Chapter  122:  Knowledge and Access to Information on Recruitment of Underrepresented Populations to Cancer Clinical Trials

Key Question 1: Methods to Study Recruitment Strategies

We analyzed 14 articles to identify methods (e.g., survey studies, focus groups) that have been used to study strategies to recruit underrepresented populations into cancer prevention and treatment trials. ⁵–¹⁸

• All 14 studies were of U.S. origin, primarily based in community settings, and targeting patients/participants.

• The reported study designs of the 14 studies varied, including descriptive (n = 4), randomized controlled trials (n = 3), quasi-experimental (n = 1), comparisons of two or more interventions (n = 2), survey (n = 1), qualitative (n = 1), case study (n = 1), and other (n = 1).

• There was substantial variability across the studies in the reporting of demographic variables such as age, gender, income or education levels of participants; information regarding the racial or ethnic distributions of the participants was available for only eight of these studies.

• The reporting of the studies limited our ability to accurately categorize age groups (e.g., adolescents, elderly), socioeconomic status, or residence (urban versus rural).

Overall, the evidence indicated the need for greater consistency in the reporting of target population characteristics, so that key findings may be considered in relation to specific populations. This would make it feasible, when the sample size is adequate, to conduct subgroup analyses to assess whether barriers to recruitment vary by sociodemographic and cultural factors.

Key Question 2: Measures of Success

We sought to identify what measures of success (e.g., proportional representation relative to the U.S. population, or proportional representation relative to incidence in a specified population) have been used to evaluate the efficacy and/or effectiveness of strategies for recruitment of underrepresented populations into cancer prevention and treatment trials.

• All studies (n = 23) were from the U.S. and 22 studies targeted patients/participants for the recruitment intervention; and over 50 percent of the studies were based on multi-center cancer clinical trials conducted in community settings (n = 9) or hospital centers (n = 7). ⁵–⁹, ¹¹, ¹³, ¹⁶, ¹⁷, ¹⁹–³²

• Most of the reports were based on retrospective review of enrollment to a single or multiple cancer trials.

• Only two articles reported having a recruitment goal for the underrepresented group prior to enrollment in the study. The majority of studies either defined recruitment success as equaling the proportion of underrepresented selected by the researcher (n = 13) for various reasons or as the disease-specific proportion of underrepresented (n = 9). The rest of the studies defined recruitment success as equaling the geographic proportion of underrepresented (n = 2), or the local research institution's proportion of the underrepresented (n = 1).

• Very few studies evaluated recruitment success in underrepresented groups especially those with low socioeconomic status, Asian/Pacific Islanders, adolescents, and rural populations (less than three studies in each group). No study reported recruitment success measures for American Indians/Alaska Natives.

The evidence reviewed indicated that success in recruitment of underrepresented populations is defined primarily by the goal of each study. When reporting on cancer trials, investigators should give careful thought to success measures for recruitment of underrepresented populations, avoid setting such measures arbitrarily, and report recruitment results based on the recruitment strategies for individual cancer clinical trials.

Key Questions 3 and 4: Methods to Study Recruitment Strategies

We sought to identify recruitment strategies (e.g., media appeals, incentives, etc.) that have been shown to be efficacious and/or effective in increasing participation of underrepresented populations in cancer treatment and prevention trials. We found a total of five eligible articles. ⁶, ⁷, ⁹, ¹¹, ¹⁷

The results of the interventions varied from no observed improvement to an increase in recruitment into cancer clinical trials. Two studies examined enrollment differences between two different intervention methods. Two other studies compared enrollment differences between interventions to a control group. These control groups were either no intervention (usual medical care from physicians) or a standard recruitment “intervention” of mailed letters and telephone contact. However, whether various interventions had a true effect (null, positive, or negative) was somewhat unclear. Some authors cautioned that their results could be due to factors such as changes in recruitment strategy during the duration of the intervention. To give a clearer picture, each of the five studies is discussed in detail.

Linnan et al. investigated the differences between passive and active recruitment into a home-based cancer prevention randomized trial among employees.¹⁷ In the passive employee contact arm, the research team contacted the employees from a list of employee names and telephone numbers provided by the company. In the active employee contact arm, employees actively signed up to participate. While lower enrollment and higher attrition were observed in the passive recruitment arm, the passive method enrolled a more diverse group of participants than did the active recruitment method.

Brewster et al. examined differences in recruitment into cancer prevention clinical trials between a clinic registry method and a media campaign targeting Latina women.⁶ In the media recruitment strategy, the study was advertised in flyers placed in local community businesses, and advertised in community and regional newspapers in English and Spanish. The odds of presenting to the clinic and of recruitment were nearly three times more successful via the media campaign than via the clinic registry.

Paskett et al. examined the effect of an intervention program aimed at physicians and the community to increase the number of rural patients with breast cancer or colorectal cancer in clinical trials.⁷ The intervention program consisted of the installation of a rapid tumor-reporting system to improve data quality and to expedite the receipt of information on cancer patients from physicians, a nurse facilitator who would notify physicians of clinical trials, a quarterly newsletter mailed to physicians about cancer treatment and clinical trials, and a health educator who trained lay health educators and provided community-based information about cancer screening, treatment, and clinical trials. Five counties in North Carolina received an intervention program while five counties in South Carolina served as controls where usual medical care was practiced. The rates of enrollment into clinical treatment trials did not improve significantly in the intervention communities.

Moinpour et al. reported the results of a randomized trial in increasing participation of minorities.⁹ Minority recruitment strategies were designed and implemented in five pilot sites: African Americans in four sites and Hispanics in one site. While each site had a minority recruiter who was given requirements and a set of tasks, the specific details of the minority recruitment interventions for each site were not given. The overall impact was minimal, and it was unclear if, and at which site the intervention was fully implemented.

Ford et al. examined recruitment differences among African Americans randomized into either a control group or three increasingly intensive intervention arms.⁷ The control group used a standard method of recruitment such as a standard recruitment letter, African-American or Caucasian interviewers for eligibility screening, baseline information collection via mailed packets, and reminder phone calls and mailings for completion of the mailed packets (Arm D). The basic intervention arm (Arm A) attempted to reduce potential sociocultural and individual barriers through the use of an enhanced recruitment letter and eligibility screening by African-American interviewers. The second more intensive intervention arm (Arm B) did not use mailed packets for baseline information collection but telephone interviews to facilitate ease of participation in addition to the enhanced recruitment letter. The third, and most intensive, intervention arm (Arm C) did not use a mailing packet or telephone interview but a church-based project site to gather baseline information in addition to the enhanced recruitment letter and eligibility screening telephone calls by an African American. The authors reported significantly higher enrollment yield (3.9 percent) in the most intensive church-based, face-to-face recruitment intervention arm (Arm C), compared to the other two intervention arms (2.5 percent [Arm A] and 2.8 percent [Arm B]) or the control group (2.9 percent [Arm D]) (p < 0.01).

There is only scant evidence in support of specific interventions to improve recruitment to cancer clinical trials, as indicated by the small number of studies comparing interventions.

Key Question 5: Barriers and Promoters of Recruitment

We sought to identify the documented barriers and promoters of participation for underrepresented populations in cancer prevention and treatment trials. Our search yielded 45 eligible studies that were conducted in a variety of settings.³, ⁵, ⁸, ¹³, ²², ²³, ²⁵, ²⁶, ²⁹, ³³–⁶³ Among the underrepresented populations, the available studies targeted African Americans primarily (n = 27), as well as Latinos/Hispanics (n = 7); American Indian/Alaska Native (n = 4 ); the elderly (n = 14); adolescents (n = 3);rural populations (n = 2); and Asian/Pacific Islanders (n = 2). While a large proportion of the available studies included populations with low socioeconomic status, only one did so by design.⁹ The search strategy yielded 40 U.S.-based studies, and we included evidence from 5 non-U.S.-based studies that featured relevant evidence.²²

Barriers and promoters of participation in cancer prevention and treatment trials

Types of barriers and promoters identified. Overall, the eligible studies identified 118 distinct barriers to accrual to cancer clinical trials, including 97 barriers to accrual to therapeutic trials, 18 barriers to accrual to prevention trials, and 32 barriers to accrual to both therapeutic and prevention trials. There were more reported barriers to opportunity (n = 80) than to awareness (n = 7) or acceptance (n = 40) of clinical trials. Of the 59 distinct promoters of enrollment, most (n = 29) were promoters of the opportunity to participate in a cancer trial.

Barriers and promoters of accrual of African Americans to cancer treatment trials. Overall, there were 19 studies of accrual of African Americans to cancer therapeutic trials, which reported 85 barriers to accrual to therapeutic trials, including barriers to opportunity (n = 56), barriers to acceptance (n = 28), and awareness (n = 6). Of the 28 barriers to acceptance, the most frequently reported were perceived harms of clinical trial participation (n = 8), mistrust of research, researchers, and the medical system (n = 10), and fear (n = 5). Promoters were predominantly of promoters of awareness (n = 6). Of the reported 14 promoters of opportunity, the most frequently reported were culturally relevant education about trials (n = 3), and providing transportation (n = 2). Of the 14 promoters of acceptance, the most frequently reported were altruism (n = 3), perceived benefits of trial participation (n = 5), and incentives (n = 5).

Barriers and promoters of accrual to therapeutic trials in other underrepresented populations. Latinos/Hispanics. Four studies reported evidence on barriers to accrual of Latinos/Hispanics to cancer therapeutic trials. The reported eight barriers to opportunity were dominated by transportation (n = 2), age (n = 1), toxicity of treatment (n = 1), comorbid conditions (n = 1), and disease stage (n = 1). Of the seven barriers to acceptance, the most frequently reported was mistrust of research and the medical system (n = 2). Only two of the eligible studies for this question reported evidence on promoters of enrollment of Latinos/Hispanics. Brewster and colleagues found that a media-based strategy was superior to a clinic based strategy in recruiting Latino-Hispanic women.⁶ Others have reported the lack of adequate health insurance, incentives, culturally relevant education about trials and the perceived benefits of trial participation as additional promoters of accrual for Latinos/Hispanics.⁵, ⁵⁵

American Indians/Alaska Natives. The amount of evidence available for the American Indians/Alaska Natives population with regard to accrual to clinical trials, in general, was very limited. The aggregate number of American Indians/Alaska Natives in all of the eligible studies for which data on population subgroups was reported, was 19.³⁵, ⁴¹, ⁴³, ⁵⁵

Asian and Pacific Islander. We did not find any evidence regarding barriers or promoters of participation in cancer prevention or treatment trial for the Asian and Pacific Islander population.

The elderly. In the 11 available studies, barriers and promoters of opportunity were predominant in this population. Of the 22 barriers to opportunity, the most frequently reported were age (n = 2).

Adolescents. Only two of the available studies yielded evidence, and reported the lack of available trials as a significant barrier to enrollment of adolescents. Promoters of participation for this population included the perceived benefits of trial participation, including a chance for better treatment, and altruistic motives.

