eligible participants were recruited via targeted mailings, we probed further into potential differences between sources of targeted mailings. Of the 244 eligible participants, 33% were recruited via the American Cancer Society, 34% from Cancer Registries, 18% from Hospitals/Clinics and 14% from Organizational Recruitment Services (e.g., Army of Women). Of those eligible American Cancer Society participants, 59% were ultimately randomized. Of those eligible Cancer Registry participants, 75% were randomized compared to 44% of eligible Hospital/Clinic participants and 54% of Organizational Recruitment participants. This constituted a statistically significant difference between sources with respect to transition from eligible to randomized (p<.01).
Participant Randomization and Retention
Of those randomized (n=161), 88.2% were recruited via targeted mailings, 1.2% via paid print advertisements, 5.0% via non-print advertisements, 5.0% via social media/email blasts, and 0.6% via brochure/flyer. Taken together, data suggest a statistically significant difference in recruitment rates by recruitment strategy (p<.001) with greatest proportion of recruited participants being reached via targeted mailings.
Participant Characteristics
A complete description of the final screened, randomized and retained samples are presented in Table 1. It should be noted that given the samples were not independent (randomized participants were also screened), the groups could not be statistically compared, although differences between randomized and retained samples could be and were compared. Results did not suggest statistically significant differences between groups with respect to socio-demographic or cancer history variables.
Table 1.
Baseline Characteristics of Screened Sample, Randomized Sample, Retained Sample
| Screened Sample, N=874 | Randomized Sample, N=161 | Retained, N=126 | |
|---|---|---|---|
| Age, years | 57.92(11.08), Range: 26–92 | 57.33(10.85), Range: 28–84 | 57.44(10.29) |
| Race | |||
| White | 542(62%) | 131(81.4%) | 102(81%) |
| Asian | 5(.6%) | 1(.6%) | 1(.8%) |
| Black | 201(23%) | 27(16.8%) | 21(16.7%) |
| Other | 126(14.4%) | 2(1.2%) | 2(1.6%) |
| Ethnicity | |||
| Not Hispanic/Latino | 733(83.9%) | 157(97.5%) | 123(97.6%) |
| Hispanic/Latino | 20(2.3%) | 3(1.9%) | 3(2.1%) |
| Unknown | 121(13.8%) | 1(.6%) | 0 |
| Education1 | |||
| HS | 65(7.4%) | 9(5.6%) | 78(6.3%) |
| Some College | 220(25.2%) | 25(15.5%) | 16(12.7%) |
| College Graduate | 478(54.7%) | 127(78.9%) | 102(81.0%) |
| Unknown | 111(12.7%) | 0 | 0 |
| Cancer Stage2 | |||
| 0 | 100(11.4%) | 26(16.1%) | 22(17.5%) |
| 1 | 253(28.9%) | 75(46.6%) | 58(46%) |
| 2 | 224(25.6%) | 44(27.3%) | 33(26.2%) |
| 3 | 105(12%) | 16(9.9%) | 13(10.3%) |
| 4 | 7(.8%) | 0 | 0 |
| Unknown | 185 (21.2%) | 0 | 0 |
| Radiation | |||
| No | 244(27.9%) | 49(30.5%) | 45(35.7%) |
| Yes | 483(55.3%) | 108(67.1%) | 80(63.5%) |
| Unknown | 147(16.8%) | 4(2.5%) | 1(.01%) |
| Chemotherapy | |||
| No | 330(37.8%) | 74(46%) | 62(49.2%) |
| Yes | 394(45.1%) | 83(51.6%) | 60(47.6%) |
| Unknown | 150(17.2%) | 4(2.5%) | 4(3.2%) |
Mean(Standard Deviation) for continuous variables; N(%) for categorical variables. Differences between columns were not tested statistically for differences as the samples were non-independent and sample sizes imbalanced which precluded comparisons.
Categories for education were different in the screener vs. baseline assessment.
Cancer stage in the screener was self-reported.
Targeted Mailings vs. Other Recruitment Methods
When comparing targeted mailings to the other recruitment methods, there were no statistically significant differences in baseline characteristics (p’s>.05) or retention at end of treatment (p>.05). However, when probing further into the type of targeted mailings, there were differences between targeted mailing types with regards to employment, education and income. Specifically, those recruited via American Cancer Society and Tumor registries reported statistically significantly more years of education (college education) and statistically significantly greater income compared to other methods (p’s<.001). Those recruited via Organizational Recruitment Services or American Cancer Society were statistically significantly more likely to report office work (clerical or professional) compared to other methods, p<.001. There were no differences with respect to age, race, ethnicity, cancer stage or treatment, p’s>.05. These details are presented in Table 2.
