HHS Removed the Recommendation. Pfizer Still Owes the Trial.

The 4,000-person pregnancy study that became 346 women is not a vaccine-injury story. It is an evidence-integrity story.

Sec. Robert F. Kennedy Jr. now controls the federal health department that once told pregnant women to accept COVID-19 vaccination under population-level recommendation. His department has moved COVID-19 vaccination away from a universal posture: CDC’s pregnancy page now frames 2025-2026 COVID-19 vaccination under “individual decision making,” while still telling pregnant women that pregnancy increases the risk of severe COVID-19 and that studies covering more than one million pregnant women have not linked vaccination before or during pregnancy to increased health risks. ^[1,2]

That change matters. It does not settle the question.

Popular Rationalism already named the ethical frame: special protections for pregnant women exist because thalidomide and diethylstilbestrol showed what happens when medicine converts uncertainty into reassurance before evidence can bear the weight. The previous article put the matter plainly: “blanket statements of ‘safety’ without evidence betray that legacy.” We also placed vaccine liability and data access on the table, arguing that a closed compensation-and-surveillance system cannot substitute for civil discovery, independent review, and public access to vaccine-safety data. ^[3,4]

Kennedy’s current policy movement creates a paradox. In 2021, universal recommendation helped destroy the feasibility of Pfizer’s placebo-controlled pregnancy trial. In 2025 and 2026, Kennedy’s department began removing or reclassifying federal vaccine recommendations in the name of informed consent, transparency, and trust. But removing a recommendation does not repair the missing randomized evidence. It changes policy posture. It does not produce the trial the public was promised.

On February 18, 2021, Pfizer and BioNTech announced that the first participants had been dosed in a global Phase 2/3 pregnancy study. The company described a randomized, placebo-controlled, observer-blind trial in approximately 4,000 healthy pregnant women, 18 years or older, vaccinated at 24 to 34 weeks of gestation. The study would evaluate the safety, tolerability, and immunogenicity of two doses of BNT162b2 or placebo, administered 21 days apart. Pfizer also said infants would be monitored through approximately six months of age. ^[5]

That was the public evidence promise.

FDA’s August 2021 clinical review shows why that promise mattered. The adult licensure trial, C4591001, generated pregnancy reports, but it did not generate pregnancy-outcome evidence adequate for pregnancy-specific conclusions. FDA recorded 42 pregnancies among BNT162b2 recipients by March 13, 2021. FDA also recorded that birth outcomes, unknown pregnancy outcomes, and ongoing pregnancies were not included in the study report because the applicant did not collect that information in its standard clinical database. ^[6]

FDA therefore knew the pregnancy evidence remained thin at the point of licensure. The same review described the pregnancy and pregnancy-outcome data from C4591001 as limited and identified additional postmarketing pregnancy follow-up, including a pregnancy registry and the C4591015 randomized pregnancy trial. ^[6]

The randomized pregnancy trial did not mature into the announced 4,000-person study.

Pfizer’s final clinical study report synopsis says C4591015 began February 16, 2021, ended July 15, 2022, and ran at 41 sites in Brazil, South Africa, Spain, the United Kingdom, and the United States. Enrollment stopped on October 25, 2021, with approximately 350 participants enrolled. Pfizer attributed the stop to changes in universal COVID-19 vaccination recommendations for pregnant women and increased global vaccine availability, which made recruitment into a placebo-controlled pregnancy trial difficult. ^[7]

The sponsor’s own explanation deserves a fair reading. Pfizer says universal recommendations and increased vaccine availability made recruitment into a placebo-controlled pregnancy study difficult. The final CSR says enrollment numbers changed to approximately 200 in Phase 2 and approximately 150 in Phase 3. Pfizer’s later public statement says the study no longer had sufficient sample size for the primary immunogenicity objective and that continuation of the placebo-controlled study could no longer be justified under global recommendations; Pfizer also says the proposal was shared with and agreed to by FDA and European marketing authorization. That last sentence remains sponsor-side evidence unless the corresponding FDA/EMA decision memo appears. ^[7,8]

