Participation in Cancer Clinical Trials: Race-, Sex-, and Age-Based Disparities

Context Despite the importance of diversity of cancer trial participants with regard to race, ethnicity, age, and sex, there is little recent information about the representation of these groups in clinical trials.

Objective To characterize the representation of racial and ethnic minorities, the elderly, and women in cancer trials sponsored by the National Cancer Institute.

Design, Setting, and Patients Cross-sectional population-based analysis of all participants in therapeutic nonsurgical National Cancer Institute Clinical Trial Cooperative Group breast, colorectal, lung, and prostate cancer clinical trials in 2000 through 2002. In a separate analysis, the ethnic distribution of patients enrolled in 2000 through 2002 was compared with those enrolled in 1996 through 1998, using logistic regression models to estimate the relative risk ratio of enrollment for racial and ethnic minorities to that of white patients during these time periods.

Main Outcome Measure Enrollment fraction, defined as the number of trial enrollees divided by the estimated US cancer cases in each race and age subgroup.

Results Cancer research participation varied significantly across racial/ethnic and age groups. Compared with a 1.8% enrollment fraction among white patients, lower enrollment fractions were noted in Hispanic (1.3%; odds ratio [OR] vs whites, 0.72; 95% confidence interval [CI], 0.68-0.77; P<.001) and black (1.3%; OR, 0.71; 95% CI, 0.68-0.74; P<.001) patients. There was a strong relationship between age and enrollment fraction, with trial participants 30 to 64 years of age representing 3.0% of incident cancer patients in that age group, in comparison to 1.3% of 65- to 74-year-old patients and 0.5% of patients 75 years of age and older. This inverse relationship between age and trial enrollment fraction was consistent across racial and ethnic groups. Although the total number of trial participants increased during our study period, the representation of racial and ethnic minorities decreased. In comparison to whites, after adjusting for age, cancer type, and sex, patients enrolled in 2000 through 2002 were 24% less likely to be black (adjusted relative risk ratio, 0.76; 95% CI, 0.65-0.89; P<.001). Men were more likely than women to enroll in colorectal cancer trials (enrollment fractions: 2.1% vs 1.6%, respectively; OR, 1.30; 95% CI, 1.24-1.35; P<.001) and lung cancer trials (enrollment fractions: 0.9% vs 0.7%, respectively; OR, 1.23; 95% CI, 1.16-1.31; P<.001).

Conclusions Enrollment in cancer trials is low for all patient groups. Racial and ethnic minorities, women, and the elderly were less likely to enroll in cooperative group cancer trials than were whites, men, and younger patients, respectively. The proportion of trial participants who are black has declined in recent years.

Ten years have passed since Congress responded to concerns about unequal access to clinical trials and enacted the National Institutes of Health (NIH) Revitalization Act, which encouraged representation of women and minority patients in NIH-sponsored research.¹ Ensuring broad access to research studies has subsequently been an important aim of national research policy. The National Cancer Institute (NCI) has addressed concerns about enhancing heterogeneity of trial populations through multiple mechanisms, including the creation of Minority Community Clinical Oncology Programs, the development of focused initiatives in partnership with the Centers for Disease Control and Prevention and academic medical centers, and emphasis on trials that focus specifically on the elderly.^2,3 Moreover, while the NCI budget has grown from approximately $2 billion in 1993 to approximately $4 billion in 2002,⁴ and the number of trial participants has increased, it is not clear if minorities, women, and the elderly have shared in this growth.⁵

Investigations of race- and sex-based disparities in cancer trial participation have yielded conflicting results. A study of black and Hispanic participation between 1991 and 1994 concluded that these groups were represented in cancer trial populations in equal proportion to the cancer disease burden in these groups.⁶ This report was confirmed by a subsequent analysis of enrollment by the Southwestern Oncology Group that compared the proportion of blacks and Hispanics enrolled in trials for 15 cancer types with their share of the US cancer population between 1993 and 1996.⁷ Notably, the authors found that representation of minorities varied substantially with the type of cancer. Blacks with leukemia, head and neck, or prostate cancer were more likely than white patients to enroll, while blacks with lymphoma or ovarian cancer were less likely than whites to enroll.⁷ Similarly, studies have found equivalent representation of women and men for some non–sex-specific conditions, and a sex disparity for others.^7-10 The study of Southwestern Oncology Group enrollment (1993-1996) noted sex disparities in only 3 of the 11 cancer types.⁷ However, it is unclear whether race and sex disparities are present in more recent studies, or whether they persist after accounting for age.