Rural populations. Only two of the available studies focused on recruitment of rural populations to cancer clinical trials, including cross-sectional surveys of physicians,¹¹ and focus groups. The studies reported numerous barriers to awareness, opportunity and acceptance of trial participation. They also reported altruism and incentives (financial and otherwise) as promoters.⁴³

Barriers and promoters of accrual to prevention trials in African-American populations. Overall, there were 13 studies of barriers and promoters of accrual of African Americans to cancer prevention trials. We did not include studies of accrual to other types of primary prevention trials (e.g., diet and exercise) in this systematic review. Among the 41 barriers to accrual to prevention trials, barriers to opportunity (n = 24) were predominant, and of the 13 barriers to acceptance, mistrust of research and the medical system (n = 8), and the perceived harms of clinical trial participation (n = 4) were the most frequently reported. Promoters included provision of transportation (n = 1) and incentives (n = 2).

Chemoprevention trials. ⁹, ³⁷, ⁵⁶ On average, each of the chemoprevention trials reported two barriers (range: 1 to 2). There were no barriers to awareness, two barriers to opportunity, and three barriers to acceptance with mistrust of research reported in two studies. In one study, promoters included preference for the study's principal investigator to be black, and the perception that it is better to be treated by research doctors.

Smoking cessation trials.⁵, ⁵⁶, ⁶⁰ Out of the three smoking cessation trials in African-American populations, only one trial reported barriers to accrual, and not being ready to quit may have been a confounding factor. The reported promoters were incentives, support, encouragement, prayer, the certainty of receiving actual medication, and the impact of diagnosis on risk perception.

Screening trials. ⁷ The results of this one study were discussed in detail under key questions 3 and 4.

Barriers and promoters of accrual to prevention trials among other underrepresented populations.

Latinos/Hispanics. Overall, there were four studies reporting primarily barriers to opportunity (n = 7), especially transportation (n = 2). Mistrust of research and the medical system (n = 2) and family considerations or issues (n = 2) were the most frequent barriers to acceptance.

American Indians/Alaska Natives and Asian and Pacific Islanders. As discussed in the section on accrual to therapeutic trials, very little information is available on these two populations.

The elderly. Overall, there were three studies of barriers and promoters of enrollment to cancer prevention trials in the elderly. These studies reported three distinct barriers, age being the most frequently cited (n = 2). Among the three promoters of accrual to prevention trials, there were no promoters of awareness or acceptance. The three reported promoters of accrual were the entry criteria, age and low-income status.

Rural populations. The available evidence on barriers and promoters of accrual of rural populations to cancer prevention and treatment trials is based on two studies that we discussed in the section on barriers and promoters of accrual to therapeutic trials.¹⁰, ⁴³

Key Question 5a: Effects of Demographic Factors

Overall, the available evidence for key question 5 suggested that accrual to or intention to participate in a trial varied by the following sociodemographic factors: age (n = 16); gender (n = 3); socioeconomic status (n = 4); and race/ethnicity (n = 4). These barriers and promoters related most frequently to study design barriers, including exclusion by age (n = 6), study duration and visit structure (n = 4), comorbid conditions (n = 7), and functional status (n = 6). Few trials were available for adolescents; and as expected, parental influence was reported as a factor in decision-making only in this population. However, the available evidence did not suggest age as a factor that reduced awareness or acceptance of participation.

Key Question 5b: Effects of Cultural Factors

Three of the studies reported that barriers or promoters of enrollment varied by cultural factors,³⁹, ⁵⁵, ⁶⁰ however, it is not entirely clear whether such “cultural factors” refer to cultural norms, values or beliefs. For the elderly population, enrollment barriers and promoters did not vary by culturally relevant factors other than race or ethnicity. The heterogeneity of the available evidence and the definitional overlap among several of the underrepresented populations limited our ability to synthesize the evidence regarding whether some barriers or promoters vary by cultural factors.

Key Question 6: Role of Provider Attitudes and Perceptions

Nine studies presented data on how provider attitudes/perceptions were barriers to and promoters of accrual to cancer clinical trials. Four studies found provider atttitudes as a barrier to enrollment¹¹, ²³, ⁴⁴, ⁶⁴ while one study found provider attitudes to be a promoter of patient accrual.²³ The studies also reported that patient age,²³, ⁵⁹, ⁶⁵ comorbidity,²³, ⁵⁹ disease stage,²³ mistrust of researchers,²³, ³¹ and lack of physician awareness about trials⁴⁴, ⁵² were factors that prevented providers from enrolling their patients into clincial trials. Two studies⁶⁴, ⁶⁶ found that provider communication or method of presentation were barriers to patient enrollment, whereas one study found it to be a promoter of trial enrollment.⁴¹

For studies that targeted minority populations,²⁹, ⁴⁴, ⁵² mistrust of researchers and lack of provider awareness about trials were leading provider barriers ⁴⁴, ⁵² that decreased patient enrollment in clinical trials. Additionally, concerns about patient noncompliance and a lack of available protocols were reasons cited for not talking to patients about clinical trials.²⁹ For studies that targeted the elderly, provider attitudes regarding clinical trials prevented them from sharing information about trials with their patients in one study,²³ and increased their willingness to enroll patients in clinical trials in another study. ⁴¹

Research Quality

Since the enactment of the National Institutes of Health (NIH) Revitalization Act in 1993,⁶⁷ cancer researchers have put increased emphasis on recruitment of underrepresented populations to clinical trials. However, this aspect of the human research enterprise has received attention primarily in the secondary analysis of ongoing clinical trials, rather than as an area of focused scholarship. This reality is clearly reflected in the quality of studies available for this evidence report. One of the positive aspects of the studies available for our review is that they have described a number of barriers and promoters of participation in clinical trials. However, most of the evidence is not based on rigorous studies, and a large proportion of the available studies were not driven by any clear hypotheses. A major weakness of the available evidence is the limited number of studies that compared two or more interventions, especially randomized controlled trials. The quality of the evidence summarized raises some questions about its adequacy to answer our questions regarding barriers and promoters of participation in cancer clinical trials. However, because of the consistency and patterns of occurrence of the identified barriers and promoters, it does provide important insights into future research directions.

Key Questions 1 and 2

• Much of the available body of evidence was developed as “evidence by convenience” in the context of recruitment difficulties, or in retrospective analyses of recruitment of underrepresented populations across multiple clinical trials. There is a need for well-designed, controlled studies of strategies to improve accrual to cancer prevention and treatment trials. These studies should be hypothesis-driven, and include defined measures of success. They should also meet the usual standards of the NIH peer review process.

• Investigators should give careful thought to success measures for recruitment of underrepresented populations, and they should avoid setting such measures arbitrarily. Additionally, researchers should evaluate and report recruitment results for underrepresented groups more consistently.

• More attention should be focused on issues of trial design. If studies are not designed to address problems that are relevant to patients in underserved communities, then even the best recruitment strategies will be ineffective. Similarly, trials that exclude patients with chronic conditions will preferentially exclude the elderly, members of minority groups, and patients with lower socioeconomic status, because they are more likely to have chronic conditions. Hence, recruitment efforts must proceed hand-in-hand with initiatives to design relevant and pragmatic trials.⁶⁸

Key Questions 3, 4, 5, and 6

• Because of many underrepresented populations' mistrust of researchers and of research institutions, research efforts to improve participation of underserved populations in cancer clinical trials should be developed within the framework of community-based participatory research, with community involvement through all phases of the research.

• The need remains for community-based studies to understand barriers to accrual in the community, including attitudes toward clinical trial participation. Whenever possible, such studies should be linked to the implementation of cancer clinical trials, and include actual recruitment as a major outcome. For example, several studies have suggested culturally relevant education as a strategy for improving accrual to cancer clinical trials. There is a need to further investigate the efficacy of culturally relevant education as a strategy to improve accrual to cancer prevention trials and cancer treatment trials.

• There is an urgent need to understand why participation of the Asian American/Pacific Islander and American Indian/Alaska Native populations in cancer clinical trials is minimal to non-existent. Studies of barriers and promoters of their participation should be linked to opportunities to participate. New research initiatives in this area may require several years before they are fruitful in terms of trial enrollment results.

• Similarly, there is a continuing need to better understand and improve upon strategies for recruitment of African-American males and Latinos/Hispanics into cancer clinical trials. Ideally, such studies should include documentation of existing barriers within a population as a basis for tailored interventions across the spectrum of barriers and promoters, including awareness, opportunity and decision-making.

• There is a need for further investigation of effective communication strategies, including investigations on the best approach to deliver information about clinical trials, both at the community level and at the point of interaction with the potential participant.

• In communities lacking established efforts to promote awareness about clinical trials, sufficient time should be allowed for relationships to be built with community members, including community-based providers, before accrual can begin. The period for building such relationships may take several years, but it would vary depending on the community and the existing relationships prior to an intervention.

• Some interventions (e.g., media-based strategy for Hispanic women) have been shown to be effective in increasing accrual to clinical trials. Such interventions should be replicated, and where appropriate, the results should be disseminated widely.

• To advance the evidence regarding efficacious strategies for improving enrollment to cancer clinical trials, intervention studies will need to be linked to one or more clinical trials, depending on sample size requirements. The studies should include collection of baseline information regarding prevalent risk factors in the study population. Systematic data collection about barriers and promoters of trial participation should be linked to concrete plans for designing interventions to address such barriers. Moreover, the next generation of studies of barriers and promoters of accrual should be multidisciplinary, including the involvement of community-based participatory researchers, social and behavioral scientists, as well as health economists.

• There are many barriers to care, and it is unlikely that piecemeal strategies to address these barriers will be effective to promote participation in cancer clinical trials. There is a need for a cost-effective strategy to address barriers to care on multiple levels, and in a manner that can be integrated into the context of the health care system and of the research team. To facilitate the integration of recruitment interventions into health care systems, especially the research team, a study should compare the efficacy of a recruitment intervention specialist to that of usual, opportunistic recruitment practices. The recruitment intervention specialist would be a professional or paraprofessional staff member who is appropriately trained to promote awareness about clinical trials in the community and to help patients overcome barriers to opportunity. Ideally, the recruitment intervention specialist would be indigenous to, or at least have extensive familiarity with, the community targeted by the recruitment effort. Thus, this role would be analogous to that of a patient navigator for clinical trials, and its cost-effectiveness should be investigated.

• Research to improve enrollment of underrepresented populations in cancer clinical trials must interface with other ongoing initiatives designed to address cancer health disparities through discovery, development, and delivery. Such efforts must overcome the critical disconnect between discovery and development on the one hand, and delivery of cancer care on the other.

• Substantial resources will need to be dedicated to research efforts to build upon the existing evidence on strategies for improving enrollment of underrepresented populations in cancer clinical trials. Many of the initiatives that contributed to the available evidence were probably not funded. NCI should dedicate adequate funds for well-designed studies of barriers and promoters of accrual to cancer clinical trials.

