By Justin Yamashita, MSc. Benchtop, site, CRO: three levels of basic and clinical research, explained without spin.

The receipts for everything Monday and Wednesday claimed.

The deeper dive: trial methodology, mechanisms, and the science behind the story.

Of 86 approved drugs with sex specific pharmacokinetic data, 76 produced higher drug exposure in women at the same dose. That's the finding this issue is built on, and the asymmetry underneath it is stronger than the headline number.

Monday's Root Room site selection meeting and Wednesday's Plain Talk zolpidem case both sit on top of it. Here's the evidence, the method, and the regulatory timeline that moved twice in 2025 while almost nobody covered it.

The Prediction Runs in One Direction

A 2020 analysis in Biology of Sex Differences assembled the pharmacokinetic data on 86 approved drugs where sex specific numbers were available. For 76 of them, women showed higher drug exposure than men on the same dose, measured mainly as area under the curve, which is the total amount of drug the body is exposed to over time, along with peak concentration and how fast the body clears it. Elimination times ran longer in women too.

Then the authors asked the harder question. Of those drugs, 59 had clinically identifiable sex biased adverse reactions. Did the pharmacokinetic difference predict which sex got hurt? In 52 of the 59, yes. 88%.

Two numbers, two denominators. The 76 of 86 is about exposure. The 88% is about harm, and it's calculated on the 59, not the 86. They're close to each other arithmetically by coincidence, and they get fused constantly. Keep them apart.

The asymmetry inside the 88% is the part worth carrying. Where the pharmacokinetics ran higher in women, higher adverse reaction rates in women followed in 96% of cases. Where they ran higher in men, male predominant reactions followed in only 29%.

Two limits on that asymmetry, stated up front. The 59 drugs entered the analysis because they already had an identifiable sex bias in adverse reactions, which is selection on the outcome being tested and makes the relationship look tighter than a random sample would. And detection isn't symmetric: women report adverse drug reactions more often, so a female biased reaction is more likely to be recorded as one. The male biased arm is also the smaller of the two, so the 29% rests on few drugs and thin numbers move. The direction of the finding is well supported. The precise spread should be held as an estimate.

The authors' own conclusion is that sex differentiated pharmacokinetics predict adverse reactions in women but not in men, and they recommend evidence based dose reductions for women. The paper opens by noting women experience adverse drug reactions at nearly twice the rate of men. That figure is in Zucker and Prendergast's own text, which is why it's used here and not attributed elsewhere.

If sex differences in drug response is a thread you want to pull, the books that cover it properly are on the Root to Rx reading list at RootToRx.com/the-roots. Those are affiliate links, so a purchase supports the newsletter at no extra cost to you.

The Enrollment Gap, With the Method Attached

An analysis published in Communications Medicine in December 2025 examined 341 Phase III trials supporting FDA drug approvals from 2017 to 2023, covering 348 approved drugs, using the FDA's own Drug Trial Snapshots data.

The test was specific, and the specificity matters. For each trial, enrollment was compared by chi squared test, which is just a test of whether an observed count differs from an expected count by more than chance, against US Census representation for the four largest racial and ethnic groups: Asian, Black, Hispanic and White. A trial counted only if it cleared the threshold for all four at once. Twenty trials did. 6%.

Two findings underneath the headline. Black and Hispanic enrollment declined across the study period, while Asian enrollment rose and White enrollment held roughly flat. And trials carrying Breakthrough Therapy designation, an FDA status that speeds review for drugs showing early promise against serious conditions, were more balanced than the rest, which is a lead worth pulling on rather than a conclusion.

One note on citation hygiene: the paper's own title says "341 US FDA-approved drugs." The abstract says 341 Phase III clinical trials. The trials figure is the accurate one, and if someone corrects you to drugs, they're reading the title, not the study.

Three limits on the 6%, stated up front so nobody has to catch me on them.

