eMethods. Trends in EID over time with regression analysis
eTable. Comparison of racial/ethnic composition of trials leading to drug approval with the expected trial demographic distribution in the US based on published 5-Category Race/ethnic prevalence data
eFigure 1. Enrollment Incidence Disparity (EID) and Enrollment Incidence Ratio (EIR) for each pharmaceutical agent relative to expected NEI prevalence rates.
eFigure 2. Enrollment Incidence Disparity (EID) and Enrollment Incidence Ratio (EIR) for each pharmaceutical agent relative to expected prevalence from Stein et al. 2011, Vanderbeek et al. 2011, and US Census data
Customize your JAMA Network experience by selecting one or more topics from the list below.
Berkowitz ST, Groth SL, Gangaputra S, Patel S. Racial/Ethnic Disparities in Ophthalmology Clinical Trials Resulting in US Food and Drug Administration Drug Approvals From 2000 to 2020. JAMA Ophthalmol. 2021;139(6):629–637. doi:10.1001/jamaophthalmol.2021.0857
Were clinical trials leading to US Food and Drug Administration (FDA) ophthalmology drug approvals representative of the racial/ethnic distribution in the US from 2000 to 2020?
In this cohort study of 31 clinical trials of 13 drugs with 18 410 participants, the racial/ethic distribution of trial participants was different from the expected distribution for 12 drugs, with Black, Hispanic or Latinx, and other non-White participants being underrepresented.
In this cohort study, racial/ethnic distribution of participants in clinical trials leading to FDA ophthalmology drug approvals differed from the expected disease burden and racial/ethnic distribution in the US, suggesting that further efforts to increase enrollment of minority groups in clinical trials is warranted.
Diverse, representative enrollment in pivotal clinical trials is vital to sufficiently power subgroup analyses and ensure equity and validity of trial results.
To evaluate the racial/ethnic representation, trends, and disparities in clinical trials leading to US Food and Drug Administration (FDA) ophthalmology drug approvals from 2000 to 2020.
Design, Setting, and Participants
This cohort study used data from participants in clinical trials of drugs for neovascular age-related macular degeneration (AMD), open-angle glaucoma (OAG), and expanded indications for diabetic retinopathy (DR) from January 1, 2000, to December 31, 2020. Trial data were sourced from FDA reviews, ClinicalTrials.gov, and relevant linked studies. National expected racial/ethnic proportions were sourced from public National Eye Institute prevalence data as well as published rates scaled using US Census Bureau data.
Main Outcomes and Measures
The primary outcome measures were the distribution of and change over time in the racial/ethnic proportion of participants in clinical trials leading to FDA approval of drugs for AMD, OAG, and DR.
During the 20-year period, 31 clinical trials were identified for 13 medications with 18 410 participants. The distribution of trial participants was different from the expected trial distribution for most approvals with regard to race/ethnicity (12 drugs) and sex (10 drugs). Compared with the first decade (2000-2010), trials conducted in the second decade (2011-2020) showed increases in enrollment of Asian (odds ratio [OR], 2.30; 95% CI, 1.97-2.68; P < .001) and Hispanic or Latinx participants (OR, 1.74; 95% CI, 1.49-2.03; P < .001) for AMD, Asian participants (OR, 2.21; 95% CI, 1.46-3.42; P < .001) for DR, and Black (OR, 1.60; 95% CI, 1.43-1.78; P < .001) and Hispanic or Latinx participants (OR, 10.31; 95% CI, 8.05-13.35; P < .001) for OAG. There was a decrease in Black participants in DR trials (OR, 0.58; 95% CI, 0.42-0.79; P < .001). Based on these trends, the enrollment incidence ratio is expected to worsen by 2050, with overrepresentation of white participants vs underrepresentation of Black and Hispanic or Latinx participants in trials of drugs for AMD (1.08 vs 0.04 vs 0.77), DR (1.83 vs 0.87 vs 0.59), and OAG (1.62 vs 0.90 vs 0.37).
