Enrollment of women in heart failure randomized controlled trials.
Enrollment of nonwhites in heart failure randomized controlled trials.
Explicit exclusion in heart failure randomized controlled trials.
Mean age of patients enrolled in heart failure randomized controlled trials.
Mean age and participation of women in heart failure randomized controlled trials.
Heiat A, Gross CP, Krumholz HM. Representation of the Elderly, Women, and Minorities in Heart Failure Clinical Trials. Arch Intern Med. 2002;162(15):. doi:10.1001/archinte.162.15.1682
Appropriate representation of specific groups of patients in randomized controlled trials (RCTs) and generalizability of evidence obtained have been questioned. We sought to compare the characteristics of patients in heart failure (HF) RCTs with those of patients with HF in the community, and to document these trends over time.
MEDLINE was searched from 1985 through December 1999 for all HF RCTs. Additional trials were obtained through bibliographies of the reviewed articles, previous meta-analyses, and overviews of the HF trials. Original reports of RCTs of interventions for chronic HF were selected if they were published in the English language and included 50 or more participants. Publication year, sources of support, location of principal investigator, sample size, type of intervention, inclusion and exclusion criteria, characteristics of participants, and the use of invasive diagnostic tests were extracted.
Participants in the 59 HF RCTs we reviewed were markedly different from the patients with HF in the community. Patients in RCTs were younger, more often male, more likely to have subnormal systolic ejection fraction, and were most commonly white. We did not find a significant improvement in the representation of the trials, with respect to these characteristics, among those published recently compared with those from the late 1980s and early 1990s.
Clinical trials are focusing on a relatively small segment of the HF population. The consequences of underrepresenting minorities, women, and elderly are unknown but may be particularly important for HF. Future clinical trials should adequately include populations that carry the burden of the disease.
RANDOMIZED controlled trials (RCTs) are considered to provide the highest level of evidence for clinical practice, but several studies have suggested that RCTs often do not enroll subjects who are representative of patients seen in the community.1- 6 The infrequent enrollment of older patients, women, and minorities has raised specific concerns about the generalizability of medical evidence to these important patient groups. In response to concerns about clinical trial generalizability and equity, the US Congress included requirements for the inclusion of women and minorities, but not the elderly, in the National Institutes of Health (NIH) Revitalization Act of 1993 (Public Law 103-143).7,8
This issue is particularly important for heart failure (HF), which is primarily a disease that affects older adults and may manifest itself differently in the elderly. More than 80% of patients hospitalized with HF in the United States are older than 65 years and more than 20% are 85 years or older.9 Evidence suggests that older patients who have HF are more likely to have diastolic dysfunction, as well as a higher prevalence of comorbid illnesses than their younger counterparts.
Heart failure also significantly impacts women and minorities; women constitute almost 57% of patients with HF in hospitals9; 30% of these patients are nonwhites.10 This heterogeneity is clinically relevant, as prior studies have suggested that characteristics and outcomes of HF may vary with sex and race.11- 18
Despite the burden of illness that HF imposes on members of these underrepresented populations, little attention has been focused on their representation in HF clinical trials and how their representation has changed over time. To address this issue, we performed a systematic review of the RCTs that enrolled patients with HF from 1985 through 1999. Our objective was to determine the enrollment of elderly, women, and minorities in those trials, examine secular changes, and assess the representation of trial subjects compared with patients with HF in the general population.
We performed a systematic search of English-language literature to identify all major clinical trials of treatment for chronic HF. MEDLINE was searched from January 1985 through December 1999 for articles indexed under the subject headings heart failure and randomized controlled trials or clinical trials. Additional trials were obtained from the reference lists of the reviewed articles, previous meta-analyses, and overviews of the HF trials. These led to 642 titles. We examined the title and abstract of those articles based on our exclusion criteria and excluded 460 articles; 182 remaining articles needed complete evaluation. Another 137 articles were excluded after close and entire evaluation. We selected 45 studies. We obtained an additional 14 trials from the reference lists of the reviewed articles and previous meta-analyses and overviews of the HF trials. Our search led to 59 clinical RCTs on the treatment of chronic HF.