Recommendations for enrollment of elderly patients were conspicuously absent from the NIH Revitalization Act. Although elderly patients represent approximately two thirds of cancer patients, they account for only 25% to 30% of clinical trial participants.^7,11 Elderly minorities, theoretically subject to the combined barriers to participation that characterize elderly and minority communities, may have the greatest risk of low trial participation.^12-14 However, there is a paucity of contemporary data about the enrollment of elderly minority patients in cancer research trials.

Given the importance of broad representation in cancer trials and the considerable emphasis that Congress has placed on achieving this goal, we conducted 2 complementary analyses of recent enrollment in NCI-sponsored Clinical Trial Cooperative Group therapeutic trials. First, we compared patients enrolled in breast, colorectal, lung, or prostate trials in 2000 through 2002 with population-based cancer incidence data. We determined whether age, sex, or racial and ethnic disparities in participation exist, and whether the magnitude of disparities varies according to age group and cancer type. We then assessed temporal changes in enrollment among racial and ethnic groups, comparing patients enrolled in 2000 through 2002 with those enrolled in 1996 through 1998.

Enrollment data for NCI-sponsored therapeutic trials for 1996 through 2002 were obtained from the NCI Clinical Data Update Service, a database that contains patient-level information about participants in all NCI-sponsored Clinical Trial Cooperative Group trials. Cancer incidence data (1995 through 1999) were obtained from the NCI Surveillance, Epidemiology, and End Results (SEER) program,¹⁵ which incorporates incidence data from 11 cancer registries across the country and is regarded as the most authoritative representation of national cancer trends.¹⁵ We obtained information on age exclusion criteria for the 50 largest trials from the NCI's PDQ Database of Clinical Trials.¹⁶ Because none of these databases included personal identifiers, this study received a waiver of the informed consent requirement.

National and SEER county population estimates by 5-year age group, race, ethnicity, and sex were obtained from the bridged single race estimates provided in the US census.^17,18 In the 2000 census, 2% of the US population indicated more than 1 racial category.^18,19 To reconcile the census 2000 data with other databases that only allow a single race for each individual, these bridged estimates use a complex formula to assign races to individuals who had indicated more than 1 race.¹⁷

Our study included participants in breast, colorectal, lung, and prostate cancer trials, as these were the 4 most common causes of cancer-related mortality during the study period.²⁰ Only female participants were included for breast cancer trials since the number of male participants was negligible. We only included participants in cooperative group therapeutic trials who were 30 years of age and older and had a single documented racial or ethnic group, age, and sex. Trials of surgical interventions were excluded.

Race and Ethnicity. From 1996 through 2001, Cancer Therapy Evaluation Program (CTEP) trial participants were assigned to 1 of 5 mutually exclusive categories: white, black, Asian/Pacific Islander, American Indian/Alaskan Native, or Hispanic. In 2002, CTEP required that race and ethnicity (Hispanic or non-Hispanic) be coded as separate categories. To reconcile the 2002 race/ethnicity definitions with the CTEP data from prior years, we defined all participants who indicated that they were of Hispanic ethnicity as "Hispanic." Population data from the US census were assigned to 1 of the following mutually exclusive groups: Hispanic, white non-Hispanic, black non-Hispanic, Asian/Pacific Islander non-Hispanic, and American Indian/Alaskan Native non-Hispanic.

Cancer Incidence. Cancer incidence rates in each 5-year age, race, and sex category were estimated by dividing the number of cases in SEER by the corresponding population of the SEER counties. These rates were then applied to the corresponding categories in the US population and then summed over all categories for national estimates of annual cancer incidence for each category of interest.^7,21 Cancer incidence was assumed to be fixed during the study period.