Conclusions and Relevance
In this cohort study, Black, Hispanic or Latinx, and other non-White participants were underrepresented in clinical trials leading to FDA ophthalmology drug approvals compared with the expected disease burden and racial/ethnic distribution in the US. Although there was meaningful improvement from 2000 to 2020, further efforts to increase minority enrollment in clinical trials seem to be warranted.
Epidemiological studies have demonstrated myriad racial/ethnic and sex variations in the prevalence and disease course of common ocular diseases, including open-angle glaucoma (OAG), age-related macular degeneration (AMD), and diabetic retinopathy (DR). For example, there have been significant racial/ethnic differences in the rates of OAG, AMD, and DR among minority groups in landmark studies, including the Baltimore Eye Survey,1,2 the Barbados Eye Study,3-5 the Salisbury Eye Evaluation Project,6-8 the Proyecto Ver Study,9-11 the Los Angeles Latino Eye Study (LALES),12-14 the Multi-Ethnic Study of Atherosclerosis15,16 as well as studies using data from the National Health and Nutrition Examination Survey.17-19 These population-based epidemiological studies established critical racial and demographic differences, but historically they have not been sufficiently nuanced to be representative of the entire US population.20
Given the importance of access to emerging therapeutics, efforts have been made to address the impact of racial/ethnic disparities in clinical research during the past 30 years.21-23 The US Food and Drug Administration (FDA) specifically addressed inclusion of demographic subgroups in clinical trials in 2012,24 with guidance for National Institutes of Health (NIH)–defined phase 3 clinical trials in 201725 and suggestions to promote inclusivity in November of 2020.26 Still, a review27 of race/ethnicity reporting in the ophthalmology literature suggested that only 43% of manuscripts in ophthalmology journals in 2019 included race and/or ethnicity data. A prior study28 on sex and ethnic composition of clinical trials for ophthalmological new molecular entities (NMEs) found no significant change in demographic composition from 2006 to 2016. This finding could be problematic if the trial composition is not representative of the US population, with similar evidence of disparities in clinical trials across various medical and surgical fields.29-33
In this study, we evaluated potential disparities in race/ethnicity and sex between reported disease prevalence in the US population and a cohort of patients enrolled in clinical trials ultimately leading to FDA approval for NMEs and expanded indications. We specifically performed our analysis for the leading causes of blindness: neovascular AMD, DR, and OAG.
This cohort study did not qualify as human subject research and was deemed to be exempt from institutional review board approval by the Vanderbilt University institutional review board, with a waiver of informed consent. The study adhered to the tenets of the Declaration of Helsinki34 and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Drugs identified by NME and basic licensing agreements from January 1, 2000, to December 31, 2020, were searched on Drugs@FDA database.35 For diabetic eye disease, the anti–vascular epithelial growth factors aflibercept and ranibizumab with recent efficacy data and new indications for diabetic macular edema or related retinopathy were included and limited to the phase 3 efficacy studies referenced on the FDA drug label. Demographic data were sourced from the FDA statistical reviews baseline demographics table, FDA medical reviews, ClinicalTrials.gov, linked publications, and/or the new drug application. For sex categories, whether the data were self-reported or investigator determined could not be ascertained.
The NIH advocates for common language for racial/ethnic categories. Hispanic or Latinx is an ethnicity that is not dependent on race; race categories include American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White.36 Before widespread adoption of this convention, Hispanic was listed as an exclusive race category for several trials and the reference demographic benchmark studies. For these cases, the remaining participants were considered non-Hispanic. Therefore, we used the term Hispanic or Latinx as a separate category in racial category analysis for comparability in this analysis. American Indian or Alaska Native and Native Hawaiian or Other Pacific Islander were grouped into the “other” race category because they were inconsistently reported in clinical trials and in the reference literature. Of importance, identification of race/ethnicity, sex, and gender were fluid during the study period, with concomitant expansion and variation of the reported demographic categories. Therefore, the categories used reflect the categories that remained consistent during the study period; whether the demographic data were self-reported or investigator determined could not be ascertained.