The test is conjunctive. A trial passes only if it clears the threshold for all four groups at once. Stack four conditions and the joint pass rate falls sharply on arithmetic alone, so a trial well matched on three groups and marginally short on the fourth scores identically to a trial that enrolled almost no one. 6% measures how many trials cleared a demanding compound test. It doesn't cleanly measure how far off any individual trial was.

The comparison population is the US census. Many pivotal trials enroll across multiple countries, and for a drug whose disease isn't evenly distributed, disease prevalence would be the more defensible denominator than national population.

Chi squared is sensitive to sample size. In a very large trial it flags small deviations. In a small trial it misses large ones. Pass rate is therefore partly a function of how big the trials were.

What none of that touches is the trend. Black and Hispanic enrollment declined across the study period, measured the same way every year in the same dataset. A changing yardstick could manufacture a level. It can't manufacture a slope.

The Legal Timeline, 1977 to Right Now

1977. FDA recommended excluding women of childbearing potential from Phase I and early Phase II trials, a direct response to the thalidomide tragedy of the 1950s and 1960s. The intent was protective. The effect was decades of dosing data drawn mostly from men.

June 10, 1993. Seven years of NIH policy memos hadn't been law. This was. The NIH Revitalization Act, Public Law 103-43, sections 131 through 133, added section 492B to the Public Health Service Act, now codified at 42 U.S.C. 289a-2. Three requirements, and the third is the one nobody quotes. Women and members of minority groups must be included in NIH supported clinical research, with outreach for recruitment. Phase III trials must be designed to allow a valid analysis of whether the variables studied affect those groups differently. And the cost of inclusion "is not a permissible consideration" in deciding whether inclusion is inappropriate. That sentence is in the statute because cost was the standing excuse.

December 29, 2022. FDORA, part of the Consolidated Appropriations Act 2023, Public Law 117-328, the same bill and the same day as the FDA Modernization Act 2.0 covered in Issue 021. It required sponsors of certain late stage trials to submit a Diversity Action Plan before enrollment: goals by age, sex, race and ethnicity, the rationale, and how the sponsor intends to meet them.

June 2024. FDA published draft guidance implementing that requirement, docket FDA-2021-D-0789, with the Federal Register notice at 89 FR 54010. The comment period closed September 26, 2024. FDORA sets a clock for final guidance after comments close, reported as nine months, which would put the deadline around June 26, 2025. I haven't read that clock in the public law text myself, so treat the exact date as approximate and the direction as certain. The deadline is behind us and no final guidance exists.

January 2025. The draft guidance page was removed from fda.gov following executive orders on federal diversity programming. FDA gave no formal announcement of the removal.

February to July 2025. The removal was challenged and reversed in court. In Doctors for America v. Office of Personnel Management, et al., No. 1:25-cv-00322 (D.D.C.), Judge John D. Bates granted a temporary restraining order on February 11, 2025, an emergency order that freezes things in place before a case is decided, requiring defendants to restore the removed medical webpages and datasets by end of day. Agencies covered included OPM, HHS, CDC and FDA, and the DAP guidance was among the restored pages. The case didn't stop there. On July 3, 2025 Bates ruled for the plaintiffs without a trial, which is what summary judgment means, finding the agency memoranda behind the removals broke the Administrative Procedure Act, the law governing how federal agencies make decisions. His words were "arbitrary and capricious," which is that law's own standard rather than a characterization of anyone's motives. He cancelled the memoranda and ordered the websites restored to their January 29, 2025 state. Restoration was confirmed complete in December 2025.

July 17, 2025. NIH issued NOT-OD-25-131, revising its own inclusion policy. It renamed the policy, replaced the phrase sex and gender with sex throughout, and updated its racial and ethnic categories to OMB Statistical Policy Directive No. 15, the government wide standard for how federal agencies collect race and ethnicity data, which adds Middle Eastern or North African as a category.