Further investigation is needed on barriers to recruitment of all of the underrepresented populations, as defined in this report, into cancer-related clinical trials. The specific populations are: African Americans (especially men), Hispanics, American Indians/Alaska Natives, Asian and Pacific Islanders, adolescents, the elderly, and rural populations. Future studies should include the evaluation of culturally tailored strategies to promote awareness about cancer clinical trials among underrepresented populations. Different types of intervention approaches should be considered to promote accrual to cancer therapeutic trials and cancer prevention trials. Research and evaluation of recruitment strategies may yield stronger evidence about ways to improve participation of underrepresented populations in cancer clinical trials. The principal need is for hypothesis-driven research, and ultimately randomized controlled trials, to evaluate the most promising strategies for recruiting underrepresented populations into cancer treatment and prevention trials.

Sources

Our comprehensive search plan included electronic and hand searching. In March 2004, we searched the following electronic databases: MEDLINE^®, the Cochrane CENTRAL Register of Controlled Trials (Issue 1, 2003), the Cochrane Database of Methodology Reviews (CDMR), the Cumulative Index of Nursing and Allied Health Literature (CINAHL^®), the Psychological Abstracts (PsycINFO), and The Campbell Collaboration's Social, Psychological, Educational, and Criminological Trials Register (C2-SPECTR). This electronic search strategy was not limited by year of publication, and the MEDLINE database goes back to 1966

Hand searching for possibly relevant citations took several forms. Our experts identified 34 journals that were thought to be most likely to contain relevant studies (see Appendix B). We scanned the table of contents of each issue of these journals for relevant citations from January 2003 through July 2004.

For the second form of hand searching, we used ProCite^® (ISI ResearchSoft, Berkeley, CA), a reference management software package, to create a database of reference material identified through an electronic search for relevant guidelines and reviews, through discussions with experts, and through the article review process. The investigators reviewed the articles identified as being possible review articles during the abstract review process. The references in these review articles were searched to identify any additional articles for consideration. We also used MEDLINE to search for articles published by selected experts known to have interests related to our Key Questions.

Finally, we examined the reference lists of eligible articles to identify any potentially relevant ones. This task was completed by the second reviewer as part of the article review process (see description of article review process below).

Search Terms and Strategies

Search strategies, specific to each database, were designed to maximize sensitivity. Initially, we developed a core strategy for MEDLINE, accessed via PubMed, based on an analysis of the Medical Subject Headings (MeSH) and text words of key articles identified a priori. The PubMed strategy formed the basis for the strategies developed for the other electronic databases (see Appendix C).

Organization and Tracking of Literature Search

Whenever possible, the results of the searches were downloaded and imported into ProCite. We used the duplication check in the bibliographic software ProCite to include in the Recruitment Citations Database only articles that were not previously retrieved. This database was used to store citations and to track the search results and sources. We also used this database to track the results of the title and abstract review processes and the retrieval of the full text copies of articles (see Figure 2).

Inclusion and Exclusion Criteria

During the abstract review process, emphasis was placed on identifying all articles that could have original data pertinent to the questions about recruitment of underrepresented populations into cancer prevention and treatment trials. As previously described, representatives from NCI were consulted during the development of inclusion and exclusion criteria.

In evaluating titles and abstracts, the following criteria were used to exclude articles from further consideration:

• Not written in English.

• Did not include human data.

• No original data.

• Meeting abstract only (no full article for review).

• Did not address cancer treatment or prevention.

• Did not report a controlled trial or discuss recruitment to a controlled trial.

• Article did not apply to any of the study questions.

For citations relevant to Key Questions 3 and 4, we limited eligibility further to studies that compared at least two groups. This comparison group could include groups from randomized and non-randomized designs with concurrent or historical controls. Non-controlled designs (e.g., case series with no comparison groups) were excluded for Key Questions 3 and 4, but remained potentially eligible for Key Questions 1 and 2.

When the initial exclusion criteria were developed, “not a U.S.-based study” was an exclusion criterion. As the review process progressed, however, the EPC team determined that this exclusion criterion might delete some articles from the review that could provide important information. Abstract reviews were then conducted on articles initially excluded because they were not U.S.-based studies. A number of articles were then eligible for article review that would not have been eligible otherwise. We included studies that addressed underrepresented minority populations using whatever definitions of race/ethnicity were provided by the study authors.

Abstract Review Process

We reviewed abstracts by printing titles and abstracts of all articles retrieved by the literature search on a standardized form and distributing them to two reviewers (see Appendix D ^*). The reviewers independently screened the abstracts for eligibility and classified them by research question addressed. When reviewers agreed that a decision regarding eligibility could not be made because of insufficient information, the full article was retrieved for review.

The results of the abstract review process were entered into the Recruitment Citations Database. Deleted citations were tagged with the reason for exclusion. Citations were returned to the reviewers for adjudication if there was disagreement on eligibility.

Quality Assessment and Data Abstraction

Forms were developed to confirm eligibility for a full article review, to assess study characteristics, and to extract the relevant data to address the study questions (see Appendix E). The forms were developed through an iterative process that included the review of forms used for previous EPC projects, discussions among team members, and pilot testing. This process was complex because of the heterogeneity of the literature. We developed a general form and question-specific content review forms for the abstraction of data to address Key Questions 1 and 2, 3 and 4, and 5 and 6. We used one form to assess the quality of articles eligible for Key Questions 3 and 4, and a separate form to assess the quality of articles eligible for Key Questions 5 and 6. A quality assessment form was not used for articles eligible for Key Questions 1 and 2 because the questions called for only a listing of methods that have been used. The forms were color coded to aid reviewers.

Article Review Process

We used a serial article review process. In this process, the quality assessment and data abstraction forms were completed by the primary reviewer. The second reviewer, after reading the article, performed an independent assessment of the study's quality using a separate quality assessment form and then checked each item on the data abstraction form for completeness and accuracy. The reviewer pairs were formed to include personnel with both clinical and methodological expertise. A third reviewer re-reviewed all articles that were marked as “ineligible” by the first two reviewers to ensure consistency in the classification of the articles. Reviewers were not masked to the articles' authors, institution, or journal. In most instances, data were directly abstracted from the article. If possible, relevant data were also abstracted from figures.

All information from the article review process was entered in a relational database (Recruitment Evidence Database). The database was used to maintain and clean the data, as well as to create detailed evidence tables (see Appendix F ^†) and summary tables (see Tables 1 to 7).

Identification of Relevant Articles

As indicated in Figure 2, we found 13 studies that were relevant to this question. Details about these studies are shown in Table 1, and Evidence Tables 1-1, 1-2, and 1-3 in Appendix F.^b

Study Characteristics

All 13 studies were of U.S. origin, and most were based in a community setting.⁴¹–⁴⁹ Studies by Advani et al. and Mauer et al.⁵⁰, ⁵¹ were hospital-based, with Advani et al. involving outpatients and Mauer et al. involving inpatients. One study⁵² was set in the workplace or workers' homes. The specific setting for one study was unclear.⁵³ The reported target population for all of the studies was patients/participants except one study⁴⁴ where the target population was patient /participants and physicians. Three studies⁴⁵, ⁴⁶, ⁴⁹ targeted physicians only. The reported study designs in all studies varied. A randomized controlled trial was the reported study design for two studies. ⁴³, ⁵² One study was quasi-experimental, ⁴¹ and two studies compared two or more interventions.⁴², ⁴⁶ Descriptive study designs were used in four studies.⁴⁴, ⁴⁷–⁴⁹ One study used a survey design, ⁵⁰ and another employed a qualitative design.⁵³ One used a case study design.⁵¹

These studies dealt with a range of cancers. A few investigated trials for a specific cancer; for example, breast cancer was specified in three studies, ⁴⁶, ⁴⁷, ⁵⁰ and lung cancer was specified in three.⁴³, ⁴⁸, ⁵² The balance of the studies dealt with multiple types of cancer. Two reported a general focus that included all cancers.⁵³ One investigated trials for any cancer type where there was an active protocol.⁴⁴

Four studies did not report the dates on which recruitment into the study started or ended.⁴⁹, ⁵⁰, ⁵², ⁵³ The recruitment ending date for one of the studies was reported as ongoing.⁴⁵ All of the studies investigated the effectiveness of methods or the delineation of barriers to recruitment of individuals into trials (Evidence Table 1-1, see Appendix F).

Target Population Characteristics

Of the 13 eligible studies, six reported mean age, ⁴², ⁴³, ⁴⁶, ⁴⁷, ⁵², ⁵³ which ranged from 43 years to 62 years (see Evidence Table 1-2 in Appendix F). Within each of these studies the actual age range of participants was quite wide. For example, Brewster et al. reported a within-study range of 17 to 78 years of age.⁴² Five studies reported an age range.⁴², ⁴³, ⁴⁸, ⁴⁹, ⁵³ The reported age range in four of these studies was narrower than it was for Brewster et al.⁴²

Only eight studies reported the gender of the participants.⁴¹–⁴³, ⁴⁶, ⁴⁸, ⁴⁹, ⁵², ⁵³ Four studies reported that the majority of participants were male.⁴³, ⁴⁶, ⁵⁰, ⁵² The studies did not report usable information about the income or education levels of participants.

Information regarding the racial or ethnic distributions of the participants was available for only a few of these studies. Four studies reported that 6 percent, 100 percent, one percent, and 100 percent, respectively, of participants were African American.⁴¹, ⁴³, ⁴⁷, ⁵⁴ Sears et al. reported that 9 percent of participants were Asian/Pacific Islanders, ⁴⁷ and Advani et al. reported that 33 percent of its participants were American Indian.⁵⁰ Sears et al. reported that 85 percent of participants in their study were Caucasian.⁴⁷ In other studies, Advani et al and Randall-David et al. reported 67 percent and one percent of participants as Caucasian.⁵⁰, ⁵³ Brewster et al. and Berman et al. reported that one percent of their participants were Latino/Hispanic.⁴¹, ⁴² None of the studies reported having any adolescent participants. Ford et al., Zhu et al., and Thornquist et al. reported that the proportions of their participants who were elderly were 100 percent, one percent, and 25 percent, respectively.⁴³, ⁴⁹, ⁴⁸ Over one-third (38 percent) of the participants in the study by Ford et al. reported low socioeconomic status.⁴³ Approximately 64 percent of the participants in the study by Advani et al. were from a rural area.⁵⁰ Three other studies reported that one percent of their participants were from rural areas.⁴⁶, ⁵¹, ⁵³

Recruitment Methods Findings

All but two of the studies defined successful recruitment into cancer-related controlled trials as actual participation of the targeted group⁵⁰, ⁵³ (Evidence Table 1-3, see Appendix F), and none of the studies set a priori recruitment goals. Eight of the studies reported recruitment start and end dates, ⁴¹–⁴⁴, ⁴⁶–⁴⁸, ⁵¹ with the average reported time spent on recruitment being 33 months. Advani et al. defined successful recruitment as participants' willingness to participate in the study and not their actual participation.⁵⁰ We defined “methods” broadly as methods used to a) study recruitment, b) design the studies, c) identify the target populations, and d) recruit individuals into the clinical trials. Seven studies did not report the method used to design their recruitment strategies.⁴³, ⁴⁴, ⁴⁷, ⁴⁸, ⁵¹, ⁵²The primary outcomes of the studies varied and could be organized into primary, secondary, or tertiary categories. Linnan et al. and Zhu et al. were primary prevention projects.⁴⁹, ⁵² One study was a survivorship/recurrence-prevention project, ⁴⁵ and two were chemoprevention studies.⁴⁴, ⁴⁸ Each of the 13 studies initially mailed letters to recruit participants as part of the overall recruitment strategy. In addition to these mailings, telephone calls and fliers or posters strategically placed around the worksite or community were used in an attempt to recruit participants into the studies. One study advertised in community and regional newspapers that served as an important source of information for the area's Spanish speaking population.⁴² These advertisements were printed in both English and Spanish. This study also used the local Spanish language television station, faith-based organizations, mental health facilities, and free clinic facilities in an attempt to get people to participate. One study used third party insurers to enlist participants.⁴⁸ Four studies also enlisted the assistance of physicians to recruit participants into their project. ⁴⁴–⁴⁶, ⁴⁸ Incentives, both monetary and material, were used in Paskett et al.⁴⁵ One study enlisted the aid of local businesspersons to help in the recruitment of participants.⁴³ Finally, none of the studies reported having set a goal for recruiting a specific percentage of participants before the start of the study.

