Demographic comparison of subjects in FDA approval trials in the United States to disorder specific demographics using Real World Data

Abstract

The Food and Drug Administration (FDA) has recommended that clinical trial study populations accurately reflect the patients likely to use the product, if approved. The FDA has not provided specific guidance on how cohort sizes of clinically relevant demographic characteristics should be determined. Therefore, the present study was designed to compare demographic characteristics reported in US-only FDA approval trials to the demographic characteristics of the related medical disorders in an electronic health records database of >150 M patients in the United States (US). The results demonstrate that comparative disparities in demographic cohort proportions are common, yet inconsistent, and highlight the need to define disorder specific demographic cohort proportion goals in future clinical trials.

1. Introduction

The Food and Drug Administration (FDA) has stressed the need to ensure that clinical trial study populations are representative of the patients likely to use the product, if approved [1,2]. Multiple recent publications have raised concerns that significant underrepresentation of some demographic cohorts exist in clinical trials [[3], [4], [5], [6]]. However, it remains unknown if a consistent pattern of demographic disparities exists across all clinical disorders, since the demographics of medical disorders can vary significantly between conditions. Therefore, the goal of the present study was to assess the demographic characteristics of clinical trial populations from US-only, FDA approval trials compared to the demographic characteristics of the medical disorders derived from Real World Data (RWD) in an electronic health records database of more than 150 M individuals in the US.

2. Methods

2.1. FDA approvals clinical Trial demographic data

Since 2016, the FDA has published Annual Drug Trials Snapshots Summary Reports that contain demographic data from subject populations studied in FDA approval trials [7]. From 2017 to 2020, the FDA Summary Reports [[8], [9], [10], [11]] included the percentage of subjects involved in approval trials within the US. Demographic data in these 4 annual FDA Reports consisted of 5 specific cohorts: Women, White, Black or African American, Asian and Hispanic. These data were abstracted and used in the present analysis.

2.2. TriNetX Analytics Network data

The TriNetX Analytics Network [12] database contains electronic health records from >150 million individuals obtained from>70 Health Care Organizations (HCOs) across the US. The database includes commercial, Medicaid, Medicare, and VA providers and is comprised of anonymized diagnoses, procedures, medications, laboratory values and genomic information. TriNetX, LLC is compliant with the Health Insurance Portability and Accountability Act (HIPPA), the US federal law which protects the privacy and security of health care data, and any additional data privacy regulations applicable to the contributing HCO. Because this study used only de-identified patient records and did not involve the collection, use or transmittal of individually identifiable data, Institutional Review Board approval was not required.

Database queries were generated for ICD-10 diagnostic codes that were consistent with the FDA approved clinical indication. Demographic data were obtained from US-only database entries for the specific ICD-10 codes. The query results provided data on the sex, race, and ethnicity of patients in the TriNetX dataset with the specific ICD-10 code.

2.3. Comparative analysis

Cohort proportion sizes were used to calculate the proportion ratio of the FDA approval trial data compared to the RWD from the TriNetX database. Quantitative definitions of the terms “under-representation” and “over-representation” have yet to be provided by the FDA. Therefore, in the present study, “under-representation” of a trial cohort proportion size vs. the RWD cohort proportion size was defined, arbitrarily, as a cohort proportion ratio <0.75 and “over-representation” of a trial cohort proportion size vs. the RWD cohort proportion size was defined, arbitrarily, as a cohort proportion ratio >1.25.

3. Results

3.1. Demographics of US only subject populations from FDA approved therapeutics from 2017 to 2020

There were 207 new therapeutic agents approved by the FDA from 2017 to 2020 listed in the FDA Drug Trials Snapshots Summary Reports [[8], [9], [10], [11]]. Amongst this group, there were 27 product approvals based on clinical trials that were performed with 100% of the subjects having been studied in the US.

Each of the 27 US-only FDA approval trials were then evaluated to determine if the approval indications were associated with a specific ICD-10 diagnostic code. A total of 17 (of the 27) US-only FDA approval trials were for clinical indications that were related to specific ICD-10 Codes. The abstracted data consisted of the proportion of 5 separate demographic subject cohorts in each of the 17 clinical development programs, as shown in Table 1.

Table 1.

Demographic Characteristics of US-only Subjects in FDA Approval Trials vs. RWD from the TriNetX Database (as of September 2023).