We selected RCTs of drugs, procedures, lifestyle-related interventions, and health services for treatment of chronic HF if they were published in the English language and included 50 or more participants. Trials were excluded if they were currently ongoing, were subgroup or follow-up analyses of the original study, were reporting on the prevention or incidence of HF, or were focused on patients with a recent myocardial infarction or acute HF as a result of cardiac arrest or cardiogenic shock.
We developed an instrument for data abstraction and a data dictionary with detailed standardized definitions. We extracted the following information: year of publication, sources of support, location of principal investigator, sample size, type of intervention, inclusion and exclusion criteria, demographics and characteristics of the study sample, and invasive diagnostic tests used as part of the study protocol.
Heart failure RCTs were assigned to 3 five-year periods (1985-1989, 1990-1994, and 1995-1999), according to their year of publication. Intervention type for each trial was categorized as medications, procedures, lifestyle-related programs (patient education, diet, and exercise), or health services (disease management/home-based interventions, care coordinator, and use of critical pathways/clinical guidelines). Specific attention was focused on the age, sex, and race distribution of study participants and whether these characteristics were the source of any explicit exclusion in these trials. For each trial, the first complete published report was our source of information. However, for completeness, other published articles on the study, including subgroup analyses, were also reviewed to supplement the original report.
We studied the general characteristics of HF trials and evaluated their variations over time. Mean age was calculated as weighted mean, considering the sample size of each trial. We also compared the characteristics of trial participants with those of patients in the general population in the United States. The latter was obtained from the existing national or community-based studies.10,19- 24 The χ2 test was used to evaluate differences for categorical variables. To test for a trend in demographic characteristics of the patient samples across our study period, we performed 3 simple linear regression tests using year of publication as the independent variable. The dependent variables were the mean age of participants, the proportion of women, and the proportion that were nonwhite; each study was weighted by the sample size. The simple linear regression tests were also used to examine the association between age and sex. Results are also presented in 3 five-year periods (1985-1989, 1990-1994, and 1995-1999), according to the year of trial publication for the purposes of illustration. All calculations were performed using the software system STATA 5.0 (STATA Corp, College Station, Tex). P values of .05 or less were considered to be of statistical significance.
A total of 59 RCTs met the entry inclusion criteria; their characteristics are summarized in Table 1. Fewer HF RCTs were performed in the earlier periods compared with the latest period (14, 17, and 28 for 1985-1989, 1990-1994, and 1995-1999, respectively). The principal investigators for most HF clinical trials were located in the United States during the 1985 through 1989 and 1990 through 1994 periods, and in Europe during 1995 through 1999. Among the 47 trials for which sources of support were reported, 38 (81%) received complete or partial financial support from the pharmaceutical industry.
A total of 45627 patients were randomized in these 59 trials; the number of participants per trial ranged from 50 to 7788 (median, 253). Heart failure trials tended to include a larger number of patients over time. Studies published between 1995 and 1999 included more than 65% of participants in all RCTs included in this study. Eighty-eight percent of HF trials studied the effectiveness of drugs, 5% assessed health services interventions, 5% studied lifestyle-related interventions, and 2% evaluated different procedures. Of note, health services and lifestyle-related interventions, as well as procedures, were studied exclusively in trials published from 1995 through 1999. Invasive diagnostic procedures were used in only 4 (7%) studies as part of the study protocol.
The mean age of the HF study population (reported in 53 trials [90%]) was 61.4 years (SD, 6.4). Participation of patients older than 80 years in HF trials was very poorly documented; information on the age distribution was reported for only 13 (22%; Table 1) trials. Only 4 trials (7%) with 2176 participants (5%) included patients older than 80 years and the exact number of these patients was not reported. Another study, which did not report the age range but reported on the age distribution, included 282 patients and 50% of these patients were 79 years or older. Only 2 studies (7%), with a total of 1004 participants, exclusively enrolled patients older than 65 years.
Information on sex was provided in 57 studies (97%) and women constituted 21% of study participants in those trials (Figure 1). Twelve studies (20%), which included 51% of the total number of participants, provided information on racial distribution (Table 1). Only 15% of participants in those studies were nonwhite (Figure 2). None of the trials specifically targeted women or nonwhite populations.