Identification of Relevant Articles

As indicated in Figure 2, we found 23 studies that were relevant to this question (see Table 2). Details about these studies are shown in Evidence Tables 2-1, 2-2, and 2-3, Appendix F.

Study Characteristics

All studies except one were from the United States.⁵⁴ Twelve of the studies were reviews of multiple cancer clinical trials and, therefore, had multiple study settings.⁴⁴, ⁵⁵–⁶⁵ Nine of the studies occurred in a community setting⁴¹–⁴⁴, ⁴⁶, ⁴⁸, ⁴⁹, ⁶⁶, ⁶⁷ and seven involved a hospital center⁴², ⁴⁴, ⁵¹, ⁵⁸–⁶⁰, ⁶³. Twenty studies targeted patients/participants for the recruitment intervention⁴¹–⁴⁴, ⁴⁸, ⁵¹, ⁵², ⁵⁴–⁶¹, ⁶³–⁶⁷, two studies targeted physicians, and one study targeted patients/participants and researchers.⁶² Most of the study designs to evaluate recruitment success were descriptive⁴⁸, ⁵⁴–⁶¹, ⁶³, ⁶⁵ presenting recruitment results of a retrospective review of multiple cancer trials (n=11), ⁴⁸, ⁵⁴–⁶¹, ⁶³, ⁶⁵ descriptive studies of enrollment into a cancer clinical trial (n=4), ⁴⁴, ⁶², ⁶⁴, ⁶⁷ or a description of enrollment into a cancer trial using a survey, qualitative, or quasi-experimental study design (n=2). The rest of the study designs were two controlled clinical trials, ⁴², ⁴⁶ two randomized controlled trials, ⁴³, ⁵² a case-control study, ⁶⁶ and a case series.⁴⁶

Nine studies evaluated recruitment into any cancer trial⁴⁴, ⁵⁴–⁶¹; the remaining studies evaluated recruitment success for trials on breast cancer (n=6), ⁴⁶, ⁴⁹, ⁵¹, ⁶³–⁶⁵ colorectal cancer (n=5), ⁴³, ⁴⁶, ⁵¹, ⁶³, ⁶⁵ prostate cancer (n=5), ⁴³, ⁵¹, ⁶², ⁶³, ⁶⁵ lung cancer (n=5), ⁴⁸, ⁵¹, ⁵², ⁶³, ⁶⁵ or cervical cancer (n=2).⁴², ⁶⁷ Sixteen of the studies evaluated recruitment or reviewed active trial enrollment during the 1990s.⁴¹–⁴⁴, ⁴⁶, ⁵¹, ⁵⁴, ⁵⁵, ⁵⁶, ⁵⁹, ⁶¹–⁶⁵, ⁶⁷ Three studies reported that recruitment/enrollment began in the 1980s, ⁴⁸, ⁵⁸, ⁶⁰ one in the 1950s⁵⁷; three did not report when recruitment began.⁴⁹, ⁵², ⁶⁶ The studies evaluating recruitment success were cancer treatment trials alone (n=14), ⁴⁴, ⁴⁶, ⁵¹, ⁵⁴, ⁵⁶–⁵⁹, ⁶¹–⁶⁶ cancer prevention trials alone (n=7), ⁴¹–⁴³, ⁴⁸, ⁴⁹, ⁵², ⁶⁷ or involved treatment and prevention trials (n=2)⁵⁵, ⁶⁰ (see Evidence Table 2-1 in Appendix F).

Target Population Characteristics

For the eligible studies reporting mean age, the majority of the participants were middle-aged adults with a mean age between 40 to 60 years old.⁴², ⁴⁶, ⁵², ⁵⁷, ⁶⁴, ⁶⁶ The age range when documented was mostly between 40 and 80 years.⁴², ⁴³, ⁴⁸, ⁴⁹, ⁶², ⁶³ Three studies evaluating cervical cancer, smoking cessation and adolescent participation in clinical trials had a lower mean age from 20 to 40 years.⁴¹, ⁶⁰, ⁶⁷ The gender distribution varied across studies from 0 to 100 percent male. Four studies reported information on low socioeconomic status or low annual income (less than $30,000 per year)⁴², ⁴³, ⁵², ⁴⁹ while four studies reported having greater than 45 percent of participants with a high school education or less.⁴¹, ⁴⁹, ⁵², ⁶⁶

In these studies, we found that the percentage of study subjects who were in underrepresented groups varied widely: elderly (ranged from 15 to 100 percent with the majority ranging from 20 to 50 percent), ⁴³, ⁴⁸, ⁴⁹, ⁵⁴, ⁵⁶, ⁵⁷, ⁶¹, ⁶³, ⁶⁵, ⁶⁶ African Americans (ranged from 4 to 100 percent with the majority between 4 to 30 percent), ⁴¹, ⁴³, ⁴⁹, ⁵⁶, ⁶³, ⁶⁴, ⁶⁷ Latinos/Hispanics (ranged from 3 to 85 percent with the majority ranging from 3 to 6 percent), ⁴¹, ⁴², ⁵², ⁵⁶, ⁵⁸, ⁶³, ⁶⁷ Asian /Pacific Islanders (ranged from 2 to 3 percent), ⁶³, ⁶⁷ American Indians/Alaska Natives (0 to 2 percent), ⁶⁷ rural (100 percent in both articles evaluating rural populations), ⁴⁶, ⁵¹ adolescents (ranged from 21 to 33 percent), ⁶⁰ and low socioeconomic status/low annual income (27 to 91 percent with the majority between 27 to 39 percent)⁴², ⁴³, ⁵¹, ⁵² (see Evidence Table 2-2 in Appendix F).

Recruitment Goals and Measures of Success

All studies defined recruitment as actual participation in a cancer trial. Only two articles reported having a specific recruitment goal for the underrepresented group prior to enrollment in the study. One was a goal for the rural population based on the recruitment results of a prior study, ⁵¹ and the other was a goal for African American men 55 years old or older based on geographic representation (i.e. percent African American elderly men in the U.S.).⁶² The study by Maurer et al on rural populations met this recruitment goal while the Moinpour et al. study on African Americans did not, recruiting 4 percent instead of 8 percent.⁵¹, ⁶²

Recruitment success was defined differently depending on the study. To better evaluate the recruitment success measures, we grouped them into four different categories: 1) geographic proportion underrepresented (i.e., proportion of U.S. population in an underrepresented group); 2) disease-specific proportion underrepresented (i.e., proportion underrepresented with cancer or a specific type of cancer in a geographic area such as the U.S.); 3) proportion underrepresented selected by the researcher for various reasons (e.g., statistical power, convenience, or data from prior studies); and 4) research institution's proportion underrepresented (i.e., proportion of population in the institution that was in an underrepresented group).

Most of the studies defined recruitment success as equaling the proportion underrepresented that was selected by the researcher for various reasons such as convenience (n=13), ⁴²–⁴⁴, ⁴⁶, ⁴⁸, ⁴⁹, ⁵¹, ⁵², ⁵⁴–⁵⁵, ⁵⁷, ⁶⁶, ⁶⁷ or as the disease-specific proportion underrepresented (n=9).⁵⁴–⁵⁶, ⁵⁸, ⁶⁰, ⁶¹, ⁶³–⁶⁵ The rest of the studies defined recruitment success based on the geographic proportion underrepresented (n=2), ⁴¹, ⁶² or the institution's proportion underrepresented at a research institution (n=1).⁵⁹

When examining the 12 articles on cancer treatment trials, recruitment success was mainly defined as being equal to the proportion set by the researcher, ⁴⁴, ⁴⁶, ⁵¹, ⁵⁴, ⁵⁷, ⁵⁹, ⁶⁶ or the U.S. disease-specific proportion underrepresented.⁵⁴, ⁵⁶, ⁵⁸, ⁵⁹, ⁶¹, ⁶³ For the eight articles on cancer prevention trials alone, recruitment success was defined mainly as being equal to the proportion set by the researcher.⁴², ⁴³, ⁴⁸, ⁴⁹, ⁵², ⁶⁷ For the 3 studies that evaluated both prevention and treatment trials, the disease-specific proportion underrepresented was the major success definition used.⁵⁵, ⁵⁸, ⁶⁴

Christian and Trimble in an article that did not meet our strict eligibility criteria (it contained no original data), reviewed data from the cancer therapy evaluation program that reports the NCI cancer clinical trials by race (American Indian/Alaska Native, Asian/Pacific Islander, African American, and Latino/Hispanic). They used the data from the 2000 U.S. census as a comparison, thereby providing another example of geographic representation as a measure of success.⁷³

A priori recruitment goals were rarely specified for underrepresented groups when evaluating recruitment success of cancer clinical trials. Additionally, the trials had no standard measure of recruitment success. The majority of recruitment success measures were defined either as the proportion set by the researcher, often arbitrarily, or the disease-specific proportion of underrepresented. This was true across all underrepresented groups and regardless of whether the study addressed cancer treatment, prevention, or both. Very few studies evaluated recruitment success in underrepresented groups especially those with low socioeconomic status, Asian/Pacific Islanders, adolescents, and rural populations (less than three studies in each group). No study reported recruitment success measures for American Indian/Alaska Natives (see Evidence Table 2-3 in Appendix F).

Identification of Relevant Articles

As indicated in Figure 2, we found 5 studies that were relevant to these questions.⁴², ⁴³, ⁴⁶, ⁵², ⁶² Details about these studies are shown in Tables 3 and 4, and Evidence Tables 3/4-1, 3/4-2, 3/4-2, 3/4-4, and 3/4-5 in Appendix F.