The final randomized denominator tells the story. Pfizer randomized 174 maternal participants to BNT162b2 and 174 to placebo. The maternal safety population contained 173 BNT162b2 recipients and 173 placebo recipients. The infant safety population contained 167 infants born to mothers randomized to BNT162b2 and 168 infants born to mothers randomized to placebo. ^[7]

The placebo comparison also narrowed after delivery. Participants originally randomized to placebo could receive BNT162b2 after unblinding at the one-month postdelivery visit; 152 received a first postdelivery BNT162b2 dose and 148 received a second. That crossover may have followed the ethics of the recommendation environment. It also limited long placebo-controlled maternal and infant follow-up after early postpartum unblinding. ^[7]

The statistical analysis plan records the methodological collapse. SAP version 2 deleted a hypothesis-test section. SAP version 3, tied to protocol amendment 5, clarified that no hypothesis test would occur because of sample-size reduction, removed model-based geometric mean ratio analysis and sensitivity analysis, and changed the nonpregnant comparator selection to 1:1 age matching. ^[9]

This was not just a smaller trial. It became a descriptive trial.

The SAP states that, because of early enrollment termination and reduced sample size, all endpoints would be analyzed descriptively without formal hypothesis tests. Missing e-diary data, antibody results, and laboratory results would not be imputed. ^[9]

FDA’s current COMIRNATY label now gives the regulatory result. Study 9, NCT04754594, enrolled COVID-vaccine-naive pregnant women who received two doses of COMIRNATY or placebo, with the first dose at 24 to 34 weeks’ gestation. The label records 173 COMIRNATY recipients and 173 placebo recipients who received at least one dose. It then states that the trial failed to reach target accrual, had originally been designed for 4,000 participants, terminated prematurely, and created uncertainty regarding differences in outcomes. ^[10]

That sentence should anchor every public discussion of this trial and the use of the vaccine during pregnancy.

The label also states that no COMIRNATY data exist before 24 weeks’ gestation to inform risks for major birth defects and miscarriage. That limitation does not require interpretation. Pfizer designed and conducted the randomized pregnancy trial around late second-trimester and third-trimester dosing. It could not adjudicate first-trimester teratogenic risk because it did not test first-trimester exposure. ^[10]

The preeclampsia imbalance remains unresolved. After excluding participants with HIV infection, FDA’s label records preeclampsia in 6 of 161 COMIRNATY recipients and 2 of 163 placebo recipients. FDA does not call that imbalance causal. FDA says the data cannot establish or exclude a causal relationship. That is the right conclusion. It also means the trial failed to resolve an imbalance it observed. ^[10]

The congenital-anomaly table follows the same pattern. FDA records major or minor congenital anomalies or chromosomal abnormalities in 10 of 156 infants born to women in the COMIRNATY group and 7 of 159 infants born to women in the placebo group. FDA then states that the available trial data cannot establish or exclude vaccine-associated risk because COMIRNATY was not administered in the first trimester and because predisposing factors, including genetic and maternal risk factors, appeared in cases in the COMIRNATY group. ^[10]

Again: not proof of harm. Not proof of absence. Failed adjudication.

The efficacy analysis did not rescue the trial. Pfizer’s CSR says maternal vaccine-efficacy results were “uninterpretable” because of the very small sample size and low number of COVID-19 cases. One analysis estimated 3.8% vaccine efficacy with a 95% confidence interval from -1227.8% to 93.0%, based on 2 cases in the vaccine group and 2 in placebo. Another analysis using the CDC case definition estimated 36.2% vaccine efficacy with a 95% confidence interval from -456.5% to 94.7%, based on 2 cases in the vaccine group and 3 in placebo. ^[7]

Intervals that wide do not estimate clinical protection. They document the cost of under-accrual.