The expected disease prevalence for each disease state, stratified by race and sex, was sourced from the National Eye Institute reported eye health data,37 which were based on US Census data and studies on OAG,38 DR,39,40 and AMD.41 To avoid the effects of large sample size using the US population as a reference, an expected trial participant distribution was created for each NME using an equivalent number of total participants. For example, if a trial enrolled 1000 participants, the actual demographic distribution was compared with the expected demographic distribution for a 1000-person trial using national prevalence data. To assess the difference in enrollment over time, the current demographic distribution of trials in the study period were assumed to remain constant over time. These distributions were compared with projected prevalence rates for 2030 and 2050 from the National Eye Institute and were subject to their methodologic assumptions.37
Stata/IC, version 16 (StataCorp LLC) was used to analyze the data. Statistical significance was set at a 2-sided P = .05. The racial/ethnic composition of trials leading to each pharmaceutical agent’s approval was compared with the expected prevalence based on the US population using the χ2 test. Proportions and odds ratios (ORs) with corresponding 95% CIs were calculated for approvals in the first (2000-2010) and second (2011-2020) decades to assess the trends in composition over time.
Trials were pooled by disease category to compare the enrollment incidence disparity (EID) and enrollment incidence ratio (EIR), which were calculated as described elsewhere.29 The EID represents the absolute difference between the proportion of patients of a particular race/ethnicity among trial participants and the estimated proportion of patients of that particular race who received a diagnosis of a specific eye condition among the US population. The EIR represents the proportion of patients of a particular race/ethnicity among trial participants divided by the estimated proportion of patients of that particular race who received a diagnosis of a specific eye condition among the US population. Supplemental regressions were used to explore the association between year of approval and trial composition (eMethods and eFigures 1 and 2 in the Supplement).
The National Eye Institute data pooled Hispanic as a separate race category and did not contain prevalence rates for the category of Asian. Therefore, a secondary analysis used prevalence rates from a separate epidemiological study that reported prevalence rates for Asian American individuals in the US with OAG42 and exudative AMD43 in the same population. These prevalence rates were mapped to the most recent US census44 to create an analogous reference US population prevalence (eTable in the Supplement).
During the 20-year period (2000-2020), 31 clinical trials were identified for 13 medications with 18 410 participants. Ten trials supported approval of an NME for neovascular AMD (brolucizumab, aflibercept, ranibizumab, pegaptanib sodium, and verteporfin), and 16 trials supported approval of an NME for OAG (unoprostone isopropyl, travoprost, bimatoprost, tafluprost, latanoprostene bunod, and netarsudil); 5 trials supported expanded approval of existing NMEs for sequela of diabetic retinopathy (aflibercept and ranibizumab) (Table 1).
National Eye Institute population data were leveraged to calculate expected distribution of clinical trial cohorts based on contemporaneous demographic data. The actual racial/ethnic distribution of trial participants was different from the expected trial demographic distribution for most approvals (12 drugs). Likewise, there were significant differences in sex distribution compared with the expected sex distribution for most approvals (10 drugs) (Table 2).
The trials supporting approvals during the first decade (2000-2010) and second decade (2011-2020) by disease category showed increases for enrollment of Asian individuals (287 [5.19%] vs 472 [11.16%]; OR 2.30, 95% CI, 1.97-2.68; P < .001) and Hispanic or Latinx participants (315 [5.69%] vs 402 [9.51%]; OR 1.74, 95% CI, 1.49-2.03; P < .001) for AMD NMEs. Similarly, there was an increase in Asian participants (32 [4.22%] vs 112 [8.86%]; OR 2.21, 95% CI, 1.46-3.42; P < .001) in DR trials. However, there was a decrease in black participants in DR trials (93 [12.25%] vs 94 [7.44%]; OR 0.58, 95% CI, 0.42-0.79; P < .001). In OAG trials, there was an increase in Black (682 [14.55%] vs 930 [21.38%]; OR 1.60, 95% CI, 1.43-1.78; P < .001) and Hispanic or Latinx participants (74 [1.58%] vs 617 [14.19%], OR 10.31, 95% CI, 8.05-13.35, P < .001) (Table 3).