In reviewing the current literature on effective and/or efficacious recruitment strategies in improving participation of underrepresented populations in cancer clinical trials, we found a total of five eligible articles. Four articles examined recruitment strategies of underrepresented populations into cancer prevention trials, ⁴², ⁴³, ⁵², ⁶² and the fifth article examined strategies of recruiting underrepresented populations into cancer treatment trials.⁴⁶

The settings of the studies ranged from participants' homes and community environments⁴², ⁴³, ⁴⁶, ⁵² to community-based clinics and research sites⁴², ⁶² (Evidence Table 3/4-1 see Appendix F). Three studies focused on recruitment of underrepresented populations at the participant level;⁴², ⁴³, ⁵² one study evaluated the recruitment of underrepresented populations through participants and physicians⁶² and one evaluated recruitment of underrepresented population through physicians⁴⁶ (Evidence Tables 3/4-1 and 3/4-2 in Appendix F). The underrepresented populations examined included blue-collar workers in manufacturing companies, African-American and Latino/Hispanic men, Latina/Hispanic women in southern California, and cancer patients in rural counties of North and South Carolina. All five studies used randomized or non-randomized controlled trials to evaluate the efficacy or effectiveness of interventions to promote recruitment to cancer clinical trials (Evidence Table 3/4-3, Appendix F).

Quality of Studies

The studies on Key Questions 3 and 4 varied in their limitations as summarized in Quality Evidence Table 3/4-4 in Appendix F. We specifically examined each article based on representativeness, bias and confounding, recruitment description, outcomes and follow-up, statistical quality and interpretation, and conflict of interest.

Linnan's study reported a good amount of detail on the description of the study setting and participant characteristics, inclusion and exclusion criteria, and recruitment methods. However, the authors did not adequately discuss their statistical analyses, address methods to account for bias and confounding such as randomization or blinding, state any information about definition of successful recruitment or recruitment barriers and promoters, or report their source of funding.⁵²

Brewster's study overall was considered to be of fair quality. The authors described in great detail the setting and population, inclusion and exclusion criteria, key participant characteristics, information about non-participants, and statistical analyses. In addition, the researchers explained differences between the two recruited populations and the recruitment description. However, the authors did not provide adequate information about the definition of successful recruitment or recruitment barriers and promoters, or about their source of funding.⁴²

While Paskett's study provided some details about the setting and population, and reported their source of funding, it was also of fair quality, due to the lack of information in certain areas. The researchers did not provide adequate information on participant characteristics or their recruitment methods, address issues of randomization or blinding, explain their statistical analyses, define recruitment success, or report reasons for withdrawal.⁴⁶

In a good quality study, Ford and colleagues described in great detail the study setting, inclusion and exclusion criteria, participant characteristics, recruitment description, outcomes and follow-up, statistical analyses, and the source of funding. However, the researchers did not explain their randomization methods or how they addressed blinding.⁴³

Moinpour's study was not graded due to the lack of information.⁶²

Results of Studies

The results of the interventions varied from no observed improvement⁴⁶ to an increase in recruitment into cancer clinical trials⁴², ⁴³, ⁵² (Tables 3 and 4, and Evidence Table 3/4-4 in Appendix F). Two studies examined enrollment differences between two or more intervention methods.⁴², ⁵² Two other studies compared enrollment differences between interventions and a control group.⁴³, ⁴⁶ These control groups featured either no intervention⁴⁵ (usual medical care from physicians) or a standard recruitment “intervention” of mailed letters and telephone contact.⁴³ However, whether various interventions had a true effect (null, positive, or negative) was somewhat unclear. Some authors cautioned that their results could be due to factors such as changes in recruitment strategy during the duration of the intervention. To get a clearer picture, we discuss the five studies in detail.

Linnan et al. investigated the differences between passive and active recruitment into a home-based cancer prevention randomized trial among employees.⁵² In the passive employee contact arm, the research team contacted the employees from a list of employee names and telephone numbers provided by the company. In the active employee contact arm, employees actively signed up to participate. The research team examined differences in enrollment, reach, and attrition along with organizational factors and behavioral risk factors between the two recruitment arms. Compared to the active employee contact, the passive employee contact reached a higher proportion of employees (74.5 percent vs. 24.4 percent) and had a higher attrition rate (46 percent vs. 29.9 percent). However, the active employee contact arm enrolled a larger proportion (77.5 percent vs. 40.9 percent, P<0.0001). Besides having a higher proportion of smokers and higher consumption of high-fat diets in passive employee contact arm, no significant differences were observed in other behavioral or organizational factors. While lower enrollment and higher attrition were observed in the passive recruitment arm, the passive method enrolled a more diverse group of participants than did the active recruitment method. The authors concluded that with the differences observed, the results provided insight into the advantages and disadvantages that researchers may encounter when designing and implementing recruitment strategies. They also noted that these different recruitment methods might not necessarily apply to all types of work sites.

Brewster et al. examined differences in recruitment into cancer prevention clinical trials between a clinic registry method and a media campaign targeting Latina/Hispanic women.⁴² In the clinic registry method, all women who had visited clinics as patients or a guardian were identified from the clinic registries. In the media recruitment strategy, the study was advertised in flyers placed in local community businesses, and advertised in community and regional newspapers in English and Spanish. Women who called in response to these announcements were screened for eligibility and subsequently scheduled if eligible. Compared to the clinic registry method, the media recruitment method resulted in a significantly larger proportion of women screened as a result of the media-based strategy were eligible (81 percent vs. 75 percent, p <0.001), and were more likely to give telephone consent to participate (83 percent vs. 65 percent, p <0.001). The odds of presenting to the clinic were three times higher for women recruited by the media campaign than for those recruited via the clinic registry (odds ratio [OR] = 3.00; 95% confidence interval [CI] 2.38 to 3.78). In addition, of the women who were screened by telephone, recruitment was nearly three times more successful via the media campaign than via the clinic registry (OR = 2.97; 95% CI 2.52 to 3.51). The authors concluded that the media campaign method was more advantageous than the clinic registry in recruiting women. The media campaign also recruited a larger number of Latina and uninsured women into the study.

Paskett et al. examined the effect of an intervention program aimed at physicians and community to increase the number of rural patients with breast cancer or colorectal cancer in clinical trials.⁴⁴ The intervention program consisted of the installation of a rapid tumor-reporting system to improve data quality and to expedite the receipt of information on cancer patients to physicians, a nurse facilitator who would notify physicians of clinical trials, a quarterly newsletter mailed to physicians about cancer treatment and clinical trials, and a health educator who trained lay health educators and provided community-based information about cancer screening, treatment, and clinical trials.

Five counties in North Carolina received an intervention program while five counties in South Carolina served as controls where usual medical care was practiced. Data collection on clinical trail enrollment was obtained from medical record data and to evaluate the enrollment differences between the North Carolina counties and South Carolina counties. As noted in Evidence Table 3/4-4, the percentage of breast and colorectal cancer patients from North and South Carolina recruited into cancer clinical trials changed in the five-year span between 1991 and 1996. In North Carolina, 15 percent (n = 24) of breast cancer patients and 4 percent of colorectal cancer patients were enrolled in clinical trials in 1991 while 6 percent (n=14) of breast cancer patients and 5 percent of colorectal cancer patients were enrolled in 1996. In South Carolina, 6 percent (n = 6) of breast cancer patients and 5 percent of colorectal cancer patients were enrolled in 1991 while 50 percent (n = 16) breast cancer patients and no colorectal cancer patients were enrolled in 1996. The authors ultimately concluded that the rates of enrollment into clinical treatment trails did not improve significantly in the intervention communities.

However, the available data remain open to interpretation despite the varied changes in percentage of enrolled patients from 1991 to 1996 in each region. The reported numbers and percentage of enrolled breast cancer patients in North and South Carolina were presented for both 1991 and 1996, but only the percentage of enrolled colorectal cancer patients was given for each area. The number of colorectal cancer patients in North and South Carolina were given as one aggregate number. Thus, without the definitive numbers distinguishing colorectal cancer patients in North Carolina from those in South Carolina, it is unknown whether the reported percentages of enrolled colorectal cancer patients in 1991 and 1996 in both regions indicate an increase or decrease in enrolled patients into cancer clinical trials. In addition, the authors noted incomplete data on breast cancer cases in South Carolina in 1996 were obtained thus dictating caution in making comparisons.

Moinpour et al. reported the results of a randomized trial in increasing participation of minorities.⁶² Minority recruitment strategies were designed and implemented in five pilot sites: African Americans in four sites and Hispanics in one site. While each site had a minority recruiter who was given requirements and a set of tasks, the specific details of the minority recruitment interventions for each site were not given. The overall impact was minimal, according to the percentage of people recruited among the five sites before and after the implementation of the recruitment strategies. Four of the sites reported a decrease in percent enrolled (-0.3 percent, -0.5 percent, -2.8 percent, -0.6 percent) while only one site reported an increase in percent enrolled (+0.5 percent). Statistical significance of these results was not reported. The authors indicated that evaluation of the effectiveness of these strategies was difficult because execution of these interventions occurred near the end of the recruitment period.

Ford et al. examined recruitment differences among African Americans into cancer screening trials who were randomized into either three increasingly intensive intervention arms or a control group.⁴³ The control group utilized a standard method of recruitment such as a standard recruitment letter, African American or Caucasian interviewers for eligibility screening, baseline information collection via mailed packets, and reminder phone calls and mailings for completion of the mailed packets (Arm D). The basic intervention arm (Arm A) attempted to reduce potential sociocultural and individual barriers through the use of an enhanced recruitment letter and eligibility screening by African American-only interviewers. The second increasingly-intensive intervention arm (Arm B) did not use mailed packets for baseline information collection but telephone interviews to facilitate ease of participation in addition to the enhanced recruitment letter. The third, and most intensive, intervention arm (Arm C) did not use a mailing packet or telephone interview but a church-based project site to gather baseline information in addition to the enhanced recruitment letter and eligibility screening telephone call by an African American. The authors reported significantly higher enrollment yield (3.9 percent) in the most intensive church-based intervention arm (Arm C), compared to the other two intervention arms (2.5 percent [Arm A] and 2.8 percent [Arm B]) or the control group (2.9 percent [Arm D]) (p <0.01).

Grading of the Total Body of Evidence

Overall, the mixed results of the interventions mentioned in these studies denote a continuing need for trials that evaluate additional interventions (Table 8). The studies were limited to specific populations in certain locations. Also, very few studies were available to answer our proposed questions. The available information necessitates cautious interpretation due to limitations in the quality of the studies as summarized in Evidence Table 3/4-4 (see Appendix F).