December 2025. FDA finalized a different guidance in adjacent territory: "Enhancing Participation in Clinical Trials: Eligibility Criteria, Enrollment Practices, and Trial Designs," docket FDA-2019-D-1264. It revises the November 2020 final guidance previously titled "Enhancing the Diversity of Clinical Trial Populations." The word diversity isn't in the new title. Retitling a guidance on revision is routine and by itself means nothing. What is not routine is what legal analysts who've reviewed the text report: that the diversity framing is absent and that it doesn't mandate Diversity Action Plans. I've confirmed the guidance, its date, docket and retitling directly on fda.gov. I haven't read the full PDF, so the characterization of its contents is theirs, not mine, and it's worth reading yourself before repeating it.

What that adds up to: the statute didn't change. The 1993 inclusion mandate stands. The 2022 Diversity Action Plan requirement stands in law. What moved is the implementing layer, twice, in the same year. Quote any single layer and you can prove almost anything.

Wednesday called this the group photo nobody checked the edges of. This is the paperwork version of the same thing. The rule saying check the edges is still on the books. The instructions for how to check them were never finished.

Who Else Requires This, and What They Actually Require

If you've only read US coverage, you'd think representative enrollment is an American policy argument. It isn't. It sits in the guideline that regulators across the US, EU, Japan, Canada, the UK, China, Switzerland and Brazil inspect against.

ICH E6(R3), the current Good Clinical Practice guideline, adopted January 2025, says it directly:

"The participant selection process should be representative of the population groups that the investigational product is intended to benefit, once authorized, to allow for generalizing the results across the broader population."

That's new in R3. The previous version didn't say it. The stated reason is generalizability, not fairness.

Now the part almost nobody reports correctly. The European Union has a binding requirement and the United States does not.

Under Regulation (EU) 536/2014, which has applied since January 2022, a member state assessing a trial application must evaluate "whether the groups of subjects participating in the clinical trial represent the population to be treated, or if not, the explanation and justification provided." Annex I requires the application to contain a justification for the gender and age allocation of subjects, and an explanation if a group is excluded or underrepresented.

That is a regulation, not a guidance. It’s directly enforceable in every member state. But read its scope carefully: it covers sex and age only. It doesn't name race, ethnicity or ancestry.

So the honest global picture is this:

The EU has an enforceable justify-or-explain duty on sex and age. The US has a statutory Diversity Action Plan duty covering age, sex, race and ethnicity, but Congress tied its start to publication of final FDA guidance, and that guidance hasn't published, so the duty hasn't been triggered. Health Canada published a draft demographics action plan guidance in December 2025 that states outright that submission is optional and creates no new legal requirements. The UK's new clinical trials legislation, in force April 2026, is silent on inclusion and diversity, and the MHRA and HRA Inclusion and Diversity Plan is explicitly voluntary, with the HRA stating that failure to submit one will not cause an application to be rejected. Japan's framework addresses ethnic factors through global-trial notifications rather than diversity requirements.

No major regulator anywhere imposes an enrollment quota. What exists is a spectrum from one binding justify-or-explain duty covering two characteristics, through a statute waiting on a trigger, down to guidance and pilots. Anyone telling you this is a mandate is wrong. Anyone telling you it is purely an American political program is also wrong, because it is written into international Good Clinical Practice.

The Skeptic’s Toolkit, Turned Around

Wednesday ran Toolkit Q5, same standards for all claims, on the industry rather than on a wellness ad, using the COAG warfarin trial. Here is that case with the numbers and the sourcing attached, because it is the most important evidence in this issue.

COAG, published in the New England Journal of Medicine in 2013, randomised patients starting warfarin to genotype-guided dosing or a clinical dosing algorithm. The primary outcome was percentage of time in the therapeutic range at four weeks.

Overall result: 45.2% genotype-guided versus 45.4% clinically guided. Adjusted mean difference of minus 0.2 percentage points, p equals 0.91. A wash.

African American subgroup: 35.2% genotype-guided versus 43.5% clinically guided. Adjusted mean difference of minus 8.3 percentage points, p equals 0.01. The genotype arm also took longer to reach a therapeutic INR and had more elevated-INR events.

