The need for adequate African American representation in clinical trials is implicit when one considers the racial differences that exist in the pharmacokinetics and pharmacodynamics of certain drugs7,25 and biologic course of certain diseases1,6. Representation is also necessary for promoting best practices based on contemporary trial evidence, and is necessary for meeting the ethical mandates of social justice in clinical trial participation26. Despite these recognitions, however, African Americans are frequently underrepresented in trials evaluating breakthrough therapies and technologies in the areas of cancer16,18, human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS)8,20,22, and cardiovascular disease (CVD)10.
African American persons are more likely to refuse participation in trials, possibly due to less trust in the health system15,22. Very small sample sizes and lack of adequate control comparisons with white persons, however, have limited causality to be established between race, distrust, and willingness to join. In one of the largest series to date3, African Americans were found to express greater distrust toward medical researchers than whites, even after adjusting for socioeconomic confounders. This study, however, only measured attitudes about distrust, and did not link distrust with any enrollment preferences or behaviors. Also previously unstudied is whether patients' perceptions toward the risks and benefits of joining clinical trials differ by race, and whether these perceptions influence the enrollment decision.
A general lack of study of these issues in the area of cardiovascular trials is reason for additional consideration, given the prolific nature of cardiovascular trials, their influence on modern clinical practice, and the access they afford for novel treatments. For these reasons, we asked the following questions: Are African American persons less willing to join a cardiovascular drug prevention trial than white persons; do African Americans express greater distrust toward medical researchers and/or have different perceptions about risks and benefits to joining clinical trials; and do altered perceptions and medical researcher distrust provide mechanisms for why African Americans might be less willing to join cardiovascular clinical trials?
We conducted a multicenter cross-sectional study in participants aged ≥18 years. Study centers were 13 Maryland-based internal medicine and cardiology clinics chosen on the basis of their geographic location (urban or suburban) and academic affiliation (academic or community) to afford a socially diverse study population. Depending on the system and volume of each clinic, individuals were approached by trained recruiters either before or after their provider visit in either consecutive fashion or a random sample of patients in high-volume clinics (using a computerized random numbering scheme, www.random.org). Due to practical limitations of the recruiting staff, the sampling fraction varied by day, depending on clinic patient volume.
Eligibility criteria required individuals to 1) be visiting the clinic for a scheduled appointment with a provider, 2) possess English fluency or ability to translate the survey, and 3) have sufficient cognitive capacity, as determined by the recruiter, to comprehend the study. The local hospital research review boards associated with each of the sites approved the study protocol and permitted use of a standardized oral consent procedure. Of 1440 individuals approached for study participation between April and October 2002, 308 were excluded due to ineligibility. Of the 1132 eligible, 925 consented to complete the survey (consent rate = 82%), with 675 completing the survey while in clinic and the remainder taking the survey home (overall response rate = 789 individuals, 70%) (Figure 1). Comparing responders to nonresponders, responders were younger (mean [SD] age = 54.1  vs. 58.5  yr, p = 0.0001), but of similar sex and race as nonresponders. Of the 789 responders, we excluded those who were neither white nor African American (n = 28), and those with missing data for WTP (n = 6) or any of the mediator or confounder variables (n = 38), assuming missing data for any 1 variable did not comprise more than 2% of responses. The missing indicator method was used to adjust for those individuals (13%) who not report household family income. The final data set consisted of 717 individuals.
Recruiters informed study participants that the general purpose of the study was to understand individuals' attitudes about medical research and physicians who perform such research, and reasons for joining or declining participation in clinical trials. Participants were blinded to the study hypotheses. Participants began the survey by reading a 1-page description of a hypothetical double-blinded, placebo-controlled CVD drug prevention trial. The description read at a Flesch-Kincaid Grade Level of 8.2 (Microsoft Word 2003, Redmond, WA) and was akin to a typical consent document used in CVD trials. Elements disclosed were the experimental nature of the study; study objectives; adverse drug event risk and type; potential benefits to joining; voluntary nature of the study; study requirements with regard to length and type of follow-up; alternative treatment options; right to withdrawal; the study sponsor; the presence of potential investigator financial conflicts of interest; and rewards for joining, which included free trial-related health care and monetary compensation5. Willingness to participate (WTP) in the trial was subsequently measured using a 5-point Likert response scale (very likely, likely, moderately, unlikely, very unlikely). The remainder of the questionnaire measured patients' attitudes toward specific elements of the described clinical trial, medical researchers, and the health system. Self-reported conditions, health status, prior trial and health care experience, and sociodemographic data were also collected.