Study Characteristics

Our search yielded 45 eligible studies that were conducted in a variety of settings³², ⁴¹–⁴⁴, ⁴⁶–⁴⁸, ⁵⁰, ⁵³, ⁵⁸, ⁶⁰–⁶², ⁶⁴, ⁶⁶, ⁶⁹–⁹⁷ including in the community (n = 25), ⁴¹–⁴⁴, ⁴⁶–⁴⁸, ⁵³, ⁶⁶, ⁷⁰, ⁷¹, ⁷³, ⁷⁵, ⁷⁷, ⁷⁹, ⁸⁰, ⁸¹, ⁸³, ⁸⁶, ⁸⁸–⁹⁰, ⁹²–⁹⁴ hospital inpatient and outpatient settings (n = 22), ⁴², ⁴⁴, ⁵⁰, ⁵⁸, ⁶⁰, ⁶², ⁶⁴, ⁶⁹, ⁷⁰, ⁷²–⁷⁴, ⁷⁶, ⁷⁸, ⁸⁷, ⁸⁹, ⁸⁵, ⁸⁷, ⁹¹, ⁹⁵, ⁹⁶, ⁹⁷ a study based on recruitment logs of clinical trials groups (n = 1), ⁶² and databases consisting of all NCI-funded clinical trials (n = 2).³², ⁶¹ Most of the studies used patients/research participants as the primary respondents, ³², ⁴¹–⁴⁴, ⁴⁷, ⁴⁸, ⁵⁰, ⁵³, ⁶⁹–⁸¹, ⁸³, ⁸⁵–⁸⁷, ⁸⁹, ⁹⁰, ⁹³, ⁹⁴, ⁹⁶, ⁹⁷ while a few included interviews of healthcare professionals, including physicians, clinical research associates and data managers.⁴⁴, ⁴⁶, ⁶², ⁶⁶, ⁷⁹, ⁸⁰, ⁸⁷, ⁸⁴, ⁸⁸, ⁹¹, ⁹², ⁹⁵ Among the underrepresented populations in cancer clinical trials, the available studies targeted African Americans primarily (n = 27)³², ⁴¹, ⁴³, ⁴⁷, ⁵⁰, ⁵⁸, ⁶², ⁶⁴, ⁶⁶, ⁶⁹–⁷¹, ⁷⁴, ⁷⁵, ⁷⁹–⁸¹, ⁸⁵, ⁸⁶, ⁸⁸–⁹⁴, ⁹⁷ and some of these focused exclusively on women (n = 5)⁴⁷, ⁷³, ⁷⁴, ⁸⁶, ⁹³ or men (n = 4).⁴³, ⁶², ⁸⁵, ⁹⁴ The available studies also targeted other under-represented populations: Latinos / Hispanics (n = 7)³², ⁴¹, ⁴², ⁶², ⁶⁶, ⁷⁴, ⁹³; American Indian/Alaska Natives (n = 2)³², ⁵³, ,⁹³; the elderly (n = 14)³², ⁴³, ⁴⁸, ⁶¹, ⁶⁶, ⁷⁵–⁷⁸, ⁸³, ⁸⁴, ⁸⁷, ⁹⁵, ⁹⁶; adolescents (n = 3)⁶⁰, ⁷², ⁷³; rural populations (n = 2)⁵², ⁵³; and Asian/Pacific Islanders (n = 2).³², ⁹¹ While a large proportion of the available studies included low socioeconomic status (SES) populations, only one did so by design⁴³; and the studies generally reported low SES in association with minority status. The designs of the eligible accrual studies were predominantly descriptive and/or qualitative, but also included case-control studies (n = 2)⁵⁰, ⁶⁶ as well as controlled trials (n = 6)⁴², ⁴³, ⁴⁶, ⁴⁷, ⁶², ⁷⁸ (both randomized and non-randomized). The studies provided evidence on enrollment to therapeutic trials (n = 29), ³², ⁴⁴, ⁴⁶, ⁴⁷, ⁵⁰, ⁵³, ⁵⁸, ⁶⁰, ⁶¹, ⁶⁴, ⁶⁶, ⁶⁹, ⁷⁰, ⁷², ⁷³, ⁷⁵–⁷⁹, ⁸¹–⁸⁴, ⁹¹, ⁹³–⁹⁶ prevention trials (n = 11), ⁴¹–⁴³, ⁴⁸, ⁶², ⁷¹, ⁷⁴, ⁸⁵, ⁸⁶, ⁸⁷, ⁹⁰ both (n = 5).⁸⁰, ⁸⁸, ⁸⁹, ⁹², ⁹⁷

Participation in a therapeutic trial typically involves accrual to a pre-randomization (or run-in) phase, followed by randomized allocation of eligible subjects to the treatment arms. While most of the eligible studies addressed barriers to enrollment, few examined barriers to randomization of those enrolled in clinical trials. This report focuses on barriers and promoters of enrollment (see Evidence Table 5-3 in Appendix F), and it does not address barriers to retention of underrepresented populations in cancer clinical trials. Moreover, most of the available studies did not provide evidence on barriers to participation in early phase versus phase III cancer clinical trials. Therefore, we did not make a distinction between these two types of trials in the review.

Target Population Characteristics

Tables 5 and 6 present selected characteristics of participants in eligible studies. The search strategy yielded 40 U.S.-based studies. In addition, because of the modest number of U.S.-based studies, we also included 5 non-U.S.-based studies⁷⁶, ⁸⁷, ⁸³, ⁸⁷, ⁹⁶ that featured evidence of interest to this systematic review, including a study about the perspective of clinical research associates.⁸⁷ The sample size varied among the studies, from 23 to 59,300, and in two studies, it was not reported. The reporting of age was not consistent across the studies, and it included reports of the mean age (n = 8), ⁴¹, ⁴⁶, ⁴⁷, ⁶⁴, ⁶⁶, ⁷⁴, ⁷⁹, ⁸¹ an age range (n = 18), ³², ⁴⁴, ⁴⁸, ⁶⁰, ⁶⁹, ⁷⁰, ⁷³, ⁷⁵–⁷⁷, ⁸³, ⁸⁴, ⁸⁶, ⁸⁷, ⁹⁰, ⁹¹, ⁹⁵, ⁹⁶ or both (n = 7), ⁴², ⁴³, ⁵³, ⁷¹, ⁷², ⁸⁹, ⁹⁷ however some (n = 12)⁵⁰, ⁵⁸, ⁶¹, ⁶², ⁷⁸, ⁸⁰, ⁸⁷, ⁸⁵, ⁸⁸, ⁹²–⁹⁴ did not provide age information. Similarly, only some studies reported the income levels of the participants (n = 5).⁴², ⁵⁰, ⁷⁵, ⁸⁹, ⁹⁰ The proportion of males ranged from 0 to 100 percent, depending on the type of cancer trial (e.g., cervical cancer vs. prostate cancer) to which recruitment efforts were directed. Few of the studies reported the participation of American Indian/Alaska Natives, Asian/Pacific Islanders, adolescents or rural dwellers. ⁵², ⁵³, ⁶⁰, ⁷², ⁷³, ⁸⁴, ⁹¹, ⁹³ More studies addressed barriers to accrual of African Americans (n = 27) than of any other racial/ethnic group. Among African Americans, there were four studies of barriers to accrual in males, and five in females.

Quality of Studies

The articles were rated in terms of quality in each of five areas or domains: 1) representativeness, 2) justification of study methods, 3) reliability and validity of data collection methods, 4) potential for bias/confounding, and 5) data analysis. Selected aspects of quality are summarized in Evidence Table 5-4 (see Appendix F). In terms of representativeness, ten of 45 studies provided adequate descriptions of study participants.⁴², ⁴⁷, ⁵⁰, ⁶⁰, ⁶⁴, ⁶⁹, ⁷⁰, ⁷⁷, ⁷⁸, ⁸⁷ For justification of study methods, six studies provided excellent documentation for the basis for their studies, ⁴³, ⁶⁹, ⁷⁵, ⁷⁹, ⁸⁴, ⁹³ and 27 other studies provided good documentation.⁴¹, ⁴², ⁴⁶–⁴⁸, ⁵⁰, ⁵⁸, ⁶⁰, ⁶⁴, ⁶⁶, ⁷⁰, ⁷²–⁷⁴, ⁷⁷, ⁸⁰, ⁸¹, ⁸³, ⁸⁵–⁹², ⁹⁷ With the exception of two studies, ⁵⁰, ⁵³ the study instruments used in the surveys and data collection methods used in the qualitative studies did not show adequate reliability and validity. While 14 studies⁴⁶, ⁵⁸, ⁶⁶, ⁷⁰, ⁷⁵, ⁸¹, ⁸⁵, ⁸⁶–⁹¹, ⁹³ made notable efforts to protect against confounding or other types of bias, only four were rated as excellent in this regard.⁵⁸, ⁷⁰, ⁸⁶, ⁹² For studies that used surveys as part of data collection, potential sources of confounding included participant response rates and selection of subjects; for studies that used qualitative techniques to collect data, the selection of subjects was a potential source of confounding. In terms of the description of the data analysis, all but three studies⁴³, ⁴⁸, ⁹⁰ were inadequate in their reporting.

Overall Barriers and Promoters of Participation in Cancer Therapeutic and Prevention Trials

Tables 5 and 6 present a summary of information regarding barriers to participation in cancer prevention and treatment trials (see Evidence Table 5-3 in Appendix F for additional detail).

Types of Barriers Identified. Overall, the eligible studies identified 118 distinct barriers to accrual to cancer clinical trials, including 97 barriers to accrual to therapeutic trials, 18 barriers to accrual to prevention trials, and 32 barriers to accrual to both therapeutic and prevention trials. Across all of the studies, these barriers include eight barriers to awareness of cancer clinical trials, 80 barriers to the opportunity to participate in a cancer trial, and 40 barriers to acceptance of enrollment. Among these, 63 barriers were relevant at the patient level, 31 at the provider level, 20 at the study design level and seven at the healthcare systems level (see Evidence Table 5-3 in Appendix F for additional detail).

Types of Promoters Identified. The eligible studies identified 59 distinct promoters of enrollment to cancer clinical trials, including 36 promoters of enrollment to therapeutic trials, 14 promoters of enrollment to prevention trials, and 17 promoters of enrollment to both therapeutic and prevention trials. Overall, these promoters include six promoters of awareness of cancer clinical trials, 29 promoters of the opportunity to participate in a cancer trial, and 25 promoters of the decision to enroll. We identified promoters at the patient level (n = 40) (see Evidence Table 5-3 in Appendix F for additional detail), at the provider level (n = 12), at the study design level (n = 6), and at the healthcare system level (n = 2) (see Evidence Table 5-3 in Appendix F for additional detail).