The mechanism, stated precisely. The algorithm used three variants: CYP2C9*2, CYP2C9*3, and VKORC1 minus 1639G to A. Those are the variants that carry the most predictive weight in European-ancestry populations, which is where the algorithms were derived. The variants that carry more weight in West African ancestry include CYP2C9*5, 8 and 11, and a variant in the CYP2C cluster region designated rs12777823. They were not in the algorithm.

Say this the accurate way and it stays accurate: warfarin does not work differently in Black patients. The drug is identical. The predictor set was wrong, because it was fitted on cohorts in which the relevant variants were rare or absent. The failure is in the model, not in the patient. Race is a social category; what varies here is allele frequency, which tracks genetic ancestry imperfectly and is not the same thing.

That distinction isn't pedantry. It is the difference between a claim that survives scrutiny and one that collapses the first time somebody points out that ancestry is continuous, that most people are admixed, and that a checkbox on an intake form is not a genome.

Two Different Histories, and Why Merging Them Breaks the Argument

There is a version of this topic that opens with Tuskegee. It's worth explaining why this issue doesn't, and the reason is analytical rather than squeamish.

The United States Public Health Service untreated syphilis study at Tuskegee ran from 1932 to 1972. Six hundred Black men were enrolled, 399 with latent syphilis and 201 without, to observe the natural history of untreated disease. There was no informed consent. Penicillin became the treatment of choice for syphilis by 1943 and the participants were not offered it. An Associated Press story in July 1972, based on documents from a Public Health Service investigator, ended it. The National Research Act followed in 1974, and the Belmont Report in 1979.

That is a history of harm from inclusion. People were used as a means without consent, and treatment they were entitled to was withheld.

Under-representation is a history of harm from exclusion. Nobody is deceived. Nobody is enrolled against their interest. The injury is the absence of evidence about people who take the drug anyway, and it lands on future patients and the clinicians who have to guess.

Here is why merging them breaks the argument: the remedies point in opposite directions. The remedy for Tuskegee is stronger protection against being enrolled, consent, oversight, the right to refuse and withdraw. The remedy for under-representation is more enrollment, broader eligibility, deliberate recruitment. Present both as one story and you get an argument that eats itself, because the protective machinery built after 1974, including the reflexive exclusion of pregnant people as a risk-management default, is itself a live cause of the under-representation problem.

The Belmont Report anticipated this. Its three principles are respect for persons, beneficence, and justice. Justice is in there precisely because the commission understood that protection can shade into exclusion, and that both over-burdening and under-including a group are failures of the same principle.

The two histories do connect, through exactly one channel, and it runs one way. The abuse history is a well-founded cause of present-day distrust, and distrust depresses enrollment. That link is real and it shouldn't be waved off.

But be careful with it, because it has its own failure mode. If low enrollment is explained primarily by community reluctance, the problem has been relocated into the patient and out of the trial design. FDA's own draft guidance names sponsor operational assumptions as a cause, not participant reluctance. The barriers regulators actually list are the operational ones: where sites are located, how restrictive the eligibility criteria are, insurance and referral pathways, visit burden, transport, childcare, language, and whether a clinician offers a trial at all. Monday's scene was about the first item on that list.

"Tuskegee explains it" is true as a contributing cause and, used as the whole explanation, is a way of not fixing the site network.

Open Right Now, and You Can Read It Yourself

The comment window on a rule that would delete the word gender from FDA's regulations closes October 2, 2026, and it's open while you're reading this. The rule is "Modification of Certain Terminology in Title 21," published May 6, 2026 at 91 FR 24380, docket FDA-2026-N-2886. It would replace the word gender with sex, or delete it, throughout 21 CFR. The comment period was reopened on August 3, 2026.

I'm not going to tell you what to think about it. I'm telling you it exists, it's open, the docket number is above, and the primary documents are free on federalregister.gov. A docket is just the public file for a rulemaking: the proposal, the comments, and eventually the agency's response to them. Anyone can read it and anyone can file. Read it and decide for yourself.

For the Record

Of 86 drugs with sex specific pharmacokinetic data, 76 showed higher exposure or slower elimination in women at the same dose: Tier 1, Zucker and Prendergast 2020.