An expert panel of cardiovascular clinical trialists reviewed the hypothetical trial for face and content validity. To internally validate that WTP responses to the hypothetical trial provide a reflection of true behavior, we correlated WTP with responses to 2 additional survey questions: "If previously ever approached to join a clinical trial, did you agree to join?" and among those who had joined a prior study, "Would you ever join another study if approached again?" Responses to these 2 questions were highly correlated with WTP, with nonparametric z-scores of 4.4 (p < 0.0001) and 5.4 (p < 0.0001), respectively. Self-stated WTP has also been documented to be predictive of actual participation in another validation study11.
Independent, Mediator, and Confounding Variables
Individuals self-categorized their race as either white (white) or African American (black or African American) race. Due to insufficient numbers of individuals from other races (Asian, n = 19; American Indian or Alaska Native, n = 2; Other, n = 7), we focused only on whites and African Americans in this analysis. Being of "Hispanic or Latino origin or descent" (n = 20) was not mutually exclusive of reported race. Mediator variables included individuals' distrust toward medical researchers, measured on the 7-point Corbie-Smith distrust index3, and perceptions of the risks and benefits of trial participation, measured using 100-point visual analog scales4. The 7 questions of the distrust index asked patients' perceptions regarding the degree to which doctors fully explain research to patients, can be freely asked questions by patients, purposefully expose patients to harm, protect patients from unnecessary risk, use people like guinea pigs, prescribe medication as a way of experimentation without consent, and have ever given patients experimental treatment without consent. To measure perceptions of risk and benefit of joining the trial, we asked patients to rate using a 0-100 point scale the "chance that [their] health would benefit from joining the study", "the chance that [they] would experience personal harm or injury, such as a side effect from the study drug, by joining the study," and "the chance that the quality of [their] health care would improve by joining the study."
Confounder variables included age (in years: <45, 45-59.9, 60-74.9, ≥75), sex, self-reported CVD risk factors/conditions (coronary artery disease, congestive heart failure, peripheral vascular disease, diabetes, hypertension, hypercholesterolemia, overweight/obesity, current smoking), education level (<12th grade, grade 12/high school graduate, some college/college graduate, some postgraduate/graduate degree), household income (no response, $0-$29,999, $30,000-$49,999, $50,000–$74,999, ≥$75,000), insurance type (Medicare, managed care/HMO, private, Champus/Tri-Care, Medicaid, or none), and employment status (part-time/full-time, disabled/retired, unemployed/other).
We compared the ordered categorical WTP responses of white patients with those of African American patients using the nonparametric Wilcoxon rank-sum test. Pairwise associations between WTP, race, and mediator variables were also examined using either the Pearson chi-square method or the Wilcoxon rank-sum test. We performed prespecified sample size calculations based on an ability to detect a 25% relative difference in unadjusted WTP between race with 90% power and an alpha = 0.025. This calculation yielded a minimum of 242 subjects per race needing to be recruited. We used multiple stepwise logistic regression analysis to examine the presence, strength, and independence of an association between race and WTP, and the potential ability for mediator variables to "explain away" any crude racial associations with WTP. While confounder variables were forced into the multivariable model, we also ruled out their collinearity by evaluating them in a covariance matrix. To facilitate interpretation of our multivariable model, we treated all categorical independent and confounder variables as indicators and collapsed WTP scores into a dichotomous outcome: "very likely/likely" ([+] response) and "moderately/unlikely/very unlikely" ([−] response). Similarly, since each of the 3 mediator variables that assessed patients' perceptions of risk and benefit were measured as continuous scores on a 0-100 point visual analog scale, we collapsed each of them into quintile categories (0-19, 20-39, 40-59, 60-79, 80-100). Sensitivity analysis demonstrated that collapsing these variables had no meaningful impact on the major study observations. In addition, we created models with robust estimators of variance through the use of generalized estimation equations28, treating unique provider of sample patients as the cluster variable. Since the output from this model was indifferent from those of our original regression models, only the original regressions are presented. All statistical analyses were performed with Stata v7.0 (Stata Corp., College Station, TX).