Barriers and Promoters of Participation in Cancer Therapeutic and Prevention Trials for African Americans

Barriers to Accrual to Cancer Therapeutic Trials in African American Populations. Overall, there were 19 studies of barriers to accrual of African Americans to cancer therapeutic trials, and on average, each reported eight barriers (range: 0 to 20). Among the 85 barriers to accrual to therapeutic trials, there were six barriers to awareness, and the most frequently reported among them were lack of education about trials (n = 5), ⁵⁰, ⁷⁹, ⁸⁰, ⁹², ⁹³ lack of dissemination of study opportunities to patients/providers (n = 3), ⁸⁸, ⁹², ⁹³ and lack of knowledge about origins of cancer (n = 2).⁷⁶, ⁹⁴ Of the reported 56 barriers to opportunity, the most frequently reported were costs (n = 5), ⁵⁸, ⁷⁹, ⁸⁰, ⁸⁸, ⁹² functional status (n = 2), ⁵⁸, ⁶⁹ time commitment or additional time required (n = 5), ⁴⁷, ⁷⁹, ⁸¹, ⁹³, ⁹⁷ lack of or inadequate health insurance (n = 3), ³², ⁶⁴, ⁸¹ and provider attitudes (n = 4).⁷⁹, ⁹¹, ⁹², ⁹⁴ Of the 28 barriers to acceptance, the most frequently reported were perceived harms of clinical trial participation (n = 8), ⁴⁵, ⁸⁰, ⁸¹, ⁸⁹, ⁹¹, ⁹², ⁹⁴, ⁹⁷ mistrust of research, researchers, and the medical system (n = 10), ⁶², ⁷⁴, ⁷⁹–⁸¹, ⁸⁹–⁹⁴, ⁹⁷ and fear (n = 5).⁸⁸, ⁹¹–⁹⁴

Promoters of Accrual to Therapeutic Trials in African American Populations. Among the 19 studies of accrual of African Americans to cancer therapeutic trials, each reported, on average, two promoters (range: 0 to 8). Among the 34 promoters of accrual to therapeutic trials, there were six promoters of awareness, including education programs for community physicians (n = 1), ⁷⁹ adequate knowledge about study (n = 1), ⁸⁰ and workshop on trials (n = 1).⁸⁰ Of the reported 14 promoters of opportunity, the most frequently reported were culturally relevant education about trials (n = 3), ⁹²–⁹⁴ and providing transportation (n = 2).⁸⁰, ⁹⁷ Of the 14 promoters of acceptance, the most frequently reported were altruism (n = 3), ⁵⁰, ⁸¹, ⁸⁹ perceived benefits of trial participation (n = 5), ⁵⁰, ⁸⁹, ⁹³, ⁹⁴, ⁹⁷ and incentives (n = 5).⁸⁰, ⁸¹, ⁹³, ⁹⁴, ⁹⁷

Barriers and Promoters of Accrual to Therapeutic Trials in Other Underrepresented Populations

Because of the limited evidence available regarding barriers and promoters of participation in cancer therapeutic trials in populations other than African Americans, we discuss these barriers and promoters together within each of the following paragraphs.

Barriers and Promoters in Other Racial and Ethnic Minority Populations.

Latinos/Hispanics. Overall, three studies reported evidence on barriers to accrual of Latinos/Hispanics to cancer therapeutic trials, and on average, each reported seven barriers (range: 0 to 9).³², ⁶⁶, ⁹³ For this population, there were two barriers to awareness, and three barriers to acceptance. The reported 13 barriers to opportunity included age (n = 1), ⁶⁶ toxicity of treatment (n = 1), ⁶⁶ comorbid conditions (n = 1), ⁶⁶ and disease stage (n = 1).⁶⁶ Two studies reported evidence on promoters of enrollment of Latinos/Hispanics to cancer therapeutic trials.³², ⁹³

American Indian/Alaska Natives. The amount of evidence available for the American Indian and Alaska Native population with regard to accrual to clinical trials was very limited. The aggregate number of American Indian/ Alaska Native participants in all of the eligible studies for which data on population subgroups was reported was 15.⁵³, ⁷², ⁷³, ⁹³ Most of the evidence regarding this population comes from focus groups conducted by Roberson (n = 28), which included African American, Latino/Hispanic and American Indian/ Alaska Native (n = 10) respondents, and addressed behavioral intention to participate in both cancer treatment and prevention trials.⁹³ When asked “Why do you think people of your race do not participate in experimental studies?”, the responses included “lack of information”, “do not get involved”, and “mistrust—do not like to be treated as guinea pigs.” All of the participating groups reported “mistrust of white people” as a barrier. In a publication that was technically ineligible for this review, some investigators found that it took several years to build trust and lay the foundation for the conduct of cancer clinical trials with an American Indian/Alaska Native population.¹⁰³ Roberson reported that the American Indian/Alaska Native participants suggested that programs and studies be set up on reservations, and details, including benefits, be explained. They also suggested videos, fliers, free food and money as potential incentives for participation.

Asian/Pacific Islanders. We did not find any evidence regarding barriers or promoters of participation in any type of cancer clinical trial for the Asian/Pacific Islander population. The only available evidence was that this population is under-represented in NCI-funded cancer clinical trials.³²

Barriers and Promoters in the Elderly. Overall, there were 11 studies of barriers and promoters of enrollment of the elderly to cancer therapeutic trials. On average, each of these studies reported four barriers (range: 0 to 9). Among the studies of accrual to therapeutic trials, there were no promoters of awareness, two studies of promoters of opportunity, and six studies of promoters of acceptance. Of the 25 barriers to opportunity, the most frequently reported were age (n = 5), ⁶⁶, ⁷⁷, ⁷⁸, ⁹⁵, ⁹⁶ comorbidity exclusions (n = 3), ⁶¹, ⁷⁶, ⁸⁴ transportation (n = 2), ⁸⁴, ⁹⁵ lack of or inadequate health insurance (n = 1), ⁸⁴ and inability to understand the trial (n = 2).⁸⁴, ⁹⁵ In this population, concern about drug toxicity among both providers and participants (n = 2)⁶⁶, ⁸⁴ was an important barrier to participation, even though such concern may not be substantiated in the literature on the basis of elderly status alone.¹⁰⁴ Interestingly, in a study of protocol exclusion criteria that reduced the participation of the elderly in cancer clinical trials, Lewis and colleagues found that trials that did not specify life expectancy as an exclusion criteria resulted in relatively greater participation of elderly subjects than those that included this exclusion criteria.⁶¹

Barriers and Promoters in Adolescents. Only two of the available studies yielded evidence for our review of barriers to accrual of adolescents to therapeutic trials. Krailo and colleagues documented the lack of available trials as a significant barrier to enrollment of adolescents.⁶⁰ In addition, Broome and colleagues reported that adolescents who enrolled in a trial had traveled long distances and tended to be highly anxious; however, neither distance nor being anxious constituted a barrier to enrollment.⁷² Similarly, financial incentives were neither a barrier nor a promoter of participation. Promoters of participation for this population included the perceived benefits of trial participation, including a chance for better treatment, and altruistic motives. However, some of the study participants had a limited understanding of a randomized controlled trial.

Barriers and Promoters in Rural Populations. Only two of the available studies focused on barriers to accrual of rural populations to cancer clinical trials. Paskett conducted serial cross-sectional surveys for an aggregate sample size of 360 physicians who serve a rural population.⁴⁶ The respondents identified the lack of education about clinical trials, and the lack of physician awareness about trials, as barriers. Additionally, they reported a number of barriers that reduced their opportunity to participate in clinical trials, including the logistical burden of trial participation, transportation, and established referral patterns. Moreover, the barriers to acceptance of trial participation were cost to the patient, perceived treatment risk, mistrust of research and of the medical system, the need to see an unfamiliar physician, and the perceived harms of trial participation. Randall-David conducted focus groups in the community on barriers to participation in prevention and treatment trials.⁵³ This study identified similar barriers and promoters to those reported by Paskett and colleagues, and included the lack of dissemination of study opportunities to providers and patients, the method of presentation of the trial to physicians and patients, and religious belief, as additional barriers. This study also included altruism and incentives (financial and otherwise) as promoters.

Barriers and Promoters of Accrual to Prevention Trials in African-American Populations

Overall, there were 13 studies of barriers and promoters to accrual of African Americans to cancer prevention trials. These included three studies of accrual to a chemoprevention trial, ⁶², ⁷¹, ⁹⁰ baseline data from the African American Men's (AAMEN) study, ⁴³ and three studies of accrual to smoking cessation trials.⁴¹, ⁸⁶, ⁹⁷ However, we did not include studies of accrual to other types of primary prevention trials (e.g., diet and exercise) in this systematic review.

Barriers to Accrual to Prevention Trials in African-American Populations.

On average, each of these studies reported five barriers (range: 0 to 20). Among the 41 barriers to accrual to prevention trials, there were five barriers to awareness, and the most frequently reported among them were lack of education about clinical trials (n = 3)⁷⁴, ⁸⁰, ⁹² and lack of dissemination of study opportunities to patient/provider (n = 2).⁸⁸, ⁹² Of the reported 24 barriers to opportunity, the most frequently reported were costs (n = 3), ⁸⁰, ⁸⁸, ⁹² the study's duration and visit structure (n = 2), ⁸⁸, ⁹² the time commitment (n = 1), ⁹⁷ and lack of transportation (n = 1).⁹⁷ Of the 13 barriers to acceptance, the most frequently reported were fear (n = 3), ⁷⁴, ⁸⁸, ⁹² mistrust of research and medical system (n = 8), ⁶², ⁷⁴, ⁸⁰, ⁸⁸–⁹⁰, ⁹², ⁹⁷ perceived harms of clinical trial participation (n = 4), ⁸⁰, ⁸⁹, ⁹², ⁹⁷ and family considerations (n = 1).⁷⁴

Chemoprevention Trials. ⁶², ⁷¹, ⁹⁰ On average, each of the chemoprevention trials reported two barriers (range: 1 to 2). There were no barriers to awareness, two barriers to opportunity, ⁶², ⁷¹ and three barriers to acceptance with mistrust of research reported in two studies.⁶², ⁹⁰

Smoking Cessation Trials. ⁴¹, ⁸⁶, ⁹⁷ Out of the three smoking cessation trials in African American populations, only one trial reported barriers to accrual.⁹⁷ The barriers were: transportation problems, time and family considerations, communication / methods of presentation, mistrust of medical research, perceived harms of trial participation, reluctance to quit smoking, and job issues.

Screening Trials. ⁴³ The AAMEN study is a randomized trial comparing four different strategies for accrual of African American men into the Prostate, Lung, Colorectal, and Ovarian (PLCO) trial. According to the baseline data from this enrollment trial, having an unlisted phone number was a statistically significant barrier to enrollment. Because the enrollment procedures included telephone-based recruitment, having an unlisted phone number functioned as a barrier to awareness about the study. Those who had been screened by the prostate-specific antigen-screening test within the previous three years were excluded from participation in the study. Interestingly, having a lower income level was a promoter of participation, suggesting both the lack of access to screening services in the targeted low-income African-American population, and the screening study as an opportunity for low-income men to access these services.

Promoters of Accrual to Cancer Prevention in African-American Populations. Among the 13 studies of barriers and promoters to accrual of African Americans to cancer prevention trials, each reported, on average, three promoters (range: 0 to 8). Among the 29 promoters of accrual to prevention trials, there were four promoters of awareness, each reported in a single study: reminder phone call, ⁹⁷ adequate knowledge about the study, ⁸⁰ a workshop on clinical trials, ⁸⁰ and culturally relevant education about trials.⁸⁸ Of the 14 reported promoters of opportunity the most frequently reported was provision of transportation (n = 2).⁸⁰, ⁹⁷ Of the 12 promoters of acceptance, the most frequently reported was the use of incentives (n = 2).⁸⁰, ⁹⁷

Chemoprevention Trials. ⁶², ⁷¹, ⁹⁰ Of the three chemoprevention trials that enrolled African Americans, only Mouton's study⁹⁰ reported promoters for a total of two promoters to acceptance: preference for the study's principal investigator to be Black, and perception that it is better to be treated by research doctors.

Smoking Cessation Trials. ⁴¹, ⁸⁶, ⁹⁷ On average, each of the three smoking cessation trials reported five promoters (range: 3 to 8). While no promoters of awareness were reported, 11 promoters of opportunity were reported including provision of transportation, ⁹⁷ communication/method of presentation, ⁴¹ and employment status.⁸⁶ The five reported promoters of acceptance were incentives, support, encouragement, prayer, the certainty of receiving actual medication, the impact of diagnosis on risk perception, and perceived benefit of trial participation.

Screening Trials. ⁴³ Because the AAMEN study targeted a low-income African-American population, having a lower income level was a promoter of participation.

Barriers and Promoters of Accrual to Prevention Trials Among Other Underrepresented Populations

Because of the limited evidence available regarding barriers and promoters of participation in cancer prevention trials in populations other than African Americans, we discuss these barriers and promoters together within each of the following paragraphs.

Barriers and Promoters in Other Racial and Ethnic Minority Populations.

Latinos/Hispanics. Overall, there were four studies that reported evidence on barriers and promoters to accrual of Latino/Hispanics to cancer prevention trials, ⁴¹, ⁴², ⁶², ⁷⁴ and on average, each reported four barriers (range: 0 to 8). For this population, there was only one reported barrier to awareness, the lack of education about clinical trials.⁷⁴ Of the reported seven barriers to opportunity, the most frequently reported was transportation (n = 2).⁴², ⁷⁴ Of the seven barriers to acceptance, the most reported was mistrust of research and the medical system (n = 2).⁶², ⁷⁴

American Indian/Alaska Natives. The limited amount of evidence available for the American Indian/Alaska Native population with regard to accrual to clinical trials in general, was discussed in the section on barriers and promoters to accrual to therapeutic trials.

Asian/Pacific Islanders. We did not find any evidence regarding barriers or promoters of awareness, opportunity or acceptance of cancer clinical trials for this population. The only available evidence was that this population is underrepresented in NCI-funded cancer clinical trials.³²

Barriers and Promoters in the Elderly. Overall, there were three studies of barriers and promoters of enrollment to cancer prevention trials in the elderly. These studies reported three distinct barriers, age being the most frequently cited (n = 2).⁴³, ⁴⁸ On average, each of these studies reported one barrier (range: 0 to 3). Among the three promoters of accrual to prevention trials, there were no promoters of awareness or acceptance. The three reported promoters of accrual were the entry criteria, ⁸⁷ age, ⁴⁸ and low-income status.⁴³

Barriers and Promoters in Rural Populations. The available evidence on barriers and promoters of accrual of rural populations to cancer prevention and treatment trials is based on two studies that we discussed in the section on barriers and promoters of accrual to therapeutic trials.⁴⁶, ⁵³

Grading of the Total Body of Evidence

The overall evidence denotes a wide variety of barriers and promoters that have an influence on the participation of underrepresented populations into cancer clinical trials. However, the overall was considered to be very low due to the limitations in quality and precision as well as the lack of strong evidence within the studies (see Table 9). In addition, a high probability of reporting bias existed within the studies. The evidence is based on 45 studies, 34 of which addressed cancer treatment trials, 16 of which addressed cancer prevention trials, and five of which addressed trials for both cancer treatment and prevention. While the studies were heterogeneous and contained over 35,000 participants in total, the majority of the studies generally focused on an African American population and an elderly population. Other underrepresented populations such as Latinos/Hispanics, Native Americans/Alaska Natives, Asian/Pacific Islanders, adolescent or rural populations were scarcely evaluated.

Identification of Relevant Articles

In reviewing the current literature on question 6, we found a total of 10 eligible articles.⁴⁶, ⁵⁶, ⁶⁴, ⁶⁶, ⁸⁴, ⁸⁸, ⁹², ⁹⁵, ¹⁰⁰, ¹⁰¹ Details about these studies are given in Table 7, and Evidence Tables 6-1, 6-2, 6-3, and 6-4 in Appendix F.

Study Characteristics

Of the 10 studies eligible for review, two were published before 1990, ⁹⁵, ¹⁰¹ three were published between 1990 and 1999, ⁵⁵, ⁸⁸, ¹⁰⁰ and five were published after 2000⁴⁶, ⁶⁴, ⁶⁶, ⁷⁶, ⁹²(see Evidence Table 6-1 in Appendix F). The study designs used were concurrent controlled trial (n=1), ⁴⁶ case-control (n=1), ⁶⁶ or descriptive (n=8)⁵⁶, ⁶⁴, ⁸⁴, ⁸⁸, ⁹², ⁹⁵, ¹⁰⁰, ¹⁰¹ with three of the studies also employing qualitative techniques to facilitate in-depth assessments of physician attitudes and perceptions. Nine were U.S.-based studies and one was a British-based study. All were recruiting into treatment trials; three of the studies were also recruiting into prevention trials.

Target Population Characteristics

Seven studies⁴⁶, ⁵⁶, ⁸⁴, ⁸⁸, ⁹⁵, ¹⁰⁰, ¹⁰¹ targeted physicians, one targeted physicians/investigators⁹² and two studies⁶⁴, ⁶⁶targeted patients/participants as the study group of interest. Specific methods and techniques employed in these studies included surveys (n=8) that were implemented either via mail, ⁴⁶, ⁵⁶, ⁸⁴, ¹⁰⁰, ¹⁰¹ telephone, ⁶⁶ or other methods. Other methods employed to assess physician barriers to enrolling patients into clinical trials were focus groups, ⁹² workshops, ⁸⁸ and face-to-face interviews⁹⁵ (Evidence Table 6-2 in Appendix F).

Two studies employed physician interventions to increase patient accrual to clinical trials⁴⁶, ⁸⁸with only one measuring the outcomes of their interventions.⁴⁶

Six of the studies targeted at least one of the underrepresented populations as a primary goal of the study. Three studies targeted either racial or ethnic minorities⁸⁸, ⁹² two studies targeted the elderly, ⁶⁶, ⁸⁴ and one study targeted rural populations.⁴⁶ Four studies did not target an underrepresented population as a primary endpoint; however, data were presented on advanced age⁹², ⁹⁵, ¹⁰⁰ and social class (non-U.S. study)¹⁰⁶ to merit their inclusion in this report.

Quality of Studies

We evaluated the quality of studies in each of five areas or domains: 1) representativeness, 2) justification of study methods, 3) reliability and validity of data collection methods, 4) potential for bias/confounding, and 5) data analysis. Selected aspects of study quality are summarized in Evidence Table 6-3 (see Appendix F).

In terms of representativeness, the studies were fairly well described (four clearly described study participants). For justification of study methods, five studies adequately documented the rationale for their studies. With the exception of one study, ¹⁰⁰ the study instruments used in the surveys and the data collection methods used in the qualitative studies had serious limitations in reliability and validity. Five studies used adequate measures to protect against bias or confounding. For studies that used surveys as part of data collection, this was based on participant response rate and selection of subjects; for studies that used qualitative techniques to collect data, this was based on the selection of subjects. In terms of the description of the data analysis, all studies were inadequate in their reporting.

Results of Studies

Role of provider attitudes. Nine studies presented data on how provider attitudes/perceptions were barriers to and promoters of accrual to cancer clinical trials. Four studies found provider atttitudes as a barrier to enrollment⁴⁶, ⁶⁶, ⁹², ¹⁰¹ while one study found provider attitudes to be a promoter of patient accrual.⁸⁴ The studies also reported that patient age, ⁶⁶, ⁹⁵, ¹⁰¹ comorbidity, ⁶⁶, ⁹⁵ disease stage, ⁶⁶ mistrust of researchers, ⁸⁸, ⁹² and lack of physician awareness about trials⁸⁸, ⁹²were factors that prevented providers from enrolling their patients into clincial trials. Two studies⁵⁶, ¹⁰⁰ found that provider communication or method of presentation were barriers to patient enrollment, whereas one study found it to be a promoter of trial enrollment.⁸⁴

For studies that targeted minority populations, ⁶⁴, ⁸⁸, ⁹² mistrust of researchers and lack of provider awareness about trials were leading provider barriers⁸⁸, ⁹²that decreased patient enrollment in clinical trials. Additionally, concerns about patient non-compliance and a lack of available protocols were reasons cited for not talking to patients about clinical trials.⁶⁴

For studies that targeted the elderly, provider attitudes regarding clinical trials prevented them from sharing information about trials to their patients in one study, ⁶⁶ and increased their willingness to enroll patients in clinical trials in another study⁸⁴ (Evidence Table 6-4 in Appendix F).

Study design barriers. Eight studies evaluated the role a clinical trial's study design plays in the ability to accrue underrepresented populations to cancer clinical trials. Four studies reported issues with the protocol itself with the length of the trial and visit structure being issues in two studies, ⁸⁸, ⁹² the strictness of the protocol in one study, ⁹² and the amount of travel, failure to use the most effective drug, and the number of required lab tests being issues in one study.⁹⁴ We also found that accrual to cancer treatment trials was decreased by study design issues including comorbid conditions, ⁸⁴, ⁹² age exclusion, ⁸⁴, ⁹² eligibility criteria, ⁵⁶ issues with randomization, ⁹² medication exclusion, ⁶⁴ treatment toxicity and side-effects, ⁹⁵ life expectancy, ⁸⁴ and disease stage/location¹⁰⁰ (Evidence Table 6-4 in Appendix F).

Health care system. Two studies⁸⁸, ⁹² reported that the healthcare system itself played a role in decreased patient accrual to cancer clinical trials. Both studies found that a lack of dissemination of study opportunities to providers made it difficult for them to match potentially eligible patients to cancer clinical trials. Additionally, Pinto et al.⁹² found that a lack of cultural competence among providers and/or staff as well as a lack of patient access to institutions conducting the clinical trials made it difficult to accrue patients to these trials. McCaskill-Stevens et al.⁸⁸ found that a lack of minority investigators and personnel impeded accrual (Evidence Table 6-4 in Appendix F).

Clearly the impact of the provider on accrural has not been completely elucidated. While our review highlights that the role of individual level provider factors, study design and system level factors may be important to the accrual process, it is still unclear how these factors relate to each other, which are most important, if temporal or spatial effects occur or if any of these factors are synergistic.

Grading of the Total Body of Evidence

As summarized in Table 10, we assessed the quantity, quality, and consistency of evidence on Question 6 and concluded that the studies merited an overall “very low” evidence grade. This grade is based primarily on the fact that all ten studies scored below the 50th percentile in the quality rating, with two scoring below the 25th percentile, based on study representativeness, justification, bias, reliability/validity, and statistical analysis. Additionally, points were deducted for imprecision or sparcity of data as well as the high probability of reporting bias, based on the study designs. Thus, the evidence was insufficient to determine with any confidence how healthcare providers' attitudes and perceptions influence the efficacy/effectiveness of strategies for recruitment of underrepresented populations into cancer treatment and prevention trials. This evidence is based on studies where approximately 1,178 providers were interviewed or surveyed. Furthermore, the studies were heterogeneous in what underrepresented population they targeted.