For each disease category, the EID and EIR showed forecasted overrepresentation of White participants as well as underrepresentation of Black and Hispanic or Latinx participants for projected years 2030 and 2050. The EIDs are expected to worsen by 2030 and 2050, with overrepresentation of white participants vs underrepresentation of Black and Hispanic or Latinx participants in trials of drugs for AMD (2030: 5.03% vs −4.67% vs −0.24%; 2050: 6.97% vs −4.36% vs −1.89%), DR (2030: 23.73% vs −1.25% vs −4.44%; 2050: 36.20% vs −1.39% vs −14.64%), and OAG (2030: 21.50% vs −2.34% vs −6.50%; 2050: 29.83% vs −1.90% vs −13.12%) (Table 4). The EIRs are also expected to worsen by 2030 and 2050 in trials of drugs for AMD (2030: 1.06 vs 0.04 vs 0.96; 2050: 1.08 vs 0.04 vs 0.77), DR (2030: 1.42 vs 0.88 vs 0.83; 2050: 1.83 vs 0.87 vs 0.59), and OAG (2030: 1.38 vs 0.88 vs 0.54; 2050: 1.62 vs 0.90 vs 0.37) (Table 5).
This cohort study revealed that, from 2000 to 2020, enrollment of racial/ethnic groups in clinical trials leading to FDA approvals of drugs for AMD, DR, and OAG was different from the expected distribution based on disease burden in the US population. Disparity subanalyses showed overrepresentation of White participants and underrepresentation of black and Hispanic or Latinx participants compared with the expected enrollment. Furthermore, the enrollment appeared to change, with increased enrollment of Asian participants in AMD and DR trials, increased Hispanic or Latinx participant enrollment in AMD and OAG trials, and increased Black participant enrollment in OAG trials. However, there was no change in enrollment of Black participants in AMD trials over time and a relative decrease in Black participants in DR trials. These results suggest potentially skewed trends in racial/ethnic diversity of clinical trial participants in pivotal ophthalmology trials.
Racial/ethnic variation in the prevalence of AMD,2 OAG,1,3 and DR4 has been known and reported before the present 20-year study period, although guidance on reporting has changed over time. The National Center on Minority and Health Disparities was established in 2000.22 In 2012, §907 of the FDA Safety and Innovation Act directly addressed demographic subgroup reporting in clinical trials.24 The required analysis of available subgroup data for the Center for Biologics Evaluation and Research applications in 2011 suggested underrepresentation of certain groups with inconsistencies between race composition in FDA applications and the expected disease prevalence based on US Census Bureau statistics.45 Similarly, the NIH Revitalization Act of 1993 mandated the establishment of guidelines for inclusion of women and other minority groups in clinical research,21 with amendments to better define clinical research in 2001,46 guidelines for NIH-defined phase 3 clinical trials in 2017,25 and further suggestions in 2020.26 These important efforts may have contributed to some of the subsequent leveling of racial groups found in this study; however, many of the trials analyzed here were likely designed or initiated before these efforts, and there may be a meaningful delay before the impact of these guidelines is observed.