General characteristics of the study participants stratified by race are displayed in Table 1. African American participants were more likely than white participants to be of younger age (p < 0.001) and female (p < 0.001). African Americans tended to be of lower socioeconomic class as measured by education and annual household family income level (see Table 1). Distribution of insurance types also differed; Medicare was more frequently cited as a primary carrier by whites, and underinsurance, defined by Medicaid or no insurance, more frequently cited by African Americans (p < 0.001). With regard to cardiovascular risk profiles, more whites reported the presence of coronary artery disease and hypercholesterolemia, while more African Americans reported the presence of hypertension, diabetes mellitus, and current smoking.
Relationship Between Race, Potential Mediator Variables, and WTP
We first determined WTP as a function of race. In unadjusted analysis, African American participants were markedly less willing to join the trial than white participants (p = 0.0002) (Figure 2). Only 28% of African Americans and 39% of whites reported being either likely or very likely to participate. Further, 54% of African Americans versus only 38% of whites reported being unlikely or very unlikely to participate.
In crude analysis, distrust toward medical researchers proved to be a powerful predictor of WTP. Scores ranging from 0 (low distrust) to 7 (high distrust) were strongly and inversely linearly associated with WTP (p < 0.0001, for inverse linear trend). Positive WTP (very likely/likely response) was 53% (100/188) among those with very low levels of distrust and a score = 0, 36% (65/179) with score = 1, 28% (41/145) with score = 2, 27% (28/105) with score = 3, and 7% (2/28) with score = 6 or 7.
We found that individuals perceiving that either their personal health would benefit or the quality of their health care would improve were markedly more likely to report positive WTP (both p < 0.0001, for positive linear trend). For instance, positive WTP was 15% (13/89) among patients perceiving the lowest chance (0-19% as measured from the 0-100 point visual analog scale) of personal health benefit, 21% (24/113) if 20%-39% chance, 33% (98/294) if 40%-59%, 51% (83/164) if 60%-79%, and 63% (36/57) if 80%-100% chance. Converse to these findings, we found that patients perceiving a greater chance of harm or injury from joining the trial were markedly less likely to report positive WTP (p < 0.0001, for inverse linear trend). Positive WTP was 46% (37/80) among patients reporting a 0-19% chance of harm, 54% (65/121) if 20%-39% chance, 36% (107/301) if 40%-59%, 23% (30/130) if 60%-79% chance, and 18% (15/85) if 80%-100% chance.
Relationship Between Race and Potential Mediator Variables
Table 2 demonstrates the racial differences in response to the 7 components of medical researcher distrust index. African American participants expressed markedly higher levels of distrust toward medical researchers than white participants (median scores [25-75 range] = 2 [1-3] vs. 1 [0-2], respectively; nonparametric z-score = 8.7; p < 0.00001). For every domain tested in the index, African Americans were more likely to express a negative or guarded ("don't know") response than whites. While the racial disparity in how individuals responded to the researcher distrust index was striking, so, too, were the aggregate responses. Approximately 58% of African Americans versus 25% of whites (p < 0.001) thought doctors very often, fairly often, or did not know how often prescribed medications as a way of experimenting on people without their knowledge or consent. Further, approximately 25% of patients from both races expressed belief that their doctors sometimes exposed them to unnecessary risk. It is additionally notable that even among those expressing WTP in the trial, African Americans still consistently reported greater overall distrust than whites (p < 0.001).
With regard to patients' perceived benefits of joining the trial, either in terms of health outcome or quality of care, African American participants and white participants did not differ in their responses (p = 0.91 for health benefit and p = 0.58 for quality of care). It is noteworthy that African Americans expressed belief that they would have a greater chance of experiencing harm or injury from participating in the trial than whites (p < 0.0001) (Figure 3). For instance, 45% of African Americans versus only 22% of whites perceived a 60% or higher chance of experiencing harm or injury, such as a study drug-related adverse event, from joining the trial.