Of the 59 of those drugs with clinically identifiable sex biased adverse reactions, the pharmacokinetic difference predicted the direction in 52 cases, 88%, breaking down as 96% of female biased cases and 29 percent of male biased cases: Tier 1, Zucker and Prendergast 2020. Limitation: the 59 were selected on the presence of a sex bias, female biased reactions are more likely to be detected, and the male biased arm is small.

Only 6% of 341 Phase III pivotal trials from 2017 to 2023 cleared a chi squared test against US Census representation for all four of the largest racial and ethnic groups simultaneously: Tier 1, Zaaijer and Groen 2025, using FDA Drug Trial Snapshots data. Limitation: the four way conjunctive test drives the pass rate down independent of any single trial's shortfall, and the census is not the same denominator as disease prevalence.

Black and Hispanic enrollment declined across the 2017 to 2023 period while Asian enrollment rose and White enrollment held roughly flat: Tier 1, Zaaijer and Groen 2025. This trend is measured consistently within the dataset and is not affected by the limitations on the 6% figure.

The NIH Revitalization Act of 1993 made inclusion of women and minority groups in NIH funded research a legal requirement and stated that cost is not a permissible consideration for exclusion: Tier 2, Pub. L. 103-43.

FDA required the recommended zolpidem starting dose for women be halved, immediate release 10 mg to 5 mg and extended release 12.5 mg to 6.25 mg: Tier 2, FDA.

FDORA's Diversity Action Plan requirement is in force by statute; FDA's implementing guidance remains in draft, past its approximate June 2025 deadline, and a differently titled final guidance issued in December 2025: Tier 2, Pub. L. 117-328 and FDA dockets FDA-2021-D-0789 and FDA-2019-D-1264.

The December 2025 guidance omits the diversity framing and the Diversity Action Plan mandate: Tier 3, legal analysis, attributed and not independently confirmed against the guidance text.

Try the Move Yourself

Every claim in the section above carries its tier. That's the whole habit. Does It Have a Receipt is our two minute game on doing that to claims in the wild, before you repeat one. Free, no signup, at openlabelmedia.com/games.

Refer a friend

If you read this far, you're the reader this thing is for. The best thanks is sending it to one person who'd read it too. Series 1 cards are the reward.

The Skeptic's Toolkit, Applied to This Issue

Q6, am I seeing the full picture, or only part of the evidence? The full picture here has three layers and most coverage shows one. There's a statute, which is stable. There's an implementing guidance layer, which moved twice in 2025. And there's an outcome, measured at 6%, that hasn't tracked either one closely.

Q5, am I applying the same standards to all claims? Test it on yourself. If "there's a federal law requiring inclusion" counts as evidence when it supports the optimistic reading, then "the implementing guidance was never finalized" has to count when it doesn't. And if the unfinished guidance counts as evidence that nothing is required, then the statute has to count as evidence that something is. Same category, same weight, both directions.

Next Issue

Next issue: the room you're not allowed to sit in, and what an independent board can do to a trial while it's running, without ever telling you.

NEW TO ROOT TO RX? START HERE

Issue 003b: How to evaluate any health claim using the Skeptic's Toolkit. Start here.

Issue 001: Why medical distrust is rational, and what to do with it anyway.

Issue 004: How a drug gets from a lab to your pharmacy. The pipeline in plain language.

All back issues are at RootToRx.com.