The findings of this study are consistent with disparities in clinical trials in other fields of medicine, including oncology,29 rhinology,30 neurology,31 dermatology32 and cardiothoracic surgery.33 In ophthalmology, a recent review of sex and ethnicity enrollment in trials of NMEs demonstrated that predominantly women and White individuals were enrolled in 9 studies.28 That study did not compare the enrollment population with expected US burden or stratify by disease subgroup, which may explain the lack of significant changes over time. Our analysis included expanded clinical indications and a broader 2-decade study period to adequately power an analysis of enrollment changes over time. More importantly, we provided comparisons with expected disease incidence based on previously published methods.29
In this study, enrollment disparity was found after accounting for the expected prevalence of disease. For example, approximately 89% of US persons with AMD identify as White and at least 4% identify as Black.37 However, over the past 20 years, Black participants comprised 0.177% of AMD trial participants. Given that African American individuals may be disproportionately affected by surgically treatable or preventable causes of blindness, such as cataract, glaucoma, DR,8 it is critical to promote equitable participation in trials for emerging therapeutics. This is especially important for diseases with strong racial/ethnic components of presentation and severity, such as OAG.47
Barriers to research enrollment are well studied in medicine and may be structural, clinical, attitudinal, demographic, and/or socioeconomic.48 The most concerning explanation for the underrepresentation of certain demographic groups is an underlying systemic barrier to enrollment in rigorous clinical studies. Financial resources, transportation, employment, and other factors may additionally prevent the consistent follow-up required for clinical trial protocols. These many obstacles to successful participation in clinical research may in part account for the suboptimal outcomes seen in real-world settings.49,50 We believe a goal for investigators should be to reduce barriers to enrollment in ophthalmology clinical trials.
In addition to access issues, participation in clinical studies requires trust in the health care system and sufficient health literacy to personally evaluate a complex risk-reward tradeoff. Certain populations may have reservations about participating in trials involving non-FDA approved therapies. For example, for prostate cancer, Black men were found to be more likely than White men to harbor suspicion of the health care system, and this was associated with less willingness to discuss clinical trials.51 Attitudinal barriers to clinical trial participation are not necessarily exclusive to 1 race, ethnicity, or sex but rather may be associated with numerous demographic or contextual factors. These barriers are complex and may be historically rooted, deserving attention in future studies.
Another explanation for the disparities identified in the study is inadequate demographic reporting.29 Reporting was inconsistent in the present study, with variable race/ethnic categories within the safety analysis data sets, full analysis sets, and per protocol sets. This variability was compounded by the variable sources of data available for demographic data (Table 1), and the specific data set or source may account for minor variations in subsequent analysis. Likewise, the present analysis was restricted to phase 3 studies owing to overlapping participation with earlier phase 1 and 2 studies. Although these are limitations of the present study, we believe inconsistencies in reporting may represent a need for comprehensive and standardized reporting of demographic characteristics in clinical trials. Demographic data for each participant subset, the methods for assessment, and the degree to which the subgroup analysis is underpowered should be provided for subsequent studies.
Of importance, Hispanic was often not identified as an ethnicity or was unreported, especially in earlier trials, suggesting the increased proportion of Hispanic or Latinx participants in trials of medications for AMD and OAG may be attributable primarily to the correct inclusion of Hispanic or Latinx as a nonexclusive ethnic category for reporting. This is unlikely to entirely explain the increase in Hispanic or Latinx participation, especially in trials of medications for glaucoma. The LALES was published in 200413 and demonstrated the high prevalence of OAG in the Hispanic population. In the present analysis, the early trials of NMEs for OAG (2000-2010) were clustered from 2000 to 2001, and the later trials (2011-2020) had approvals between 2012 and 2017. Therefore, the LALES and other landmark studies may be responsible for the trend in increasing Hispanic or Latinx inclusion in these clinical trials. This finding highlights the importance of promulgation of epidemiological study findings. Although the EID and EIR projections suggest an increasing potential for disparity by 2050 if current enrollment patterns are unchanged, there is an opportunity to mitigate this disparity by designing trials with attention to both disease prevalence and the changing US population.