Multivariable Adjustment to Evaluate Independent Effect of Race on WTP
Using multiple stepwise logistic regression modeling, we assessed whether the greater medical researcher distrust and perceived chances of harm expressed by African Americans could explain the racial disparity that existed with regard to WTP, while adjusting for health and sociodemographic confounders (Table 3). In crude analysis, African American race was associated with a lower WTP (crude odds ratio [OR], 0.59; 95% confidence interval [CI], 0.43-0.83; p = 0.002), relative to that of whites. As demonstrated in multivariable model 1, adding potential confounder variables to the WTP equation, including categorical age, sex, socioeconomic status, and cardiovascular health profile, had no real impact on the relationship between race and WTP (multivariable adjusted OR, 0.58; 95% CI, 0.39-0.85; p = 0.006). However, given the strong association between the Corbie-Smith 7-point index of medical researcher distrust and lower WTP as previously reported5, the stepwise addition of medical researcher distrust to the model appeared to explain a substantial degree of the association between race and WTP (explanatory model 2: OR, 0.77; 95% CI, 0.51-1.16; p = 0.21). In the full explanatory model, adjustment for both distrust and perceived chance of harm further explained away and eliminated any meaningful association between race and WTP (explanatory model 3: OR, 0.84; 95% CI, 0.54-1.30; p = 0.44).
Numerous investigators have raised concerns that low social class accounts for fewer numbers of African Americans willing to consider participating in clinical trials2,12,19. Our data, however, provide compelling evidence that factors other than socioeconomics may represent core drivers of racial enrollment disparities. We found that African American participants, on average, were 41% less willing to participate in a cardiovascular drug prevention trial than white participants. This lesser willingness was unaffected after adjusting for socioeconomic status. African Americans, however, expressed markedly higher levels of distrust toward medical researchers and a greater perceived chance of experiencing harm or injury from joining the trial. Using stepwise multivariable regression analysis, we demonstrated that these attitudes and perceptions are important mechanisms by which African Americans are less willing to join. After adjusting for medical researcher distrust and perceived chance of harm, race, by itself, had no meaningful impact on WTP.
The issues of minorities' distrust of the health care system and minorities' underrepresentation in clinical trials have often been studied in parallel, but not in sequence. Looking at racial disparities in trust of the medical community, a study on recruitment for the Women's Health Initiative in oncology demonstrated that African American women were 8 times as likely as whites to believe that scientists cannot be trusted (33% vs. 4%)15. This study, however, did not address how such distrust related to clinical trial enrollment. Another study examining both distrust in the African American community and enrollment in AIDS trials showed that distrust was the strongest predictor of lower WTP; given that this study only sampled African American patients, however, it was unable to assess distrust as it relates to WTP independent of race, leaving inadequate control comparisons22. Similarly, observations from the African-American Antiplatelet Stroke Prevention Study showed that fear of being used as "a guinea pig" related to decreased WTP, but failed to examine distrust and WTP among non-African American patients9.
Although 2 studies did not find racial differences in WTP17,21, the sample size of 1 study was small, limited to only a single center, and the majority of surveyed patients were already concurrently enrolled in another separate cardiac study21; meanwhile the other study, from the same center, inquired about WTP in trials of more invasive cardiac procedures such as percutaneous transluminal coronary angioplasty (PTCA) or coronary artery bypass grafting (CABG)17, as opposed to the current study of WTP in placebo-controlled drug trials. However, in a community-wide survey across African American and white patients, similar to our multicenter design, authors of a study evaluating the relationship between race and WTP did indeed find that African Americans' distrust of the medical community provided an important link between African American and lower WTP23. Unlike our study, however, their smaller sample size and crude race-stratified analysis to evaluate WTP, as opposed to comprehensively evaluating race in a multivariate model, did not enable the opportunity to assess for causal inference while controlling for confounders.
Greater distrust of medical research and researchers among African American participants relative to white participants may stem from a general distrust of mainstream society14; gross past examples of unethical medical research with ethnic minorities such as the United States Public Health Service observational syphilis study, also known as the Tuskegee study14; and perpetuated racial and ethnic biases, both perceived and overt, affecting minority experiences within the health care system13. To our knowledge, no prior studies have quantified risk perceptions by race as we did in this study, within the personalized context of a clinical trial. Our data, however, are supported by several past qualitative studies, which have thematically shown African Americans' and other minorities' fear of medical experimentation2,24.