Table 2.
Baseline Characteristics by Targeted Mailing Type among those Randomized, N=142
| American Cancer Society, N=43 | Army of Women, N=19 | Tumor Registries, N=60 | Hospital/Clinic, N=19 | |
|---|---|---|---|---|
| Age, years | 55.05(10.64) | 49.26(11.69) | 61.08(9.31) | 59.58(11.12) |
| Race | ||||
| White | 35(81.4%) | 18(94.7%) | 50(83.3%) | 12(63.2%) |
| Asian | 0 | 0 | 0 | 1(5.3%) |
| Black | 7(16.3%) | 1(5.3%) | 9(15.0%) | 6(31.6%) |
| Other | 1(2.3%) | 0 | 1(1.7%) | 0 |
| Ethnicity | ||||
| Not Hispanic/Latino | 42(97.7%) | 19(100%) | 57(95.0%) | 19(100%) |
| Hispanic/Latino | 1(2.3%) | 0 | 2(1.3%) | 0 |
| Unknown | 0 | 0 | 1(1.7%) | 0 |
| Education* | ||||
| HS | 3(7.0%) | 1(5.3%) | 4(6.7%) | 1(5.3%) |
| Some College | 6(14.0%) | 2(10.5%) | 12(20.0%) | 3(15.8%) |
| College Graduate | 34(74.1%) | 16(87.2%) | 44(71.3%) | 15(78.9%) |
| Unknown | 0 | 0 | 0 | 0 |
| Occupation* | ||||
| Skill/Craft | 0 | 0 | 1(2.4%) | 0 |
| Scientific/Technical | 1(3.3%) | 0 | 2(4.9%) | 1(8.3%) |
| Service Work | 2(6.7%) | 0 | 3(7.3%) | 0 |
| Clerical/Office | 4(13.3%) | 2(15.4%) | 5(12.2%) | 1(8.3%) |
| Professional/Managerial/Admin | 21(70.0%) | 11(84.6%) | 29(70.7%) | 9(75.0%) |
| Sales | 1(3.3%) | 0 | 0 | 1(8.3%) |
| Other | 1(3.3%) | 0 | 1(2.4%) | 0 |
| Employment* | ||||
| Full Time | 23(53.5%) | 9(47.4%) | 31(51.7%) | 10(52.6%) |
| Part Time | 6(14.0%) | 4(21.1%) | 11(18.3%) | 2(10.5%) |
| Retired | 9(20.9%) | 3(15.8%) | 14(23.3%) | 6(31.6%) |
| Homemaker | 2(4.7%) | 3(15.8%) | 2(3.3%) | 0 |
| Medical Leave | 1(2.3%) | 0 | 0 | 0 |
| Unemployed | 2(4.7%) | 0 | 2(3.3%) | 1(5.3%) |
| Cancer Stage | ||||
| 0 | 5(11.6%) | 2(10.5%) | 14(23.3%) | 2(10.5%) |
| 1 | 16(37.2%) | 9(47.4%) | 28(46.7%) | 13(68.4%) |
| 2 | 16(37.2%) | 3(15.8%) | 14(23.3%) | 3(15.8%) |
| 3 | 6(14.0%) | 5(26.3%) | 4(6.7%) | 1(5.3%) |
| 4 | 0 | 0 | 0 | 0 |
| Unknown | 0 | 0 | 0 | 0 |
Means(standard deviations) reported for continuous variables; N(%) reported for categorical variables.
p<.05
Although there were no differences between targeted mailings vs other recruitment methods with respect to study retention (p>.10), those recruited via Cancer Registries were statistically significantly more likely to withdraw from the study before 12 months (p=.04).
Costs
The total recruitment costs amounted to $58,880. The costs for targeted mailings included printing of brochures (range $10.00-$1,175), mailing expenses (range $15,00 to $8,422) and data request charges (e.g., $3000 to the NC Cancer Registry and $3000 to the Army of Women): the total expenses for targeted mailings was $29, 381. Paid advertisements costs ranged from $1,072 (Charlotte Observer newspaper) to $5,000 (Breakthrough Research Magazine) and the total costs were $15, 145. Non-print advertisement costs such as radio spots ranged from $800 to $3000 for a total of $8,964. Printing of brochures for distribution cost $5,148. Printing of flyers cost $242. There were no costs associated with e-mail blasts and social media recruitment.
