Representation of Minoritized People with Cystic Fibrosis in CF Therapeutics Development Network Clinical Trials

Abstract

Rationale:

Black and Hispanic people with CF (PwCF) experience worse health outcomes related to pulmonary disease compared to White, non-Hispanic PwCF. Improving outcomes in minoritized PwCF is a crucial need in the CF community, but prior evidence suggests that these communities are underrepresented in CF clinical research.

Objective:

To determine whether CF clinical studies are representative of the overall CF population.

Methods:

The study cohort included participants from recent cystic fibrosis transmembrane conductance regulator modulator studies (CHEC-SC, PROMISE, PROSPECT, and GOAL). Persons seen at the same centers who were eligible for these studies based on inclusion and exclusion criteria, but who did not enroll in these studies were used as a comparator cohort from the CFF Patient Registry (CFFPR). Rates of study enrollment were assessed separately for Black race and Hispanic ethnicity versus non-Hispanic White PwCF. Propensity weighted Poisson regression models were used to account for potential confounders in the relationship between race and study participation.

Results:

Demographic characteristics were compared between 3,594 PwCF in the four modulator studies with 14,888 individuals from the CFFPR who were eligible, but did not enroll. Enrollees were younger (median age 19 years), more likely to have had at least three clinic visits per year, and live within 30 miles of the study site. Black and Hispanic PwCF were less likely to have private insurance and less likely to live far from study centers. Multivariable analysis revealed that Black individuals were significantly less likely than non-Hispanic White individuals to enroll in CHEC-SC and PROMISE, with a 17% and 32% lower likelihood, respectively. Hispanic individuals had a 34% lower enrollment rate than non-Hispanic White individuals in PROMISE but higher enrollment in PROSPECT. Differences in enrollment rates were greater at sites with fewer minority participants.

Conclusion:

These results suggest that Black and Hispanic PwCF are, overall, significantly less likely to be included in CF clinical research than non-Hispanic White PwCF. Future research is needed to better understand if these differences are upheld in studies with varying inclusion/ exclusion criteria and clinical sites.

Background:

Clinical trials in cystic fibrosis (CF) have led to dramatic advancements in CF therapeutics and clinical care that have substantially improved survival and morbidity in CF related lung disease(1, 2). Engaging in clinical research has offered significant benefits for individuals with CF. Participants in new drug studies, such as those studying the effects of cystic fibrosis transmembrane conductance regulator (CFTR) modulators, gain access to life prolonging therapies years ahead of the general population (3). Beyond this, clinical trials are vital in evaluating the efficacy of new therapies, identifying adverse effects, and expanding understanding of disease processes.

A comparison of CF clinical study participants to an overall population from the CF Foundation (CFF) Patient Registry showed that research participation in CF is high, with an overall estimated 30% participation rate (4). Importantly, this study showed that research participants were more likely to have private insurance and to identify as White race. Between 1999 and 2015, approximately 80% of CF pharmacology clinical trials did not describe race or ethnicity of participants(5). Among studies where race and ethnicity data were reported, 94.4% of study participants were identified as White and 24.4% of studies included no racially minoritized People with CF (PwCF). This work, though informative, occurred prior to the advent of CFTR modulator therapy and at a time of very different clinical outcomes. More recently, a 2023 single center study found that 35.9% of non-Hispanic White PwCF participated in onsite clinical trials while 21% of PwCF identifying with racially and ethnically minoritized groups participated in onsite studies(6). This study found that no racially or ethnically minoritized PwCF participated in off-site clinical trials. Despite clinical research being extremely important to advance clinical outcomes and provide access to therapies, there is clear evidence for lack of equitable enrollment in clinical studies for minoritized PwCF.

In 2022, 8.8% of PwCF in the CFF Patient Registry Annual Report were identified as race other than White and 10.0% of individuals were identified as Hispanic ethnicity (1). Although Black and Hispanic individuals make up an increasing share of the CF population, they face greater health challenges associated with the condition. This includes poorer nutritional status and reduced lung function among Black CF patients, a threefold higher risk of mortality among Hispanic patients, and an increased likelihood of acquiring pulmonary infections at a younger age in Hispanic patients(7–10). Lack of representation in clinical studies compounds these existing inequities. Because CFTR modulators are genotype-specific therapies, eligibility for clinical studies evaluating these treatments is often restricted to individuals with these specific CF causing genetic variants. As a result, racially and ethnically minoritized people with CF who are statistically less likely to carry genotypes responsive to current modulators are disproportionately excluded from these trials(11, 12).