Six trials, which included 15000 patients, were totally or partially funded by the NIH. Patients in those studies had a mean age of 62.3 years and 21% were women. In the 4 NIH-funded studies that reported racial distribution, about 16% of participants were nonwhite. In addition to 1 trial that was funded by the Department of Veterans Affairs and the NIH, the Department of Veterans Affairs participated in financial support for 4 trials (including 2570 patients with a mean age of 62.1 years); 0.2% of these participants were women and 28% were nonwhite.
Explicit exclusion of elderly patients was part of the formal study protocol in 17 trials (29%); these trials accounted for more than 20000 patients (44% of total number of HF trial participants; Figure 3). The age threshold for exclusion in those trials was as low as 70 to 75 years in 8 studies, 80 years in another 7 studies, and from 81 to 85 years in 2 others. Four studies (7%), including 3 that were supported by the Department of Veterans Affairs, excluded women as part of their protocol. They included 4% of the total number of participants in HF trials. Eight additional trials, which included 13% of the remaining HF study participants, excluded women with childbearing potential (Figure 3). None of the 59 trials had explicit exclusion of minorities as part of their formal protocol.
A total of 37 trials (63%), including 32000 (70%) of the overall trial participants, had explicit exclusion criteria based on left ventricular ejection fraction (EF); most of these (32 of 37) excluded patients with an EF greater than 40%. The number of trials that performed formal exclusion based on EF increased substantially over time (21%, 71%, and 79% for 1985-1989, 1990-1994, and 1995-1999 periods, respectively; Figure 3). There has been a significant increase in the practice of EF-based exclusion in the latest period, from 1995 through 1999, compared with the earlier trials, from 1985 through 1994 (P = .03).
Renal insufficiency was part of the formal exclusion criteria in 11 studies (19%) that included more than 19 000 (44%) of the HF clinical trial participants. Again, the number of trials that use this exclusion criterion was higher in the latest period compared with the 2 other periods (Table 1). In addition, 34 trials (58%) with 24 573 participants (54%) excluded patients who had important comorbidities that were generally defined as major hepatic, renal, or hematologic diseases.
The mean age of trial participants increased from 58.5 years (SD, 5) during 1985 through 1989 to 59 years (SD, 4) during 1990 through 1994, and 64 years (SD, 7) during 1995 through 1999 (Figure 4). Linear regression test showed that there was a significant increase in mean age of the participants over time (P = .001). Earlier HF trials were less likely to report mean age compared with the recent trials (71%, 88%, and 100% for the 1985-1989, 1990-1994, and 1995-1999 periods, respectively). However, only a small number of investigators provided information on age distribution and participation of older patients in the trial, and this did not improve over time (29%, 18%, and 21% for the 1985-1989, 1990-1994, and 1995-1999 periods, respectively). The number of trials that explicitly excluded older patients was smaller during 1995 through 1999 compared with 1990 through 1994 (27% vs 35%), but was higher when compared with the 1985 through 1989 period (21%; Figure 3). Nevertheless, there was no statistical difference between the number of trials that performed this explicit exclusion in 1995 through 1999 compared with the 2 earlier periods combined (27% vs 29%, respectively) (P = .86).
Participation of women in the HF trials decreased in the 1990 through 1994 period compared with the 1985 through 1989 period (17% vs 23%) and returned to its initial level during 1995 through 1999 (23%; Figure 1). Regression analysis showed that there was no significant change in the proportion of women enrolled in the HF trials over time (P = .84). However, the exclusion of women by protocol design appears to have become slightly less frequent. We also found that enrollment of women was significantly associated with the mean age of participants in HF trials (Figure 5; P<.001). We calculated that a 1-year increase in the mean age of trial participants was associated with a 5% augmentation in enrollment of women in these trials.
Documentation of participation of minorities in the HF trials improved over time, as more trials reported the racial distribution of their participants over the study period (7%, 24%, and 25% during 1985-1989, 1990-1994, and 1995-1999, respectively). Inclusion of nonwhite populations in the HF trials that provided this information decreased steadily from the 1985 through 1989 to the 1990 through 1995 and 1995 through 1999 periods (27%, 18%, and 13%, respectively; Figure 2). However, regression analysis did not show a significant change in the proportion of nonwhites (P = .17). Of note, only 10 trials were included in this analysis, which probably explains the lack of statistical significance.