Cost per-randomized participant was estimated and compared across recruitment methods. There were statistically significant differences in costs between recruitment methods (p<.001): those recruited via targeted mailing cost $207 per participant randomized, via paid print advertisements, the costs were $7573 per participant randomized, via non-print advertisement, it cost $1793 per participant randomized, and via brochures/flyers, the costs were $5390 per randomized participant. Non-paid print advertisements, social media platforms, support groups, email blasts and other methods were negligible in terms of cost. A full description of costs/method, as well as sample sizes/recruitment method, are presented in Table 3.
Table 3.
Costs by Recruitment Method among Randomized Sample, N=161
| Targeted Mailings | Paid Print | Non-Print | Brochure/Flyer | Social Media/Email | |
|---|---|---|---|---|---|
| Number Randomized (%) | 142(88%) | 2(1.2%) | 8(5.0%) | 1(0.06%) | 8(5.0%) |
| Cost/Randomized Participant, dollars | 207 | 7573 | 1793 | 5390 | 0 |
N=161 randomized participants. Costs were calculated as total costs divided by number of randomized participants.
Discussion
The goals of this study were to examine the yield of various recruitment approaches in enrolling and retaining breast cancer survivors into the randomized controlled trial focusing on the adoption and maintenance of physical activity. We also were interested in tracking the costs of recruitment and the cost per randomized participant. Finally, recognizing that different recruitment methods may vary in their yield, their costs and the sample characteristics, our goals were to explore whether or not there were differences in socio-demographic characteristics of participants who were enrolled the trial and those who were retained at the end of 12 months. We found that targeted mailings were the most effective approach in the overall yield of screened participants and in the proportion of screened participants who transitioned to become randomized participants. Within the targeted mailings we used, a state cancer registry was the most productive source with the caveat that they were the most likely to drop out before the study ended. We invested $58,000 across the various recruitment approaches and found that the targeted mailings were the least expensive -costing about $207 per randomized participant. Finally, while we did not find any statistically significant differences in socio-demographic and cancer-related variables of the women recruited via the six recruitment categories; there were statistically significant differences within the targeted mailings with women of higher education and higher incomes recruited from cancer registries and women with clerical positions recruited from organizational mailings. Unbalanced sample sizes across the recruitment methods may have contributed to these findings.
Targeted mailings especially from cancer registries contributed heavily to our success in meeting accrual goals. Other trials that used cancer registries reported a 5.7%recuitment rate [7] to 11% response rate (home-based multi-behavior intervention for cancer survivors [8]. A trial offering an intervention targeting weight loss and Metformin enrolled 333 breast cancer survivors (goal of 340 women) found that mass mailings from the California Cancer Registry was the most successful recruitment strategy: over 25,000 letters were sent at a cost of $191 per randomized participant [3]. Mailings to women on registry lists resulted in the randomization of 65% of study participants and from other recruitment methods (e.g., referrals and flyers) the yield ranged from 7 to 11%. Although the registries (State and UC San Diego registries) were critical for meeting recruitment goal, the yield was very low with 1% of mailings resulted in a randomization. In that study, physician referrals and community recruitment yielded the highest percentage of screened women who were eligible and then randomized (22% and 31% respectively): these were less expensive, but they involved staff time and the total yield was low and less predictable than registry lists. As others have found, our study team met recruitment goals largely through targeted mailings which though they involved staff time, allowed for the research team to control the timing of the mailings and the pace so that screening could be done efficiently and potential participants would not have wait to be screened. Unlike prior work [9, 10], we did not engage in intensive in-person recruitment in hospitals or oncology practices and physician referrals.
Generally, younger, white survivors and those closer to diagnosis are more likely to enroll when registries are the recruitment source [7, 11, 12]. We did not find that to be the case in our trial. Our results are more similar to a study recruiting colon cancer survivors for an exercise intervention in which 1433 letters were mailed using a population-based state cancer registry and 39 patients were randomized into the study protocol. Apart from age-older patients were less likely to inquire about the study and were less likely to be randomized, no other demographic, clinical or geographic characteristic was associated with interest in the study or randomization [12].
In our prior work using peer mentors to deliver the intervention to breast cancer survivors, we had relied heavily on participant recruitment facilitated by the ACS [4,5]. Although others have found that recruitment collaborations with ACS were effective in that people on their data-bases have received prior services and likely developed trust of the organization and may be willing to participate in research endorsed by the organization, we have to recognize that there were several steps required before randomization into our study (obtaining physician consent, wearing the accelerometer (for 7 days) that may have attenuated the effectiveness of this recruitment source.