FDA Approved Indication (ICD-10 Code)	Women			White			Black or African American			Asian			Hispanic
	% in FDA Approval Trials	% in RWD (TriNetX)	Ratio Trials/RWD	% in FDA Approval Trials	% in RWD (TriNetX)	Ratio Trials/RWD	% in FDA Approval Trials	% in RWD (TriNetX)	Ratio Trials/RWD	% in FDA Approval Trials	% in RWD (TriNetX)	Ratio Trials/RWD	% in FDA Approval Trials	% in RWD (TriNetX)	Ratio Trials/RWD
Treatment of head lice (B85.0)	84	81	1.04	97	86	1.13	1	9	0.11	0	3	0.00	66	23	2.91
Treatment of breast cancer (C50)	99	98	1.01	76	83	0.91	7	13	0.53	3	4	0.83	7	7	0.94
Detection of specific cancer lesions in men with prostate cancer (C61)	0	1	0.00	85	82	1.04	2	16	0.13	4	2	1.66	4	6	0.67
Treatment of blastic plasmacytoid dendritic cell neoplasm (BPDCN) in adults and children (C86.4)	23	35	0.66	90	85	1.06	4	10	0.40	3	3	1.20	10	13	0.77
Lowering the blood levels of phenylalanine in adults with phenylketonuria (PKU) (E70.0)	50	52	0.96	98	92	1.07	1	5	0.19	0	3	0.00	3	8	0.37
Treatment of symptoms associated with opioid withdrawal during abrupt opioid discontinuation (F11.23)	28	44	0.63	67	83	0.81	22	15	1.49	0	1	0.43	17	12	1.41
Treatment of schizophrenia (F20)	24	40	0.60	21	57	0.37	75	40	1.89	1	2	0.42	9	11	0.83
Treatment of postpartum depression (F53.0)	100	99	1.01	61	77	0.79	36	18	2.02	1	4	0.28	17	15	1.14
Treatment of tardive dyskinesia (G24.01)	42	40	1.05	57	80	0.71	39	16	2.42	0	2	0.21	28	7	4.20
Treatment of acute migraine (G43)	85	78	1.09	75	83	0.90	20	14	1.38	2	2	0.83	19	11	1.76
Treatment of acute migraine (G43)	89	78	1.14	82	83	0.99	15	14	1.04	1	2	0.42	15	11	1.39
Treatment of open angle glaucoma (H40)	62	57	1.09	73	70	1.04	25	24	1.05	2	5	0.40	19	9	2.15
Maintenance treatment of adults with a lung disease called COPD (J44.9)	51	52	0.99	90	84	1.07	9	14	0.66	1	2	0.44	4	4	0.99
Treatment of irritable bowel syndrome with constipation (K58.1)	82	82	1.00	64	85	0.75	31	13	2.42	3	2	1.29	28	7	4.14

3.2. TriNetX database data on specific clinical disorders

As of September 2023, there were >150,000,000 patients with demographic data in the TriNetX Analytic Network database derived from >75 Health Care Organizations in the US. Patient demographic data were obtained via database queries for each of the specific ICD-10 diagnostic codes associated with the approved FDA therapeutics, as shown in Table 1. The ICD-10 diagnostic codes used to screen the TriNetX database are also provided in Table 1, along with the associated RWD data on each of the 5 demographic cohorts.

3.3. Comparative demographic cohort analysis of US approval trials vs. Real World data

The proportion ratio of FDA US-only trial subjects to demographic data from the TriNetX electronic health records database is also provided in Table 1. There were wide variations observed in the proportional representation of each of the 5 demographic cohorts across the multiple diagnoses. The proportion ratio of the trial subject cohorts vs. the RWD patient cohorts ranged from 0.00 (i.e., no cohort subjects were in the trials, but were present in the RWD) to 4.20 (i.e., 4.2 times as many cohort subjects were in the trials compared to their proportion in the RWD).

A graphical summary of the demographic cohort ratios is shown in Fig. 1. Of the 85 cohorts across the 17 clinical development programs, each demographic cohort was both over-represented and under-represented in certain trial populations compared to the RWD estimates with one exception: White cohorts were not over-represented in any of the 17 study populations.

Fig. 1.

Relative Proportions of Demographic Cohorts in FDA US-only Approval Trials vs RWD from the TriNetX Database.

Women and White subjects were the most represented demographic groups in the 0.75–1.25 proportion range. Asian subjects were the most under-represented demographic group (in 11/17 trials), overall, and Hispanic subjects were the most over-represented group (in 10/17 trials). Black or African Americans subjects were the second most frequently under-represented cohort (in 6/17 trials) and were also the second most frequently over-represented cohort (in 7/17 trials).

Overall, less than half of the 85 demographic cohorts (in 38/85 trials) had proportion ratios that were between 0.75 and 1.25 of the cohort proportions in the RWD. None of the 17 US-only trial populations analyzed had all 5 demographic cohort proportion ratios that were between 0.75 and 1.25 of the RWD cohort proportions.

4. Conclusions

The major finding of the present study is that sex, race and ethnic disparities exist in clinical trial populations in the US. However, no consistent pattern of disparities could be identified amongst the 5 demographic cohorts analyzed since the pattern of “representativeness” varied significantly between the various clinical trial populations. While US-only, industry sponsored trials for therapeutics approved by the FDA between 2017 and 2020 were reasonably diverse in terms of demographic cohort proportions, when analyzed at the disorder level, wide demographic cohort disparities were sometimes observed.

Notably, 4 of the demographic 5 cohorts were sometimes over-represented (i.e., cohort proportion ratios >1.25) and all 5 of the cohorts were sometimes under-represented (i.e., cohort proportion ratios >0.75) across the disorder specific subject populations. Asian populations were the most under-represented, consistent with previous observations [13], indicating that recruitment efforts may need to be enhanced to this population.

None of the 17 US-only trial populations analyzed had proportion ratios within the 0.75–1.25 range for the 5 demographic cohorts expected based on the RWD electronic health record data from over 150 M Americans [12]. The FDA has yet to comment on the acceptable amount of variance from epidemiological or RWD that would be considered acceptable for therapeutic approvals in the future. The data in the present study demonstrate that demographic parameters should be determined on a disorder specific basis when developing diversity goals for future clinical trials.

CRediT authorship contribution statement

Stephen J. Peroutka: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The author declare that he has no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.