A community-based study in Olmsted County, Minnesota, using the Rochester Epidemiology Project, found that the mean age of patients receiving the first diagnosis of HF was 77 years (SD, 12).20 Evidence from other studies in Europe also suggests that the mean age of HF patients in the community is older than 75 years,22,23 which is substantially higher than the mean age (61 years) that we found for patients in HF trials (Figure 4). In fact, only 3 trials (5%) included patients with a mean age older than 75 years.
Women were underrepresented in the HF clinical trials when compared with their representation in the general population of HF patients, as reported in the National Health and Nutrition Examination Survey and surveys from Europe (21% vs at least 50%, P<.001; Figure 1).22- 24 Only 2 trials (3%) had patients with a mean age of 75 years or older, and had at least a 50% female composition.
Finally, there was a substantial underrepresentation of nonwhites in trials in comparison with the distribution of nonwhite patients with HF published in the Vital and Health Statistics series (15% vs ≈ 30%, P<.001; Figure 2).10 Only 4 trials (7%) included at least 25% nonwhite patients.
Fewer than 30% of patients with HF enrolled in HF trials had preserved systolic EF. However, results of community and hospital-based studies from the United States and Europe persistently show that patients with preserved systolic EF constitute more than 50% of patients with HF.19,23,25,26
Our findings reinforce current concerns about the representation of specific patient groups in RCTs. We showed that participants in HF clinical trials differ markedly from those of the general population of patients with HF in the United States. Compared with the latter, RCT patients are younger, more often male, more likely to have subnormal systolic EF, and most commonly white. We found that only 2 of the 59 trials reflected the general presentation of community patients with respect to both sex and age, only 2 studies exclusively targeted patients older than 65 years, and none particularly targeted women or nonwhite populations. We also demonstrated that members of these populations were underrepresented in government-funded trials (eg, those funded by the NIH and Department of Veterans Affairs). In addition, we did not find a significant improvement in population representation in trials published in recent years compared with those from the late 1980s and early 1990s.
As many of the trials were multicenter and multinational, we could not evaluate the representation of nonwhites in each trial compared with the racial distribution of the population in the individual trial's location. However, our goal was not to evaluate or criticize each individual trial. Rather, we aimed to document the paucity of data specific to important subgroups of patients with HF in the United States. Our study shows an important discrepancy between the general characteristics of the HF trial enrollees and those of US patients with HF in the community.
Some experts have suggested that enrolling a narrow spectrum of patients might affect the external validity or generalizability of RCTs, which have been hailed as the gold standard for establishing clinical efficacy.6 Previous studies have compared the enrollment in RCTs of specific population groups with their representation in the general population of patients with the same conclusion. These studies found that members of specific population groups were not adequately represented in clinical trials.1,27 Gurwitz and colleagues2 found that women and the elderly were widely excluded in clinical trials of acute myocardial infarction from 1960 through 1990. A recent study on the enrollment of women in National Heart, Lung, and Blood Institute–funded cardiovascular clinical trials showed that women were underrepresented in HF trials.3 Additional studies have revealed a substantial underrepresentation of elderly in cancer trials.4,5 Two studies compared the characteristics and outcomes of trial patients with acute myocardial infarction with those of nontrial patients at the study hospitals and external hospitals in the same area, or with those eligible but not enrolled in the same hospital. Those authors have raised concerns about the practice of preferential recruitment in RCTs and reported that trial participants were younger, were more often male, had fewer comorbid conditions, and had a survival advantage over the nonparticipants that was larger than expected from the intervention alone.28,29
The selected populations in the RCTs we studied appear to be the result of both explicit and implicit exclusion criteria. We found that a large number of trials explicitly excluded women and elderly patients. We also found that even when age-based exclusion was not part of the formal study design, mean age of the trial participants was markedly lower than what has been reported by community-based studies. The skewed trial populations may be the result of excluding patients with conditions that are disproportionately found in specific groups (eg, the elderly) or by implicit exclusions derived from a reluctance to enroll certain patients.
We also found that long lists of comorbid conditions and medications were part of exclusion criteria that would most certainly lead to preferential recruitment of younger patients in trials. Furthermore, the presence of EF-based exclusion might widely influence the enrollment of elderly patients and women in HF clinical trials, as studies have consistently showed that patients with HF, with preserved systolic EF, are more likely to be older and female.19,20,23,25,30 Although preserved left ventricular systolic function is a common condition among patients with HF,19,20,23,25,26,31- 33 HF trials in general did not include patients with preserved systolic EF.