Within the targeted mailings category, we had also partnered with the Army of Women – a resource that has been used in other trials-their participants have been found more likely to be college graduates, white, with lower percentage of incomes <$30K [13]. The Army of Women chiefly provides email blasts to their research registry. It should be noted that people volunteer for the registry (that includes healthy volunteers) and hence, the populations reached are likely to be more motivated. For these reasons, it has been recommended that investigators may need to reach out to hospital-based or population-based cancer registries [13]. As previously described, we did recruit from a state cancer registry and the Georgia SEER and hospital-based registries but our sample is largely white (81.4%) and well-educated (67.7% with a college degree).
A recent Cochrane review on recruitment [2] concluded that telephone reminders to non-responders, the use of opt out and open designs (participants are informed about the treatment they are receiving in the trial) as effective strategies for recruitment into RCTs. We used opt-in rather than opt out (opt out approaches could be controversial) as well as telephone reminders (a minimum of 5 follow-up calls at different times/day, as well as email reminders when addresses were available) and our trial was open as interventions were clearly described prior to the telephone screen and at several points during the recruitment phase (and not limited to the informed consent document). The Cochrane review results provided “probable” support for the use of a financial incentive for recruitment. We did not offer financial incentives for recruitment, but participants once enrolled were given a total of $125 ($25 per assessment time-point, 5 time-points) for completing assessments in our trial.
We did not find that social media and free internet sources were effective in recruitment of our sample. Some investigators have used traditional (e.g., mailings, flyer distribution) and free internet sources and found them to have lower costs per participant but that the approaches had limited reach and targeting ability and relatively slow rates of recruitment. Conversely, advertisements through FB and Google had a much broader reach, with greater ability to target certain populations, demonstrated faster rates of recruitment and required low levels of ongoing effort [14]. These results were from a web-based trial targeting smokers. It appears that a multimodal recruitment plan is needed to increase sample diversity and investigators will need to continuously monitor participant enrollment to ensure that the sample has the characteristics needed to answer the research questions.
We tried to reach participants using a variety of approaches but recognize that we had select inclusion criteria and several steps prior to randomization: these features may have reduced the study’s attractiveness to survivors who may not be intrinsically motivated to participate in such research. In considering recommendations for improving efficiency and yield of recruitment, others have recommended setting up community and patient advisory boards, developing strong collaborative relationship with community physicians, and addressing barriers to participate including cultural and health literacy barriers [13]. We carefully monitored monthly randomization rates and continued to brainstorm development of new approaches and re-approached recruitment partners (e.g., ACS) to meet our accrual goals. Clearly there were some recruitment approaches that were both expensive and yielded little or no return (paid print and non-print advertisements such as radio). Hence, investigators will need to carefully budget and balance recruitment expenses and staff time to achieve accrual goals. Targeted mailings when conducted by the research team offered the advantage of control by the team but did involve considerable staff time as well as time donated by student volunteers who assisted with the large mailings.
Investigators have also recommended the use of qualitative approaches to assist with the recruitment process and to investigate reasons for non-participation [15]. Others have developed frameworks to improve the process of recruitment to RCTs (e.g., the SEAR approach [16], Screened=to identify potentially eligible participants, Eligible=assessed against the inclusion/exclusion criteria; Approached=the provision of oral and written information and invitation to participate in the trial and Randomized=into the trial or not). Some studies have used a patient-centered approach for recruitment and retention which implies continuous engagement of stakeholders to identify the resources available and develop strategies to address the challenges to recruitment and retention [17].
The strengths of our study lie in tracking costs associated with the numerous recruitment approaches that were utilized and obtaining data on response rate, eligibility rate, randomization yield and retention of participants recruited via these approaches, over 12 months. The study limitations include our inability to obtain demographic and other key information from non-responders, unequal cell sizes (recruitment methods were unbalanced in terms of number of eligible participants) and the specific eligibility requirements for this randomized controlled trial that could limit generalizability of results.
To conclude, we found that targeted mailings were the most effective approach to recruitment of breast cancers for a randomized controlled trial. We used several types of targeted mailings but of these, cancer registries produced the largest yield for the lowest costs. However, there was greater attrition among these participants over the 12-month trial. We were not able to track staff/personnel time involved with recruitment but it is important that investigators budget adequately for staff time and expenses associated with recruitment. In addition, given that our sample had specific eligibility requirements that needed to be met before randomization, the respondent burden may have been high for our trial. For similar trials among survivor samples with specific characteristics and eligibility requirements, investigators should consider investing in a variety of approaches while balancing costs, monitor accrual yields and adjust investments accordingly.