We defined the enrollment fraction as the number of trial enrollees divided by the estimated US cancer cases in each subgroup. Pearson χ² tests of independence were performed to assess the relationship between enrollment fractions in various racial/ethnic and age groups in the 2000 through 2002 time period. Similar tests were conducted after stratifying by age and cancer type, and crude odds ratios (ORs) were estimated within each strata. We performed sex comparisons for the 2 non–sex-specific cancers in our analysis, colorectal and lung cancer.

For the analysis of temporal changes, we compared the proportion of trial participants represented by each racial or ethnic group in 2000 through 2002 with the proportion enrolled in 1996 through 1998. Patients enrolled in 1999 were omitted from this analysis, as preliminary multivariate analysis suggested that trends in representation of some racial groups began to change during this year. We used a polytomous logistic regression model to determine the relationship between time period and racial distribution of trial participants, using racial group as the outcome variable and adjusting for age, cancer type, and sex. We used this approach because the outcome consisted of multiple nominal categories (ie, 5 racial groups), and this method allowed us to use a single model to assess the relation between time and racial distribution. We performed tests for interaction between cancer type and time period and patient age group and time period. To account for the fact that participants might be independent across trials, but not necessarily within trials, we used Huber-White robust variance estimates to allow for clustering by trial.²² We conducted a sensitivity analysis by repeating the regression using only the 50 largest trials, and again after excluding the 4 trials from this sample that had age limits on eligibility. STATA 6.0 was used for all analyses.²²

From 1996 through 2002, 75 215 patients were enrolled in NCI-sponsored cooperative group nonsurgical treatment trials for breast, lung, colorectal, or prostate cancers (Table 1). Approximately 3.1% of trial participants were Hispanic, 85.6% were white, 9.2% were black, 1.9% were Asian/Pacific Islanders, and 0.3% were American Indians/Alaskan Natives. Table 1 also includes the proportion of cancer patients 30 years of age and older diagnosed in the United States in 2000 in each racial and ethnic group, as well as the distribution of race and ethnicity in the general population of adults. Breast cancer patients represented 54.2% of trial participants, although they accounted for only 27.9% of incident cancer cases. Although cancer incidence was evenly divided among the 3 age groups (30-64, 65-74, ≥75 years), the representation of trial participants was heavily skewed toward the youngest age group (68.0% of trial participants were 30-64 years old).

Trial participants represented approximately 1.7% of the total number of incident cancer cases diagnosed during the 2000 through 2002 study period (Table 2). When all 4 cancer types were considered in aggregate, Hispanics and blacks were underrepresented. Compared with a 1.8% enrollment fraction among white cancer patients, only 1.3% of Hispanic patients (OR vs whites, 0.72; 95% confidence interval [CI], 0.68-0.77; P<.001) and 1.3% of black patients (OR, 0.71; 95% CI, 0.68-0.74; P<.001) participated in trials. The enrollment fraction of Asian/Pacific Islanders (1.7%) was not significantly different than for whites. American Indians/Alaskan Natives had a higher enrollment fraction than whites (2.5% vs 1.8%) that was statistically significant (P<.001).

The enrollment fractions varied substantially with cancer type, as did racial disparities in enrollment fractions (Table 2). While breast cancer trial participants represented 3.2% of incident breast cancer patients, the enrollment fraction was only 1.9% for colorectal cancer and 0.8% for patients with either lung or prostate cancer. Black patients with breast, colorectal, or lung cancers were significantly less likely to participate in trials than were their white counterparts (ORs: 0.74, 0.64, 0.61, respectively; P<.001 for all ORs). In contrast, the enrollment fractions for blacks were slightly greater than for whites in prostate cancer trials (1.0% of blacks vs 0.8% of whites; OR, 1.18; 95% CI, 1.09-1.28; P<.001). Hispanic patients were less likely to enroll than white patients for all 4 cancer types, whereas Asian/Pacific Islanders were less likely to enroll than white patients only for lung and colorectal cancer trials.