References

1. Zucker I, Prendergast BJ. "Sex differences in pharmacokinetics predict adverse drug reactions in women." Biology of Sex Differences. 2020;11:32. doi:10.1186/s13293-020-00308-5

2. Zaaijer S, Groen SC. "Longitudinal clinical trial enrollment trends across 341 US FDA-approved drugs and their guiding role in precision medicine strategies." Communications Medicine. 2025;5:514. doi:10.1038/s43856-025-01270-2

3. NIH Revitalization Act of 1993, Public Law 103-43, signed June 10, 1993. Inclusion provisions sections 131 to 133, adding section 492B to the Public Health Service Act, 42 U.S.C. 289a-2. congress.gov

4. Food and Drug Omnibus Reform Act of 2022 (FDORA), part of the Consolidated Appropriations Act, 2023, Public Law 117-328. congress.gov

5. U.S. Food and Drug Administration. "Diversity Action Plans to Improve Enrollment of Participants from Underrepresented Populations in Clinical Studies." Draft guidance for industry, June 2024, docket FDA-2021-D-0789. Federal Register notice 89 FR 54010, June 28, 2024. fda.gov

6. U.S. Food and Drug Administration. "Enhancing Participation in Clinical Trials: Eligibility Criteria, Enrollment Practices, and Trial Designs." Final guidance, December 2025, docket FDA-2019-D-1264. fda.gov

7. U.S. Food and Drug Administration. "Questions and Answers: Risk of next-morning impairment after use of insomnia drugs; FDA requires lower recommended doses for certain drugs containing zolpidem." fda.gov

8. National Institutes of Health. NOT-OD-25-131, "Revised Policy on the Inclusion of Women and Members of Racial and/or Ethnic Minority Groups in Clinical Research," July 17, 2025. grants.nih.gov

9. U.S. Food and Drug Administration. "Modification of Certain Terminology in Title 21." Proposed rule, 91 FR 24380, May 6, 2026, docket FDA-2026-N-2886. Comment period reopened August 3, 2026, closing October 2, 2026. federalregister.gov

10. Doctors for America v. Office of Personnel Management, et al., No. 1:25-cv-00322 (D.D.C.). Temporary restraining order February 11, 2025; summary judgment July 3, 2025.

11. NIH Office of Research on Women's Health. NIH policies on inclusion, and history of women's participation in clinical research. orwh.od.nih.gov

12. Kimmel SE, French B, Kasner SE, et al. "A pharmacogenetic versus a clinical algorithm for warfarin dosing." New England Journal of Medicine. 2013;369(24):2283-2293. doi:10.1056/NEJMoa1310669

13. International Council for Harmonisation. ICH E6(R3) Good Clinical Practice, Step 4, January 6, 2025. database.ich.org

14. Regulation (EU) No 536/2014 of the European Parliament and of the Council on clinical trials on medicinal products for human use, Article 6 and Annex I paragraph 17(y). eur-lex.europa.eu

15. Health Canada. "Draft guidance document for clinical trial sponsors: SGBA Plus Demographics Action Plan in clinical trial applications." December 19, 2025. canada.ca

16. Health Research Authority and MHRA. Guidance on developing and submitting an Inclusion and Diversity Plan. hra.nhs.uk

17. U.S. Food and Drug Administration. "Enhancing the Diversity of Clinical Trial Populations: Eligibility Criteria, Enrollment Practices, and Trial Designs." Final guidance, November 2020. fda.gov

18. Centers for Disease Control and Prevention. The U.S. Public Health Service Untreated Syphilis Study at Tuskegee. cdc.gov

19. The Belmont Report. 44 Federal Register 23192, April 18, 1979. hhs.gov

New here? Root to Rx runs three pieces a week. Monday you watch a real health claim try to work on someone. Wednesday you get the tools to run it yourself. Thursday you get the full evidence so you can check my work. Written by a clinical researcher, free, at RootToRx.com.

For educational purposes only. Nothing in this newsletter is medical advice. Talk to your doctor before making any health decision.

Instagram: instagram.com/roottorx | LinkedIn: linkedin.com/company/roottorx | Facebook Lab: facebook.com/groups/theroottorxlab | Subscribe: RootToRx.com

You are receiving this because you subscribed at roottorx.com.

Disclosure: Root to Rx is published independently by Open Label Media LLC. Views expressed are personal views of Justin Yamashita and do not represent his employer or any affiliated organization. No employer resources or proprietary information are used. Every claim is sourced from publicly available materials.

Root to Rx | 228 Park Ave S, #29976, New York, New York 10003, United States