Acknowledgments
The authors acknowledge the assistance of the American Cancer Society, cancer registries and the organizations that facilitated participant recruitment. The contributions of the research staff are also acknowledged.
This study was funded by the National Cancer Institute (R01 CA 183849, 2015).
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- [1].Sully BG, Julious SA & Nicholl J(2013) A reinvestigation of recruitment to randomized, controlled, multicenter trials: A review of trials funded by two UK funding agencies. Trials, 14, 166–167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Treweek S, Pikethly M, Cook J, Fraser C, Mitchell E, Sullivan F. et al. (2018). Cochrane Database Systematic Review, 2018 (2): MR000013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Patterson RE, Marinac CR, Natarajan L, Hartman SJ, Cadmus-Bertram L, Flatt SW et al. (2016). Recruitment strategies, design, and participant characteristics in a trial of weight-loss and metformin in breast cancer survivors. Contemporary Clinical Trials, 47, 64–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Pinto BM, Rabin C, Abdow S. & Papandonatos GD (2008). A pilot study on disseminating physical activity promotion among cancer survivors: A brief report. Psycho-Oncology, 17, 517–21. [DOI] [PubMed] [Google Scholar]
- [5].Pinto BM, Stein KS & Dunsiger SI (2015). Peers promoting physical activity among breast cancer survivors: A randomized controlled trial. Health Psychology, 34, 463–472. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Sallis JF, Haskell WL, Wood PD, et al. Physical activity assessment methodology in the Five-City Project. American journal of epidemiology. Jan 1985;121(1):91–106. [DOI] [PubMed] [Google Scholar]
- [7].Adams RN, Mosher CE, Blair CK et al. (2015). Cancer survivors’ uptake and adherence in diet and exercise intervention trials: An integrative data analysis. Cancer, 121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Snyder DC, Morey MC, Sloane R. et al. (2009). Reach out to ENhancE wellness in older cancer survivors (RENEW): design, methods, and recruitment challenges of a home-based exercise and diet intervention to improve physical function among long-term survivors of breast, prostate and colorectal cancer. Psycho-Oncology, 18(4): 429–439. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Pinto BM, Trunzo JJ, Rabin C, Cady B, Fenton M, Herman A, Legare R. & Sikov W. (2004). Recruitment strategies for a home-based physical activity intervention for breast cancer patients. Journal of Clinical Psychology in Medical Settings, 11, 171–178. [Google Scholar]
- [10].Zhou E, Dunsiger SI & Pinto BM (2013). Proactive versus reactive recruitment in a physical activity intervention for breast cancer survivors: Does it matter? Clinical Trials, 10, 587–592. [DOI] [PubMed] [Google Scholar]
- [11].Millar MM, Kinney AY, Camp NJ, Cannon-Albright LA, Hashibe M, Penson DF et al. (2019). Predictors of response outcomes for research recruitment through a central cancer registry: Evidence from 17 recruitment efforts for population-based studies. American Journal of Epidemiology, 188, 5. 928–939. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Brown JC, Troxel AB, Ky B, Damjanov N, Zemel BS, Rickles MR et al. (2016). A randomized phase II dose-response exercise trial among colon cancer survivors: Purpose, study design, methods and recruitment results. Contemporary Clinical Trials, 4, 366–375. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Stanton AL, Morra ME, Miller SM, Diefenbach MA, Slavin-Peroccia R, Racih PC et al. (2013). Responding to a significant recruitment challenge within three nationwide psycho-educational trials for cancer patients. Journal of Cancer Survivorship, 7, 392–403. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Watson N, Mull K, Heffner JL, McClure JB & Bricker JB (2018). Participant recruitment and retention in remote eHealth intervention trials: Methods and lessons learned from a large randomized controlled trial of two web-based smoking interventions. J Med Internet Res. 20 (8): e10351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Elliott D, Husbands S, Hamdy FC, Holmberg L. & Donovan L. (2017). Understanding and improving recruitment to randomized controlled trials. European Urology, 72 (5), 789–798 [DOI] [PubMed] [Google Scholar]
- [16].Wilson C, Rooshenas L, Paramasivan S, Elliott D, Jepson M. et al. (2018). Development of a framework to improve the process of recruitment to randomized controlled trials (RCTs): the SEAR (Screened, Eligible, Approached, Randomized) framework. Trials:50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Chhatre S, Jefferson A, Cook R, Meeker CR, Kim J, Hartz KM et al. (2018). Patient-centered recruitment and retention for a randomized controlled study. Trials, 19: 205. [DOI] [PMC free article] [PubMed] [Google Scholar]