Although previous research has highlighted non-representative inclusion in CF studies, a comprehensive, multi-center evaluation of participant representation in CF clinical research has yet to be conducted in the modern, post-modulator era. This study evaluated the equity of racial and ethnic representation in CF modulator studies in the CF Therapeutics Development Network (CF TDN), a large clinical trial network comprised of over 90 study sites conducting more than 250 observational and interventional studies since 1998. Our study compares participation by race and ethnicity in CF TDN studies to an overall population available in the CFF Patient Registry. Portions of this work have previously appeared in the author’s published thesis.

Methods:

Study Population and Data Sources

Study eligible PwCF were identified using data from the CFF Patient Registry which captures encounter based clinical data for approximately 85% of the US CF population(13). We used study data from the CF TDN, merging study data from four large, recent observational studies which included participants who received ivacaftor, lumacaftor/ ivacaftor, tezacaftor/ ivacaftor, or elexacaftor/ tezacaftor/ ivacaftor. Studies included Characterizing CFTR Modulated Changes in Sweat Chloride and their Association with Clinical Outcomes (CHEC-SC, NCT 03350828); A Prospective Study to Evaluate Biological and Clinical Effects of Significantly Corrected CFTR Function (PROMISE, NCT04038047); Prospective Longitudinal Study of CFTR-dependent Disease Profiling in CF (PROSPECT), and the G551D Observational Study (GOAL, NCT 01521338) (see Supplemental Table e1 for study characteristics and eligibility criteria) (14–17). CHEC-SC (n= 2,998) did not have specific mutation requirements, PROMISE (n=523) included those with at least one copy F508del, PROSPECT (n=209) required two F508del mutations, and GOAL (n=216) included those with G551D, R117H or other non-gating mutations. We selected these four observational studies because they were large, recent, and provided detailed participant-level data. Comparable data from industry-sponsored interventional modulator trials were not accessible. Observational studies additionally have relevance to real world participation trends and are more generalizable to the broader CF population. Finally, assessing the clinical effects of modulators is especially important given their impact on CF care. To analyze factors associated with study participation, we identified persons eligible for each study based on inclusion and exclusion criteria from CFF Patient Registry captured during the years of each study’s accrual (see Supplemental Table e1). For participants in the four studies, we captured CFF Patient Registry data from the year of enrollment. For eligible persons who did not enroll, we randomly selected CFF Patient Registry data from one of the years in which they were eligible for the study. Persons eligible for more than one study were included in the evaluation of the most recent study. Baseline data reported corresponds to year of enrollment for enrolled PwCF and random year selected and eligible for those who did not enroll. The study was approved by the University of Washington Institutional Review Board (STUDY00004003).

Statistical Analyses

We captured race from the CFFPR, entered through a combination of self-reported and staff entered data with pre-specified, fixed categories. We followed precedent by McGarry and McColley and categorized people based on a combination of race and ethnicity (5, 11). We compared enrollment of PwCF identified as either Hispanic ethnicity, non-Hispanic Black race, or non-Hispanic White race.

To identify whether race was associated with study participation in a causal way, we constructed a causal inference diagram to identify potential confounders and to distinguish them from mediators and other relationships (Figure 1). Using Daggity, we determined that bias in identifying the direct, causal association of race with study participation could be minimized by adjusting for these measures: demographics, lung function as measured by percent predicted forced expiratory volume in one second (ppFEV1), CFTR genotype, socioeconomic status, CF related diabetes status, visit frequency, and exacerbation frequency(18, 19). Socioeconomic status variables included highest education level of patient and patient’s parents, employment, and insurance type. Distance to study center as calculated using zip code was additionally included. We approximated distance from individual residences to study sites by computing the distance between the latitude and longitude at the centroid of each zip code. We used the software python, pgeocode package, which pulls zip code information from the GeoNames database (20, 21). In regression, the distances were categorized as: <=30 miles, 31 to 60 miles, and >60 miles. For each study separately, we constructed logistic regression models with race or ethnicity as the outcome and included the above listed covariates to build the propensity for the exposure of interest (race or ethnicity). We built propensity scores separately for Black race versus non-Hispanic White race, and then for Hispanic ethnicity versus non-Hispanic White race.

Figure 1. Example Causal Diagram: association of race/ ethnicity with study participation.

In a second step, weighting on these resulting propensity scores to control for confounding, we then performed Poisson regressions that evaluated the association between race or ethnicity and the outcome of interest (study participation)(22). The exposure of interest, race or ethnicity, was the only fixed effect. Study site number was included as a random effect, as well as inverse weighting on propensity of the exposure (race or ethnicity). Inverse probability weighting equalizes the groups of different races or ethnicities over all included confounders and serves to make the groups of different races or ethnicities otherwise comparable for “risk” of the outcome (study participation). Poisson regression provides effect estimates in a more interpretable format than logistic regression: risk ratios instead of odds ratios. In these analyses, we excluded sites for whom no persons of that race/ethnicity were reported in the CFF Patient Registry (where the probability of enrolling a person of minoritized race/ethnicity was zero). Because inclusion and exclusion criteria varied between studies and different sites enrolled in each, we analyzed each study separately (see Supplemental Table e1). Analyses were conducted using R software (R version 4.4.3).

While the random term for site permits differences in enrollment rates geographically, we also wondered whether the number of persons of minoritized race or ethnicity in the site catchment area (eligible persons) might be related to any observed differences by race or ethnicity. Therefore, we additionally assessed whether any association between race or ethnicity and enrollment frequency might depend on the number of eligible persons of the race or ethnicity of interest. We did this using interaction terms between the race of the eligible CFFPR registrant and a categorical variable for number of persons of that race or ethnicity at that site.

Results:

Cohort Characteristics

We identified 3,594 unique persons who were enrolled into CHEC-SC (n= 2,998), PROMISE (n=523), PROSPECT (n=209), or GOAL (n=216) and who also had CFF Patient Registry data collected in the same year. We also identified 14,888 persons who were eligible for at least one study including genotype eligibility based on CFF Patient Registry data during the time of the studies. Characteristics of the persons enrolled and persons eligible but not enrolled are shown in Table 1. Median age for those enrolled in the modulator studies was 19 years versus 25 years for those not enrolled. A greater proportion of those enrolled in modulator studies had at least three annual visits at baseline (82% enrollees versus 54% those not enrolled) and lived within 30 miles of the study site (51% those enrolled versus 36% those not enrolled). Among 142 sites enrolling in studies, 109 had at least one Black PwCF eligible and 126 had at least one Hispanic PwCF eligible. The associations of characteristics with Black race and with Hispanic ethnicity as computed for the propensity scores are shown in Supplemental Tables e2 and e3. In regression models, 8% of persons could not be included due to missing data: 3% of persons had unknown ethnicity, 4% had unknown BMI, and 1% had unknown values for each of diabetes at baseline, visits in the current year, and zip code. As 13% of persons had unknown FEV1, those persons were retained in analyses and a category for missing FEV1 was created.

Table 1.

Baseline characteristics by whether ever enrolled into any modulator study or participating in the CFFPR and eligible but not enrolled in study.

Baseline characteristics	Modulator study eligible at CFFPR site but not enrolled N=14,888	Modulator study enrollee N=3594
Male	7,698 (52%)	1,821 (51%)
Race / Ethnicity
Non-Hispanic White	12,932 (87%)	3,168 (88%)
Non-Hispanic Black	322 (2%)	51 (1%)
Other race	328 (2%)	68 (2%)
Hispanic	800 (5%)	201 (6%)
Unknown ethnicity	506 (3%)	106 (3%)
BMI category
Underweight	885 (6%)	170 (5%)
Normal	9213 (62%)	2471 (69%)
Overweight	2849 (19%)	674 (19%)
Obese	1138 (8%)	258 (7%)
Unknown	803 (5%)	21 (1%)
Genotype
F508 Heterozygous	7,703 (52%)	2026 (56%)
F508 Homozygous	6,202 (42%)	1307 (36%)
G551D (not F508)	174 (1%)	108 (3%)
R117H (not F508)	678 (5%)	114 (3%)
Other	131 (1%	39 (1%)
Age (median, Q1, Q3)	25 (15, 36)	19 (14, 28)
Age category
<10	1993 (13%)	447 (12%)
10 to 19	3305 (22%)	1521 (42%)
20 to 29	3935 (26%)	863 (24%)
30 to 39	2946 (20%)	432 (12%)
40 to 49	1456 (10%)	188 (5%)
50+	1253 (8%)	143 (4%)
Insurance
Private	8942 (60%)	2423 (67%)
Other	5378 (36%)	1141 (32%)
Unknown	568 (4%)	30 (1%)
Education
Up to high school	2917 (18%)	533 (15%)
At least some college	5793 (39%)	1234 (34%)
Unknown	6376 (43%)	1827 (51%)
Parent Education
Up to high school	1848 (12%)	456 (13%)
At least some college	5581 (38%)	1663 (46%)
Unknown	7459 (50%)	1475 (41%)
FEV1% predicted (median, Q1, Q3)	83 (61, 98)	89 (70, 102)
FEV1% predicted
< 40	1142 (8%)	196 (6%)
40 to 89	5924 (40%)	1543 (43%)
90+	4746 (32%)	1648 (46%)
Unknown	3076 (21%)	207 (6%)
Exacerbation history at baseline
None	11804 (79%)	2643 (74%)
1 to 2	2407 (16%)	764 (21%)
At least 3	677 (5%)	187 (5%)
At least 3 annual visits at baseline	7990 (54%)	2961 (82%)
CF-related diabetes at baseline
No	10,738 (72%)	2828 (79%)
Yes	3874 (26%)	761 (21%)
Unknown	276 (2%)	5 (<1%)
Minimum distance to study site*
< 30 miles	5369 (36%)	1849 (51%)
30 to 59 miles	2934 (20%)	704 (20%)
60 + miles	6465 (43%)	1029 (29%)
Unknown	120 (1%)	12 (<1%)

^*Computed between location at centroid of home zip code to that of study participating study sites for the relevant (eligible) study.

In evaluating associations of Black race with potential confounders, Black PwCF were significantly less likely to have private insurance compared to non-Hispanic White individuals (see Supplemental Table e2). Across studies, the relative risk of having private insurance for Black individuals ranged from 0.17 (95% CI: 0.04–0.66) to 0.22 (95% CI: 0.16–0.31). In PROMISE and CHEC-SC, Black PwCF had higher rates of CFRD compared to non-Black individuals, with relative risks of 2.01 (95% CI: 1.48–2.72) and 1.89 (95% CI: 1.36–2.64), respectively. In PROSPECT, CHEC-SC, and PROMISE, Black PwCF were 76%, 71%, and 77% less likely to live more then 60 miles from a study center, respectively. Hispanic PwCF were less likely to have private insurance with risk ratios ranging from 0.12 to 0.44 (95% CI: 0.04–0.56). (see Supplemental Table e3). In PROSPECT, CHEC-SC, and PROMISE, Hispanic PwCF were also significantly less likely to live more than 60 miles from a study center, with reductions of 47%, 45%, and 36%, respectively.

Multivariable Analyses

After adjustment for measured confounders using inverse weighting in the Poisson regression analysis, Black persons were 17% less likely than White persons to be enrolled in CHEC-SC (RR=0.83, 95% CI 0.77, 0.90; Table 2a) and 32% less likely to be enrolled in PROMISE (RR=0.68 95% CI 0.57 to 0.81), which was exacerbated when there were few Black persons at the site (Table 2b). In PROMISE (N=523), there was evidence of a lower rate of enrollment in Hispanic persons relative to non-Hispanic White PwCF (RR=0.66 95% CI 0.56, 0.79), which was also most pronounced in sites with fewer Hispanic persons (Table 2b). In CHEC-SC (N=2,998), RR for study enrollment among Hispanic relative to non-Hispanic White persons was 0.94 (95% CI 0.89, 0.99). After stratifying on number of Hispanic persons at the site, there was evidence of lower rate of enrollment among sites with few (i.e., 1 to 3 persons) Hispanic persons at the site (RR=0.66, 95%CI 0.59, 0.74). There was evidence of significantly lower enrollment rate among Black PwCF in PROSPECT (RR=0.33, 95% CI 0.21, 0.53; Table 2a). This was not found in GOAL (N=216) among Black PwCF compared to the reference group, however confidence intervals were wide (RR=1.22, 95% CI 0.8, 1.74). In PROSPECT (N=209), enrollment of Hispanic PwCF compared to non-Hispanic White patients was higher (RR=2.04, 95% 1.64, 2.53). In GOAL, there was no significant difference in enrollment by ethnicity (RR=1.08, 95% 0.78, 1.58). Of note, different study centers participated in the four studies, and those who participated in multiple studies were not necessarily found to have contributed consistently to the number enrolled (see supplemental figure e1 – heat map).

Table 2a.

Association of race and ethnicity with study enrollment, both unadjusted and inverse propensity (IP) weighted for known and measured confounders.

Risk ratio for enrollment	CHEC-SC (n= 2,998)	PROMISE (n=523)	PROSPECT (n=209)	GOAL (n=216)
Black vs White
Unadjusted	0.71 (0.51, 0.98)	0.81 (0.45, 1.46)	1.08 (0.34, 3.45)	1.18 (0.51, 2.75)
IP weighted	0.83 (0.77, 0.90)	0.68 (0.57, 0.81)	0.33 (0.21, 0.53)	1.22 (0.86, 1.74)
Hispanic vs Non-Hispanic White
Unadjusted	0.85 (0.72, 1.00)	0.88 (0.57, 1.36)	1.93 (1.09, 3.41)	1.49 (0.77, 2.89)
IP weighted	0.94 (0.89, 0.99)	0.66 (0.56, 0.79)‡	2.04 (1.64, 2.53)	1.08 (0.73, 1.58)

^‡For PROMISE because of convergence issues, we excluded 17 sites with only 1 eligible Hispanic person at those sites.

Table 2b.

Regression results for Poisson regression stratified by numbers of person at each site of that race/ethnicity, inverse propensity (IP) weighted for known and measured confounders

IP weighted risk ratio for enrollment	CHEC-SC	PROMISE ѱ	PROSPECT	GOAL ѱ
Black vs White
1 to 2 POC at site	0.38 (0.32, 0.46)*	0.54 (0.43, 0.67)	1.14 (0.54, 2.40)	1.17 (0.63, 2.17)
3+ POC at site	1.00 (0.92, 1.08)	1.00 (0.75, 1.32)*	0.20 (0.11, 0.35)*	1.21 (0.79, 1.85)
Hispanic vs Non-Hispanic
1 to 3 POC at site	0.66 (0.59, 0.74)*	0.65 (0.53, 0.80)	1.27 (0.73, 2.21)	1.50 (0.73, 3.07)
4+ POC at site	1.06 (0.99, 1.13)	0.98 (0.78, 1.22)	2.21 (1.74, 2.82)*	0.93 (0.59, 1.46)

^*Interaction between influence of race/ethnicity and number of Black or Hispanic PwCF at the site on enrollment was significant, p<0.05, relative to the reference category which is always lowest number of POC at site.

^ѱPROMISE and GOAL sites tended to have more Black and Hispanic POC, and therefore the higher categories for those studies were 5+ Black POC and 9+ Hispanic POC, respectively.

Discussion:

After controlling for genotype eligibility and adjusting for key confounding variables, we found that study enrollment rate of Black PwCF was lower than would be expected as compared to the CFFPR in two large modulator studies, the CHEC-SC and PROMISE studies. This association was stronger when there were proportionally fewer Black PwCF represented at respective registry sites. Similarly, we found that the enrollment rate for Hispanic PwCF was lower than non-Hispanic White PwCF in the PROMISE study and this association was also most pronounced at sites with proportionally fewer Hispanic persons. Our sample of studies enrolled across a large period, with PROMISE and CHEC-SC enrolling more recently in 2018 through 2021. We did not see a trend that earlier studies, such as GOAL (enrolling in 2012), were less representative by race or ethnicity.

Beyond CF, studies in cardiology and oncology have identified restrictive eligibility criteria, homogenous study leadership, and burdensome follow up processes as barriers to inclusive enrollment(23–25). Historic injustices in early clinical trials and race-based clinical equations, continue to influence clinical trial participation today(26, 27). Limited access to specialized research centers, inadequate insurance coverage, and logistical barriers like transportation disproportionately affect enrollment among marginalized populations (24, 28, 29). A lack of diversity in the research workforce and disparities in funding further exacerbate these gaps (23, 30). Among Hispanic/Latinx communities, language barriers, implicit bias, and immigration-related concerns can impede participation, and the frequent treatment of this population as homogeneous overlooks meaningful diversity in national origin, language, and health beliefs (32–35). These factors are particularly relevant in diseases like CF, where genetic variation across subgroups may influence disease outcomes (34).

In these four relatively large, modern CFTR modulator studies, research participants were more likely to have private insurance and attend three or more clinic visits annually. Yet, lower enrollment among Black and Hispanic PwCF persisted even after adjusting for these and other important socioeconomic and health-related factors. In our cohort, Black and Hispanic PwCF were less likely to live further away from study centers, a pattern consistent across these modulator studies. This is notable because distance is often cited as a key barrier to study participation(25). Our findings suggest that factors other than proximity to study center contribute to these disparities. In health services research, frameworks for healthcare “access” extend beyond accessibility based on physical distance or transportation to include other considerations such as whether study resources and interactions are culturally relevant (acceptability), whether out of pocket costs limit participation (affordability), and how well study procedures align with participants’ schedules or responsibilities (accommodation)(35). These broader dimensions of access may help explain why closer proximity did not translate into higher enrollment among Black and Hispanic PwCF and more work is needed to explore these barriers precisely.

When comparing across studies, enrollment was not necessarily representative even in studies with a lower burden of follow up like CHEC-SC where there was only one study visit for sweat chloride collection. Furthermore, findings were not uniform across studies: the PROSPECT study had higher enrollment of Hispanic persons and GOAL did not have significant differences in enrollment. The demonstrated variability in enrollment of minoritized participants across studies may simply reflect study size and power (GOAL included only 216 PwCF). However, these findings could reflect differences in site level enrollment rather than differences between overall study practices (supplemental figure e1: heat map). Qualitative methods may be required to further understand differences between site practices including contextual factors, workflow of recruiting teams, staff perspectives, etc.

Looking forward, solutions to improve representation in clinical studies have not been explored in CF specifically. In 2024, the FDA updated guidance to increase clinical trial diversity by broadening eligibility, engaging communities, offering cultural competency training, and ensuring geographic access for ethnically and racially minoritized populations(36). Our finding that disparities were greatest at sites with fewer Black and Hispanic PwCF suggests that sites with greater experience caring for diverse populations may already employ more inclusive recruitment practices. More research is necessary to understand variability in recruitment practices and learn from sites enrolling diversely.

A novel and important strength of our study is the ability to identify individual persons at each study site who were eligible but did not enroll. This uniquely allows us to construct detailed, site-level and individual-level characteristics for this non-enrolled cohort, which provides a more precise method of accounting for confounding than using population averages. Additionally, we incorporated eligibility and CFTR genotype criteria into our analysis to ensure that comparisons were made among individuals who were truly eligible for each study. From these data we cannot discern whether these patients were approached or not and if they declined participation or were not given the opportunity.

Our study had several limitations. These relationships are complex and socioeconomic status may act as a mediator in the association between race and outcomes. However, since both socioeconomic status and race share a common origin in historical and ongoing structural racism (see Figure 1), socioeconomic status also serves as a confounder in the association between reported race and study enrollment. Thus, there is no analytic approach that allow us to distinctly separate the influences of socioeconomic status and race. Collapsing ethnic and racial groups with different experiences of racism into pre-set categories adds bias and imprecision. Aggregation of ethnic and racial groups may in fact mask differences in health burdens between individual cultural groups(37). Additionally, misclassification of race or ethnicity may occur based on reporting in registry or study data. Importantly, TDN sites are a subset of all CF centers, and our findings may not represent trends at all CF sites. While CFFPR data are included for most PwCF (approximately 77–85% depending on year), there may be impactful differences in individuals not participating in the registry.(38) For example, those of older age born prior to widespread newborn screening and those of minoritized race are less likely to be diagnosed and therefore less likely to be included in the CFFPR. Similarly, the CF TDN comprises a large number of CF centers but does not include all CF centers or all healthcare settings caring for PwCF. One limitation of inverse probability weighting is that it does not guarantee causal inference (39). It relies on two key assumptions: 1) all confounders must be measured and included in the model, and 2) like other regression methods, IPW can be biased if the model is misspecified, meaning for example it incorrectly assumes a linear relationship when the true relationship is more complex (e.g., quadratic). Finally, it is important to highlight that, unlike clinical trials which are experimental, these studies are observational and may offer different levels of individual benefit or appeal to patients and require varying degrees of commitment from investigators to enroll minoritized populations.

Conclusion:

We found evidence of lower enrollment rate among Black/ African American and Hispanic PwCF in recent CF modulator studies. Enrollment disparities were more pronounced at sites with fewer minoritized people. The association persisted despite adjusting for important clinical, demographic, and social variables including distance from study center. These findings underscore the need for more inclusive recruitment strategies to ensure diverse representation in clinical studies. Importantly, more recent federal policy changes removing FDA guidance to promote diversity may reduce institutional accountability and reduce resources dedicated to inclusive research practices. Future research should explore the underlying mechanisms and study design factors contributing to these disparities as well as barriers people experience to enrollment.