It is worth briefly considering the results of the current study in the context of a 2006 report by Wendler and colleagues27, who performed a comprehensive literature search to identify published health research studies that reported consent rates by race or ethnicity. Of the 20 eligible studies identified, reporting on the enrollment decisions of over 70,000 individuals, the authors found a small difference in the rates at which non-Hispanic whites and minorities agreed to participate in health research. While consent rates were comparable, minority groups were invited to participate considerably less frequently. The conclusion of the study was that efforts to increase minority participation in health research should focus on increasing minority access to trials. While we do not dispute the importance of access, our results support the notion that perceptions of adverse harm and views toward medical researchers appreciably differ between races and that these attitudes notably impact WTP decisions.
The current study has limitations that warrant consideration during the interpretation of our data. The major limitation likely relates to the fact that we proposed a hypothetical clinical trial to patients; real-world behaviors, enrollment factors, and ultimate decisions may differ from what sample individuals reported in this study. To minimize this source of bias, however, we presented to patients a trial that closely simulated a real-world scenario and underwent careful content validation by a panel of experienced clinical trialists, cardiologists, and psychometric experts. We also included several questions in the survey instrument to internally validate patients' responses to WTP; as previously described, these validation questions did indicate a strong correlation between expressed WTP and actual trial participation. Further, we based the approach to using a hypothetical trial on the experience of other investigators in this field; for instance, it has been previously demonstrated that WTP in a hypothetical trial was the single best predictor of eventual enrollment in a similarly designed, real Phase II AIDS vaccine trial11.
A second limitation pertains to the preventive CVD nature of our hypothetical trial. The present results may not necessarily be generalizable to other disease states, particularly those that are imminently life threatening. A third limitation relates to the fact that although the study was conducted in 13 clinical sites from a wide range of sociodemographic settings, the generalizability of the results outside the region are less certain. Nevertheless, the wide variety of urban and suburban, academic and community, and internal medicine and cardiology clinical sites does reflect the typical population base of participants visiting clinical trial centers in the United States. Finally, although the current study captures the attitudes of those patients who consent and are initially receptive to a research questionnaire, we believe that such a target study population likely represents the most relevant population for understanding the behavioral attitudes regarding trial participation, as we are interested in factors of WTP among individuals equipped with informed knowledge of a research study. While social response bias is always a potential factor in survey studies, we believe that we further minimized this bias by allowing patients to complete the survey at their own discretion in the place of their choice and in total anonymity. Moreover, the 82% consent rate and 70% response rate in our study are considered relatively high and respectable in population survey studies.
Despite the limitations, the current study sheds important new insight into the mediators of potential racial enrollment disparities, particularly when applied to cardiovascular clinical trials. Many investigators of preventive cardiovascular trials may not perceive the risks to joining such trials as high or as barriers to enrollment. Patients are likely to evaluate the risks very differently, however, and rely on their preformed beliefs and perceptions to guide important health care decisions, including trial participation.
Along the same lines, overcoming distrust is viewed as an important step toward improving the recruitment and retention of African Americans in clinical trials3,14. Strategies to dismantling this barrier warrant collective community-based approaches, including those that foster long-term relationships among patients, their health care team members, and community and religious leaders; education; ongoing communication, responsiveness, and interest by the scientific community; and improved sensitivity to cultural and local social norms.
We thank Leon Gordis, MD, DrPH, Vera M. Zlidar, MHS, Oyelola O. Faparusi, MBBS, PhD, and Susan Shultz (all of Johns Hopkins Medical Institutions at time of study) for their assistance.