There was a strong relationship between age and enrollment fraction. Among patients 30 to 64 years old, the enrollment fraction was 3.0%, and this decreased to 1.3% among 65- to 74-year-old patients and 0.5% for patients 75 years of age and older (Table 3). In each racial group, there were significantly lower enrollment fractions for elderly patients. The OR for trial enrollment for white elderly (≥75 years) patients vs nonelderly (30-64 years) patients was 0.14 (95% CI, 0.14-0.15; P<.001). For black cancer patients 75 years of age and older, the enrollment fraction was 0.4% compared with 1.8% for nonelderly black cancer patients (OR, 0.24; 95% CI, 0.21-0.27; P<.001). Similar disparities were noted in Hispanics, Asian/Pacific Islanders, and American Indians/Alaskan Natives (Table 3).

Men had higher enrollment fractions than women for trials involving colorectal cancer (2.1% vs 1.6%; OR, 1.30; 95% CI, 1.24-1.35; P<.001) and lung cancer (0.9% vs 0.7%; OR, 1.23; 95% CI, 1.16-1.31; P<.001 [Table 4]). When stratified by age group, there was no disparity in enrollment rates between young men and women (younger than 65 years). Among elderly subgroups, however, statistically significant disparities were noted. Among colorectal cancer patients, men 65 to 74 years of age were more likely to enroll than women (OR, 1.18; 95% CI, 1.08-1.28; P<.001); the sex disparity was somewhat larger in the 75 years and older age group (OR, 1.54; 95% CI, 1.35-1.76; P<.001). Similar findings were noted in lung cancer patients. Although there was no significant sex disparity in the youngest age group, disparities were noted in patients 65 to 74 years of age (OR, 1.40; 95% CI, 1.26-1.56; P<.001) and for the 75 years and older age group (OR, 1.71; 95% CI, 1.41-2.07; P<.001).

From 1996 through 2002, the annual number of trial participants increased from approximately 8000 to approximately 12 000 (Table 5). During this period, the percentage of trial participants who were racial or ethnic minorities decreased. Hispanic and black patients represented 3.7% and 11.0% of trial participants in 1996, respectively. These respective groups represented 3.0% and 7.9% of participants in 2002. After adjusting for age, sex, and cancer type, participants enrolled in 2000 through 2002 were less likely to be black (in comparison to whites; adjusted relative risk ratio, 0.76; 95% CI, 0.65-0.89; P<.001) than those enrolled in 1996 through 1998. There was no significant change in the representation of other racial and ethnic groups during the study period. The results did not change when analyzing only the 50 largest trials or after excluding the 4 trials from this group that had age exclusions. There were no significant interactions between time period and either cancer type or age group. When we performed 2 additional multivariate analyses, using age (≥65 years) and sex as the outcome variables, respectively, we found that there was no change in the representation of elderly or women between 1996-1998 and 2000-2002.

Although the trial enrollment rate increased by almost 50% during 1996 through 2002, the proportion of trial participants who were not white actually declined during this period. These results persisted after accounting for cancer type, age, and sex and are especially notable given the 1993 NIH Revitalization Act and the emphasis placed on increasing minority participation in cancer trials over the past decade. We also found that blacks and Hispanics were significantly less likely than whites to enroll in NCI-sponsored cooperative group trials in 2000 through 2002. In our analysis, the magnitude and direction of racial disparities varied with cancer type. While blacks had significantly lower enrollment fractions in breast, lung, and colorectal cancer trials when compared with whites, the representation of blacks and whites in prostate cancer trials was comparable. This demonstrates that equitable participation is possible and implies that there may be lessons from the design, marketing, and recruitment strategies used for prostate cancer trials that can help address underrepresentation in trials of other cancer types.

Our study also demonstrates that elderly patients, both minorities and whites, were strikingly underrepresented compared with their younger counterparts. Although the elderly accounted for approximately one third of participants in breast, lung, colorectal, or prostate cancer trials, they accounted for approximately two thirds of patients for these 4 cancer types. This strikingly low participation rate has been remarkably consistent over the past decade, as prior work has demonstrated that elderly patients represented approximately one third of trial participants in 1992, 1993-1996, and 1997-2000.^7,11,23 In effect, there has been little progress in the representation of elderly patients in cancer trials over the past 10 years. Our finding that elderly minorities are particularly underrepresented underscores the importance of identifying obstacles to trial participation for this group.