Of note, clinical trials are typically conducted at many but not all academic medical centers and private practices across the US, often with a predominantly international paired study. The demographic characteristics of the patients at these sites would not be expected to be entirely reflective of the national population demographic features. Similarly, geographic barriers to accessing these sites52 and specific inclusion and exclusion criteria26 may have contributed to our findings. These logistic hurdles may explain the skewness in our analysis despite efforts from investigators to enroll diverse populations in their trials.
The pooling of paired clinical trials by drug may increase the diversity of clinical trials and bias the distribution with non-US based racial/ethnic groups. For example, many recent trials of medications for AMD have included international collaborators, which may have led to increases in Asian participants. Epidemiological trends and clinical profiles may differ in these population. For instance, the prevalence of AMD may be different among Chinese American individuals compared with the population in China.53 The pathophysiology and specific subtype of neovascular AMD is likely different in Asian individuals,54 but the Vascular Endothelial Growth Factor Trap-Eye: Investigation of Efficacy and Safety in Wet Age-Related Macular Degeneration (VIEW2) trial was one of the first trials with sufficient enrollment of Asian participants to allow for meaningful subgroup analysis.55 Of importance, categories such as Asian can encompass markedly heterogenous subpopulations, and thus a comparison with prevalence in the US may be substantially less meaningful. In addition, study sampling methods alone can create subgroups that are not representative random samples of the general population, preventing meaningful subgroup analysis.56
The discordance between the demographic characteristics of clinical trial participants and the expected US population distribution for each disease is meaningful. These findings require interpretation in a clinical context. Specifically, given the known variation in treatment response across racial groups, whether there is potential for clinical trial data to become skewed if the enrolled cohort is not representative of the general US population is unknown. Moreover, evaluation of the next steps in addressing unequal enrollment is needed. Demographic requirements for participants may prolong trial enrollment and delay access to vision-saving therapies. A pragmatic solution to the disparities identified would be to require reporting of underrepresentation in phase 3 clinical efficacy trials as well as timely publication.57 Furthermore, postmarket surveillance data in large unbiased cohorts can capture potential racial/ethnic, sex, or other subgroup differences in clinical outcomes. These efforts can be important because small disparities in trials of NMEs may become magnified when less rigorous off-label studies58 encourage expanded use of new therapeutics across variable populations. Regardless, the findings of the present study suggest that continuing efforts to promote engagement in clinical trial enrollment, especially in minority groups at risk of disease, are necessary.
This study has limitations. There was inconsistent and variable race/ethnic category reporting. Comprehensive uniform guidelines for reporting of demographics in clinical trials may help relieve these inconsistencies for future studies. Similarly, comparison demographic data was pulled from various datasets, which may create minor variations depending on the source.
This cohort study revealed that from 2000 to 2020, Black and Hispanic or Latinx participants were underrepresented in clinical trials leading to FDA ophthalmology drug approvals compared with the expected disease burden and racial distribution in the US. The disparity has narrowed over time, and further efforts should focus on engagement of underrepresented groups. Diverse, representative enrollment in pivotal clinical trials is vital to sufficiently power subgroup analyses and ensure equity and validity of trial results.
Accepted for Publication: February 19, 2021.
Published Online: April 22, 2021. doi:10.1001/jamaophthalmol.2021.0857
Corresponding Author: Shriji Patel, MD, Vanderbilt University Medical Center, 2311 Pierce Ave, Nashville, TN 37232 (email@example.com).
Author Contributions: Dr Patel had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Berkowitz, Patel.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Berkowitz, Patel.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Berkowitz, Patel.
Administrative, technical, or material support: Groth, Patel.
Supervision: Groth, Gangaputra, Patel.
Conflict of Interest Disclosures: Dr Patel reported receiving grants from Alcon. No other disclosures were reported.
Funding/Support: This study was supported in part by an unrestricted departmental award from Research to Prevent Blindness, Inc.
Role of the Funder/Sponsor: Research to Prevent Blindness, Inc had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.