1. Bella JN, Palmieri V, Kitzman DW, Liu JE, Oberman A, Hunt SC, Hopkins PN, Rao DC, Arnett DK, Devereux RB. Gender difference in diastolic function in hypertension (the HyperGEN study). Am J Cardiol
2. Corbie-Smith G. The continuing legacy of the Tuskegee Syphilis Study: considerations for clinical investigation. Am J Med Sci
3. Corbie-Smith G, Thomas SB, St George DM. Distrust, race, and research. Arch Intern Med
4. Diefenbach MA, Weinstein ND, O'Reilly J. Scales for assessing perceptions of health hazard susceptibility. Health Educ Res
5. Ding EL, Powe NR, Manson JE, Sherber NS, Braunstein JB. Sex differences in perceived risks, distrust, and willingness to participate in clinical trials: a randomized study of cardiovascular prevention trials. Arch Intern Med
6. Earle KK, Porter KA, Ostberg J, Yudkin JS. Variation in the progression of diabetic nephropathy according to racial origin. Nephrol Dial Transplant
7. Evelyn B, Toigo T, Banks D, Pohl D, Gray K, Robins B, Ernat J. Participation of racial/ethnic groups in clinical trials and race-related labeling: a review of new molecular entities approved 1995-1999. J Natl Med Assoc
8. Gifford AL, Cunningham WE, Heslin KC, Andersen RM, Nakazono T, Lieu DK, Shapiro MF, Bozzette SA. Participation in research and access to experimental treatments by HIV-infected patients. N Engl J Med
9. Gorelick PB, Harris Y, Burnett B, Bonecutter FJ. The recruitment triangle: reasons why African Americans enroll, refuse to enroll, or voluntarily withdraw from a clinical trial. An interim report from the African-American Antiplatelet Stroke Prevention Study (AAASPS). J Natl Med Assoc
10. Hall WD. Representation of blacks, women, and the very elderly (aged > or = 80) in 28 major randomized clinical trials. Ethn Dis
11. Halpern SD, Metzger DS, Berlin JA, Ubel PA. Who will enroll? Predicting participation in a phase II AIDS vaccine trial. J Acquir Immune Defic Syndr
12. Harris Y, Gorelick PB, Samuels P, Bempong I. Why African Americans may not be participating in clinical trials. J Natl Med Assoc
. 1996;88: 630-634.
13. Institute of Medicine. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington, DC: National Academy Press; 2002.
14. Moreno-John G, Gachie A, Fleming CM, Napoles-Springer A, Mutran E, Manson SM, Perez-Stable EJ. Ethnic minority older adults participating in clinical research: developing trust. J Aging Health
15. Mouton CP, Harris S, Rovi S, Solorzano P, Johnson MS. Barriers to black women's participation in cancer clinical trials. J Natl Med Assoc
16. Murthy VH, Krumholz HM, Gross CP. Participation in cancer clinical trials: race-, sex-, and age-based disparities. JAMA
17. Peterson ED, Lytle BL, Biswas MS, Coombs L. Willingness to participate in cardiac trials. Am J Geriatr Cardiol
18. Roberson NL. Clinical trial participation. Viewpoints from racial/ethnic groups. Cancer
19. Robinson SB, Ashley M, Haynes MA. Attitude of African-Americans regarding prostate cancer clinical trials. J Community Health
20. Schilling RF, Schinke SP, Nichols SE, Zayas LH, Miller SO, Orlandi MA, Botvin GJ. Developing strategies for AIDS prevention research with black and Hispanic drug users. Public Health Rep
21. Sen Biswas M, Newby LK, Bastian LA, Peterson ED, Sugarman J. Who refuses enrollment in cardiac clinical trials? Clin Trials
22. Sengupta S, Strauss RP, DeVellis R, Quinn SC, DeVellis B, Ware WB. Factors affecting African-American participation in AIDS research. J Acquir Immune Defic Syndr
23. Shavers VL, Lynch CF, Burmeister LF. Racial differences in factors that influence the willingness to participate in medical research studies. Ann Epidemiol
24. Stark N, Paskett E, Bell R, Cooper MR, Walker E, Wilson A, Tatum C. Increasing participation of minorities in cancer clinical trials: summary of the "Moving Beyond the Barriers" Conference in North Carolina. J Natl Med Assoc
25. Strickland TL, Lin KM, Fu P, Anderson D, Zheng Y. Comparison of lithium ratio between African-American and Caucasian bipolar patients. Biol Psychiatry
26. Tilburt J, Ford JG, Howerton MW, Gary TL, Lai GY, Bolen S, Baffi C, Wilson RF, Tanpitukpongse TP, Powe NR, Bass EB, Sugarman J. Applying justice in clinical trials for diverse populations. Clin Trials
27. Wendler D, Kington R, Madans J, Van Wye G, Christ-Schmidt H, Pratt LA, Brawley OW, Gross CP, Emanuel E. Are racial and ethnic minorities less willing to participate in health research? PLoS Med
28. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics