The Black-participant share in landmark drug trials: a primer on what 'trial diversity gap' really means and how to read a race-stratified subgroup analysis

A Phase 3 major depressive disorder trial that supported the August 2022 FDA approval of Auvelity enrolled 327 patients. 116 were Black. That is 35.5 percent of the trial cohort, well above the 13.6 percent Black share of the US adult population (US Census Bureau, 2020 Decennial Census), and a figure most Phase 3 cohorts in this same primer never approach. One of them, the SKYLARK postpartum-depression trial reviewed below, did not report its cohort's race composition at all (GEMINI, Iosifescu et al, J Clin Psychiatry 2022, PMID 35649167, NCT04019704).

Across five recent FDA-decision approvals reviewed for this primer, the Black-participant share ranged from 0.67 percent to 35.5 percent. That range is the first question to bring to a prescriber. The harder question, and the one a competent prescriber should welcome, is whether the trial's race-stratified subgroup analysis was pre-specified and properly powered. For most of the drugs Black US adults are filling prescriptions for today, the answer is no.

The Black-participant share of a trial is a clinical question before it is a representation question, because race-stratified subgroup analyses, when they exist and when they are properly powered, repeatedly find different effect sizes in Black participants than the overall trial average reports. The pattern is most thoroughly documented in cardiovascular and antihypertensive trials. ALLHAT, the federally-funded outcomes trial that compared chlorthalidone, amlodipine, and lisinopril in 33,357 hypertensive participants, enrolled 15,094 Black participants and published a pre-specified race-stratified subgroup analysis in 2005 (Wright et al, JAMA 2005, PMID 15811979). The Black-subgroup result showed lisinopril produced less systolic blood-pressure reduction than chlorthalidone and higher angioedema rates than chlorthalidone or amlodipine. That subgroup finding entered clinical guidance: JNC-8 cites the ALLHAT Black-subgroup outcomes data as part of the evidence base for its 2014 recommendation that initial antihypertensive therapy in Black patients with hypertension include either a thiazide diuretic or a calcium-channel blocker rather than an ACE inhibitor (James et al, JAMA 2014, PMID 24352797).

The lesson from ALLHAT is not that race is biologically deterministic. The lesson is that population-level pharmacogenetic and pharmacodynamic differences, layered onto a US population where Black adults carry roughly twice the hypertension burden of non-Hispanic white adults (Mozaffarian et al, Circulation 2016, PMID 26673558), produce drug-response differences at the individual prescription level that get missed when a Phase 3 trial enrolls under 5 percent Black participants and does not pre-specify the race-stratified comparison. A drug that works on average in a trial population may not work on average in the population your prescriber is treating. Whether you can know that before the drug is in your system depends on whether the trial cohort and the subgroup analysis are surfaceable on the FDA label, in the published primary, or in the ClinicalTrials.gov registry.

A race-stratified subgroup analysis is the statistical test for whether the drug works the same in the Black-participant subgroup as in the white-participant subgroup or in the cohort overall. Three things have to be true for the test to give a reliable answer.

The Black subgroup has to be large enough to detect a difference. A Phase 3 trial of 500 patients with 7 percent Black enrollment yields a Black subgroup of 35. To detect a 10 percent relative effect difference between Black and non-Black participants with 80 percent statistical power, the math typically requires Black subgroup counts in the several hundreds, not in the dozens. SELECT, the trial that supported the 2024 Wegovy cardiovascular indication, enrolled 17,604 participants with 671 Black, and the registry reports the overall primary-outcome hazard ratio of 0.80 (95% CI 0.72 to 0.90) with no published race-stratified breakout in the public record (ClinicalTrials.gov NCT03574597 results module; Lincoff et al, NEJM 2023, PMID 37952131).

The analysis has to be pre-specified, not post-hoc. A pre-specified Black-vs-white comparison enters the statistical-analysis plan before the trial unblinds and counts as a confirmatory test. A post-hoc analysis is added after the data come back and counts as hypothesis-generating. Both can be useful when they are reported transparently. The methodological failure is when a sponsor presents a post-hoc finding as if it were confirmatory.

The forest plot has to be read at face value. On a forest plot, each subgroup's point estimate is a dot and the confidence interval is a horizontal line. If the line crosses the null line (typically 1.0 for hazard or risk ratios), the subgroup-specific effect cannot be distinguished from no effect at the conventional significance threshold. The most common feature of a Black-participant subgroup forest plot, in a trial that enrolled fewer than 10 percent Black participants, is a wide confidence interval crossing the null. That finding does not mean the drug does not work in Black patients. It means the trial was not powered to know.

The named methodological pitfall is the inverse: presenting "no race effect detected" from an underpowered subgroup analysis as if it were affirmative evidence of equivalent effect. Absence of evidence is not evidence of absence. The other pitfall is extrapolating from a trial population to a US-burden population the trial did not represent. Bulevirtide enrolled 1 Black participant of 150 in MYR-301, and the US chronic-hepatitis-D burden is concentrated in African-immigrant populations from sub-Saharan Africa, the Middle East, and Asia (Wedemeyer et al, NEJM 2023, PMID 37345876). The "FDA-approved" label, when it comes, does not close that epistemic gap. It leaves it open for post-marketing real-world evidence to fill.

The table makes four observations possible. Each is the kind of finding a competent prescriber should already know about the drug they are recommending.

The range is over 50-fold. Across five recent FDA-decision approvals reviewed for this primer, the Black-participant share spans from 0.67 percent (bulevirtide MYR-301) to 35.5 percent (Auvelity GEMINI). That single range is the trial-diversity gap, with each end telling a different story about how a Phase 3 cohort gets built.

Auvelity is the cleanest single-drug internal contrast. The same compound (dextromethorphan-bupropion / AXS-05) supported two different indications: major depressive disorder (GEMINI, 35.5 percent Black enrollment, approved August 2022) and Alzheimer's-associated agitation (ADVANCE Phase 2/3, 11.5 percent Black enrollment, Phase 3 program). Both trials enrolled before the FDORA 2022 Diversity Action Plan statute existed, and both exceeded the population baseline the April 2022 voluntary FDA draft guidance recommended. The 24-point spread, with the same compound at the same sponsor under the same voluntary regime, is the kind of variation a binding regime would have addressed and a voluntary one did not. Whether the spread matters for the labeled efficacy figure depends on whether each trial's race-stratified subgroup analysis was pre-specified, powered, and published.

Semaglutide is the cleanest single-molecule scale contrast. The same molecule shipped in two Phase 3 trials at very different scales and Black-participant proportions: STEP-1 for the obesity indication enrolled 5.66 percent Black at n=1,961, and SELECT for the cardiovascular indication enrolled 3.81 percent Black at n=17,604. The larger trial enrolled the smaller proportional Black cohort, the opposite of what the trial-diversity-gap discourse typically predicts. SELECT's 671-Black-participant subgroup is large in absolute terms (larger than the entire bulevirtide trial cohort by a factor of 4.5) and unaccompanied by a publicly-surfaceable race-stratified hazard ratio for the primary outcome. The ClinicalTrials.gov registry reports only the overall hazard ratio of 0.80 (95% CI 0.72 to 0.90); the NEJM published primary's supplementary appendix is not openly accessible from public sources; and no peer-reviewed secondary analysis in the four SELECT-linked PubMed entries surfaces a race-stratified primary-outcome figure (Lincoff et al, NEJM 2023, PMID 37952131; ClinicalTrials.gov NCT03574597 results module). For this drug and this indication, the enrollment-bar question and the subgroup-analysis-publication question give different answers. The 671 number is high; the published-by-race number is unavailable.

Bulevirtide is the structural-gap exemplar. MYR-301 enrolled 150 patients across Germany, Sweden, Italy, and Russia, of whom 1 was Black. The US chronic-hepatitis-D burden is concentrated in African-immigrant populations whose origin countries are not in the trial geography. When FDA reviewers consider the resubmission, the question of how to evaluate efficacy in a patient population the pivotal trial did not enroll will be load-bearing for the final label.

The trial-diversity question carries a federal statute. Section 3601 of the Food and Drug Omnibus Reform Act of 2022, signed into law December 29, 2022 as part of Public Law 117-328, added subsection (z) to section 505 of the Food, Drug, and Cosmetic Act. The new subsection requires the sponsor of any Phase 3 study, or other pivotal study of a new drug other than bioavailability or bioequivalence studies, to submit a Diversity Action Plan to FDA. The plan must include enrollment goals disaggregated by race, ethnicity, sex, and age group; the rationale for how those goals were determined; and the enrollment and retention strategies the sponsor will use to meet them (21 U.S.C. § 355(z), as added by Pub. L. 117-328 § 3601).

The statute is not yet operative at the prescribing level. Section 3601 ties the effective date to publication of a final FDA guidance document. FDA published a draft of that guidance in June 2024 (Federal Register doc 2024-14284, docket FDA-2021-D-0789), replacing an April 2022 pre-FDORA draft of the same name (Federal Register doc 2022-07978). The draft remains a draft. As of June 2026, FDA has not finalized the guidance, the 180-day clock has not started, and no sponsor is required by the statute to submit a Diversity Action Plan for an active enrollment.

The regulatory pause is part of the story. On January 27, 2025 the Trump administration issued an executive order pausing federal diversity, equity, and inclusion activities, and FDA removed the draft DAP guidance from its website. A court order required HHS to restore the guidance to the FDA website on February 11, 2025. HHS Secretary Robert F. Kennedy Jr. indicated during his January 29, 2025 confirmation hearing that he intends to finalize the DAP guidance. Whether the final language will require the enrollment-goal disaggregation that the draft specifies, or weaken it, is genuinely unsettled.

Two implications for reading a Phase 3 trial today. The trial cohort behind any drug your prescriber is recommending in 2026 was almost certainly enrolled under the voluntary regime that has run since the April 2022 draft, not under a binding DAP requirement; that voluntary regime is what produced the 50-fold range named in the table above. And even when the final guidance arrives, the statute does not direct FDA to reject a submission solely for failure to hit the enrollment goals in the plan. It requires the sponsor to submit a plan. Whether the agency uses other regulatory levers to weight an approval decision against an unrepresentative trial cohort is, at this writing, an open question.

The trial-diversity gap is editorially abstract until it lands in a clinic visit. At that visit, four questions translate the methodology into a decision. A prescriber who can answer them is a prescriber who has read the label. A prescriber who is willing to find the answers when the questions land cold is a prescriber doing the work.

1. Before I start this drug, I would like to know the trial population it was approved on. What percentage of trial participants were Black, and was the trial's race-stratified subgroup analysis pre-specified and powered to detect race-specific effects on efficacy or adverse events?
2. Is there post-marketing real-world-evidence data on outcomes in Black patients taking this drug since approval, and what does it show?
3. Are there alternative drugs in the same class, with the same indication, where the trial population was more race-representative or where a race-stratified analysis is published?
4. If this drug is the right choice for me despite the trial gap, what surveillance or follow-up are you planning so that any race-specific adverse-event signal the trial was not powered to detect gets caught early?

These questions are not adversarial. They are the questions a competent clinician will welcome. A prescriber who cannot answer the trial-population question, and who declines to consult the prescribing information or a clinical pharmacist before continuing the prescription, is signaling that the trial-diversity question is not part of how they read the drug. That is information you can act on. The blackhealth.org provider directory is one place to find prescribers who do read for the trial-population question; a second-opinion visit is another. The decision stays yours.

We name the trial population and the Black-participant share in every drug-coverage piece on blackhealth.org because the methodology this primer describes is load-bearing for whether the published efficacy figure applies to the reader. Three recent pieces work the discipline in different shapes. Our Auvelity Alzheimer's-agitation coverage names the ADVANCE Phase 2/3 11.5 percent Black-participant share and the ADVANCE-2 and ACCORD-2 race-composition gaps that have not yet posted to ClinicalTrials.gov. Our zuranolone postpartum-depression coverage flags that the SKYLARK Phase 3 trial's race composition is not in the published primary abstract. Our Maryland PDAB Ozempic coverage carries the enrollment-floor question into the state-level upper-payment-limit policy debate, where a state board sets a payment ceiling for a drug whose pivotal trial enrolled 5.66 percent Black at obesity scale and 3.81 percent Black at cardiovascular scale.

When a trial cohort's race composition is not in the published primary, our standard is to flag the gap rather than to extrapolate. When the race composition is reported but the subgroup analysis was underpowered, our standard is to say so rather than to imply parity. The trial-population question is not an editorial preference. It is part of how we report on whether a drug works.

Every drug entry in the Section 3 table carries its PMID and ClinicalTrials.gov NCT in the table itself; the block below repeats them in full and adds the FDA guidance documents and statute cited in Section 4 with their docket and public-law numbers.

Primary sources confirmed as of 2026-05-31:

1. Wedemeyer H, Aleman S, Brunetto MR, et al. "A Phase 3, Randomized Trial of Bulevirtide in Chronic Hepatitis D." N Engl J Med. 2023 Jul 6;389(1):22-32. PMID: 37345876. ClinicalTrials.gov NCT03852719.
2. Wilding JPH, Batterham RL, Calanna S, et al. "Once-Weekly Semaglutide in Adults with Overweight or Obesity." N Engl J Med. 2021 Mar 18;384(11):989-1002. PMID: 33567185. ClinicalTrials.gov NCT03548935.
3. Luetkemeyer AF, Donnell D, Dombrowski JC, et al. "Postexposure Doxycycline to Prevent Bacterial Sexually Transmitted Infections." N Engl J Med. 2023 Apr 6;388(14):1296-1306. PMID: 37018493. ClinicalTrials.gov NCT03980223.
4. Iosifescu DV, Jones A, O'Gorman C, Streicher C, et al. "Efficacy and Safety of AXS-05 (Dextromethorphan-Bupropion) in Patients With Major Depressive Disorder: A Phase 3 Randomized Clinical Trial (GEMINI)." J Clin Psychiatry. 2022 May 30;83(4). PMID: 35649167. ClinicalTrials.gov NCT04019704. (Black-participant count of 116 of 327 = 35.5% surfaced from ClinicalTrials.gov NCT04019704 BaselineCharacteristicsModule, Race (NIH/OMB) measure: Black or African American 61 + 55 across arms = 116; White 180; Asian 17; total 327.)
5. Deligiannidis KM, et al. "Zuranolone for the Treatment of Postpartum Depression." Am J Psychiatry. 2023 Sep 1;180(9):668-675 (SKYLARK Phase 3). PMID: 37491938. ClinicalTrials.gov NCT04442503.
6. Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. "Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes." N Engl J Med. 2023 Dec 14;389(24):2221-2232. PMID: 37952131. ClinicalTrials.gov NCT03574597 (SELECT). Baseline race counts surfaced from ClinicalTrials.gov v2 API BaselineCharacteristicsModule (race breakdown not in the published abstract; per discipline #3, the ClinicalTrials.gov SPA URL renders only navigation chrome via WebFetch, but the v2 API path returns full results JSON; codified as a new path-discipline note for SPA-rendered registry surfaces).
7. ClinicalTrials.gov NCT03226522 (ADVANCE Phase 2/3, AXS-05 in Alzheimer's-associated agitation; primary completion 2020). No peer-reviewed primary publication indexed in PubMed as of 2026-05-31; the 11.5 percent Black-participant share (42 of 366) is cited from the ClinicalTrials.gov results record.
8. ClinicalTrials.gov NCT05557409 (ADVANCE-2) and NCT04947553 (ACCORD-2). (Race composition not yet posted as of 2026-05-25; flagged.)
9. FDA. "Diversity Plans to Improve Enrollment of Participants from Underrepresented Racial and Ethnic Populations in Clinical Trials." Draft guidance for industry, published April 14, 2022. Federal Register doc 2022-07978; docket FDA-2021-D-0789. (Superseded by the June 2024 draft of the same name; never finalized.)
10. Food and Drug Omnibus Reform Act of 2022, § 3601, Pub. L. 117-328, 136 Stat. 4459 (December 29, 2022), codified at 21 U.S.C. § 355(z). (Diversity Action Plan submission requirement for any Phase 3 or other pivotal clinical investigation of a new drug other than bioavailability or bioequivalence studies; effective date tied to publication of final FDA guidance.)
11. FDA. "Diversity Action Plans to Improve Enrollment of Participants from Underrepresented Populations in Clinical Studies." Draft guidance for industry, published June 28, 2024. Federal Register doc 2024-14284; docket FDA-2021-D-0789. (As of June 2026 this guidance remains a draft. The draft was briefly removed from the FDA website following the January 27, 2025 Trump administration executive order on diversity, equity, and inclusion, then restored on February 11, 2025 pursuant to a court order. HHS Secretary Robert F. Kennedy Jr. indicated during his January 29, 2025 confirmation hearing that he intends to finalize the guidance; the final has not been issued.)
12. Wright JT Jr, Dunn JK, Cutler JA, et al. "Outcomes in hypertensive black and nonblack patients treated with chlorthalidone, amlodipine, and lisinopril." JAMA. 2005 Apr 6;293(13):1595-1608. PMID: 15811979. (ALLHAT Black-subgroup analysis.)
13. James PA, Oparil S, Carter BL, et al. "2014 evidence-based guideline for the management of high blood pressure in adults: report from the panel members appointed to the Eighth Joint National Committee (JNC 8)." JAMA. 2014 Feb 5;311(5):507-520. PMID: 24352797.
14. Mozaffarian D, Benjamin EJ, Go AS, et al. "Heart Disease and Stroke Statistics-2016 Update: A Report From the American Heart Association." Circulation. 2016 Jan 26;133(4):e38-e360. PMID: 26673558.
15. US Census Bureau. 2020 Decennial Census, Demographic and Housing Characteristics File. (Black or African American alone, percentage of US adult population.)