There are numerous factors that discourage the enrollment of eligible patients. There may be considerable difficulty in obtaining consent from certain groups of patients. Prior studies have demonstrated considerable concern in the African American community about the ethical conduct of research.34 Important sociocultural factors, particularly considering the legacy of the Tuskegee Syphilis Study, have led many to be wary of the motivations of researchers.35 Alternatively, elderly patients may be less likely to participate in trials compared with younger patients because they have decreased mobility and limited knowledge of trials.5,36,37 Furthermore, the concentration of trials in certain centers that may have access to a narrow spectrum of patients might lead to the practice of preferential recruitment.
There is likely a relation between the participation of women in trials and the mean age of trial participants. Given that women constitute a large proportion of patients with HF in older age groups, noninclusion of the elderly may influence the inclusion of women in trials and vice versa. We showed that the mean age of participants in HF trials was directly related to the proportion of women enrolled in those trials. However, it is difficult to separate the effects of age from sex in patients with HF since they are closely related in the prevalence of this condition.
We also found that demographic characteristics of the trial participants are very poorly documented in the published reports from HF trials. We showed that only 20% of trials gave some information on the age distribution of the enrollees in their published articles. Similarly, we found very poor documentation on the participation of minorities. It is noteworthy that a substantial proportion of the HF trials were performed in countries other than the United States, which may have different distributions of racial and ethnic groups. However, because many of the trials were multicenter and multinational, we could not evaluate the representation of minorities in the trials located exclusively in the United States. Sex was the only demographic characteristic that was adequately described in the HF trials. Poor documentation of demographic characteristics of trial participants makes it very difficult for readers to determine the population representation and generalizability of trial results.
There are several reasons to strive for representative groups of trial participants. First, as participants often receive the latest medical treatments, access to trials may be considered a marker of equity. For some patients, access to a clinical trial is synonymous with access to quality care.38 Prior studies have suggested that patients in clinical trials, even in the control arms, tend to have better outcomes than patients not enrolled in trials.39,40
Second, it is important to note that HF may manifest differently in various patient populations. Certain aspects of HF, including pathophysiology, response to treatment, clinical course, and outcomes in cardiovascular diseases, vary as a function of age, sex, and race.41- 52 Age- and sex-adjusted normative values have been suggested for different cardiovascular and pulmonary tests.53- 55 In the setting of HF, age, sex, and race differences have been demonstrated in left ventricular adaptation and autonomic responses as well as in the clinical course, epidemiology, and outcomes of the disease.11- 18 Those differences may affect the efficacy of therapeutic strategies for patients with HF.
It would be difficult for every trial to recruit a sample that is exactly representative of the general population. Further, it could be argued that unless the sample size is large enough to allow valid subgroup analysis of the patient groups of interest, a representative sample may not be helpful. However, when the complete portfolio of clinical research in a given illness is considered, as we have done with HF in this analysis, the population could be better served by a more representative group of patients in trials. To accomplish this, investigators might develop trials that specifically target underrepresented groups of patients instead of designing trials to mirror the distribution of patients in the community.
Our findings suggest that published clinical trials are focusing on a relatively small segment of the HF population. The current evidence concerning HF has been mainly generated by trials on male, middle-aged, and white patients and may not apply to a significant proportion of patients with HF. The consequences of underrepresenting minorities, women, and the elderly are not known but may be particularly important for HF. We elucidated the large gap that exists between HF trial participants and patients with HF in the community. This gap is very likely to keep growing along with the growing burden of HF due to the aging of the population and the longer survival of patients with hypertension and myocardial infarction. New strategies need to be used to generate evidence that is directly relevant to the vast majority of HF patients. Future clinical trials should adequately represent the populations who carry the burden of the disease. This would be accomplished most effectively not only by striving for representative participation in individual trials, but by designing trials to specifically investigate HF treatment strategies in subgroups of interest.
Accepted for publication January 31, 2002.
We thank Yun Wang, MS, for his contribution in developing the database and his assistance with data management and data analysis.
Corresponding author: Harlan M. Krumholz, MD, Yale University School of Medicine, 333 Cedar St, PO Box 208025, New Haven, CT 06520-8025.