Of note, sex disparities were a function of age. Men and women 30 to 64 years of age with colorectal or lung cancers were equally likely to participate in trials; older men were more likely to enroll than older women. Our results, although limited to 2 cancer types, are consistent with recent reports of underrepresentation of women in heart failure trials and suggest that there is an important interaction between age and sex in trial enrollment.⁸ Further studies are needed to better understand the social, economic, logistical, and other factors that are responsible for the disproportionately low enrollment of elderly women.

Past discussions and research on minority enrollment have highlighted several barriers that are more prevalent among specific minority groups.²⁴ Past and ongoing patterns of discrimination have also resulted in decreased trust in the health care system in general among minority patients.^12,25,26 Minorities are more likely to express concerns about exploitation, dishonesty regarding risks of experimental treatment, and motivations of researchers.^27-30 Other studies have suggested that minority patients may be less likely than whites to be offered trial participation.³¹ Socioeconomic factors also present a substantial barrier, as the disproportionately low income of many minorities may be associated with lack of access to health care in general and subsequently to clinical trials.²⁴ Social and cultural factors are also important, as language and cultural differences can hinder communication between investigators and prospective participants.^26,32,33 Race-related variability in stage of diagnosis and stringent trial exclusion criteria may also contribute to disparities in trial eligibility and participation.^11,27

Applying trial results to patients who would not have been eligible to participate has been associated with harm.³⁴ It has been promulgated that "appropriate" representation of specific patient subpopulations is necessary to further understanding of race/ethnicity–based differences in presentation, prognosis, and response to therapy.^35-40 The most commonly used measure of "appropriateness" is proportionality—that the distribution of trial participants across race or age groups should mirror the distribution of cancer in the general population.^6-8 This is notably distinct from the "valid analysis" measure of inclusion that forms the foundation of the NIH Revitalization Act, which requires inclusion of adequate numbers of minority patients to " . . . conduct valid analyses of the intervention effect in subgroups."¹ However, it is unlikely that trials that include participants with the same distribution as can be found in the community will have sample sizes large enough for subgroup analysis.^41,42 Future work should explore what degree of representation is "optimal," as inadequately powered subgroup analyses serve little purpose beyond hypothesis generation.⁴¹ While ensuring that trial results are generalizable to the population is imperative, it is still unclear how best to achieve this goal. In some instances, it might be beneficial to design trials specifically to address clinical subgroups in instances when differences in presentation, prognosis, and response to therapy have been suggested by preliminary data. For instance, some studies should be designed exclusively for racial groups that face a higher incidence of specific cancer types or for elderly patients who are more likely to have significant comorbidities that affect presentation or response to therapy.

We focused on NCI trials because NCI is the single largest sponsor of cancer trials and has focused on ensuring access to trials for minority and elderly patients. The NCI group clinical trials database covers the majority of federally funded cancer trials, but it does not include NCI-funded noncooperative group trials, or trials funded by private industry, nonprofit foundations, or other branches of the government. We focused our analysis on 4 cancers not only because they were the 4 leading causes of cancer death, but also because they include more than half of NCI group trial participants.⁷ However, the types of cancer and trials we analyzed were not inclusive and thus may not be representative. It is also possible that inaccurate recording of race and ethnicity in either SEER or CTEP data could affect the accuracy of our results.⁴³ Finally, our sample size for the Hispanic, American Indian/Alaskan Native, and Asian/Pacific Islander populations was relatively small, making our results for these groups less robust.

Whether the rationale for including minorities, women, and the elderly is framed in terms of equity in access to trials, generalizability of the results, or the need for valid subgroup analyses, our findings suggest that the emphasis placed on increasing minority and elderly trial participation over the last decade has fallen short of the mark. Increases in NCI funding in the late 1990s and other efforts to enhance trial participation have enhanced the overall trial enrollment volume, but these gains disproportionately accrued to white patients. These findings are of concern given the high profile of minority and elderly underrepresentation and the efforts that the NCI has made to address this problem.^3,44-46 To ascribe trial designs or inadequate efforts by the NCI as the culprits would be an oversimplification, as the barriers likely include variations in patient health status, disease characteristics, attitudes, beliefs, sociodemographic factors, and access to care. Given the substantial cancer burden borne by minorities and the elderly, it is apparent that other policies and initiatives will be required to ensure broad access to trials and broad applicability of their results.