The infant follow-up does not support a mortality alarm. Pfizer recorded two infant deaths, one in each original maternal-randomization group, and investigators assessed both as unrelated to study vaccination. That point belongs in the article because failed evidence cuts both ways. An underpowered randomized trial cannot prove safety with precision. It also cannot justify converting isolated events into a causal injury narrative. ^[7]

The observational record must enter the analysis. CDC currently states that studies including more than one million pregnant women worldwide have not linked COVID-19 vaccination before or during pregnancy to increased health risks for pregnant women or babies. CDC also states that COVID-19 vaccination during pregnancy can reduce severe illness and help protect infants after birth. ^[2]

Large first-trimester observational data also reduce the probability of a large common teratogenic effect. Bernard and colleagues reported a French nationwide cohort study in JAMA Network Open in 2025, DOI 10.1001/jamanetworkopen.2025.38039, including 527,564 live-born infants and 130,338 first-trimester mRNA vaccine exposures. The authors found no increased risk across 75 major congenital malformations overall, by organ system, or individually; the weighted odds ratio for major congenital malformations overall was 0.98, with a 95% confidence interval of 0.93 to 1.04. The authors still noted that extremely rare outcomes cannot be ruled out. ^[11]

That evidence matters. It should block irresponsible claims that Pfizer’s failed pregnancy RCT proves teratogenic harm. It should not permit the opposite abuse: pretending that observational reassurance retroactively completed the randomized pregnancy trial.

This is where the Kennedy context becomes important.

Kennedy’s vaccine changes did not stop with COVID-19 in pregnancy. In January 2026, HHS and CDC announced a revised childhood immunization schedule after a presidential directive. CDC described three categories: vaccines recommended for all children, vaccines recommended for certain high-risk groups or populations, and vaccines administered through shared clinical decision-making. KFF’s analysis reported that the number of diseases targeted by vaccines recommended for all children fell from 17 to 11, and that rotavirus, COVID-19, influenza, hepatitis A, hepatitis B, and meningococcal vaccines no longer sat in the routine-for-all category. Johns Hopkins reported the same shift: those vaccines remained available under shared clinical decision-making, but no longer held universal recommendation status for all children. ^[12,13,14]

A federal judge later temporarily blocked major parts of the administration’s vaccine-schedule changes and ACIP decisions, finding that Kennedy’s process likely violated administrative procedure. CIDRAP reported that the judge said the government disregarded codified scientific methods and undermined the integrity of its actions. My memory tells me that ACIP’s codified scientific methods were not at all scientific. AP likewise reported that the lawsuit originally focused on Kennedy’s decision to stop recommending COVID-19 vaccination for most children and pregnant women, then expanded after HHS reduced the number of vaccines recommended for every child. ^[15,16]

That process fight does not answer the pregnancy-trial question. It reveals the same structural failure from the other side.

The old regime overclaimed certainty. The new regime risks mistaking recommendation rollback for evidence repair.

A clean evidentiary standard rejects both errors. It refuses blanket pregnancy reassurance when the randomized pregnancy trial failed. It also refuses injury claims that outrun the data. It does not treat “recommended” as proof. It does not treat “removed from routine recommendation” as proof. It asks who generated the evidence, who failed to generate it, who controlled the data, who changed the statistical plan, who approved the change, who saw the full adverse-event tables, who saw the missingness, who saw the correspondence, and who had the legal power to compel disclosure.

FDA now says the same general problem exists across drugs and biologics. In May 2026, FDA issued guidance to improve pregnancy safety-data collection, noting that pregnant women often have limited or no human safety data at approval and that health care providers need clear, reliable information. FDA’s Tracy Beth Hoeg stated that products may be recommended to pregnant women despite clinical-trial data insufficient to assess safety during pregnancy. ^[17]

That sentence belongs next to Pfizer’s pregnancy trial.

The central failure did not occur because nobody knew pregnant women deserved evidence. In fact, in matters of human subjects research, HHS regulations contain Subpart B, “Additional Protections for Pregnant Women, Human Fetuses, and Neonates,” and require added conditions for research involving pregnant women, fetuses, and neonates, including scientifically appropriate preclinical or clinical studies, least possible risk, and informed consent. FDA also maintains guidance on when and how to include pregnant women in clinical trials, recognizing that exclusion can leave clinicians and patients without direct evidence. ^[18,19]

The rules existed. The evidence still failed to mature.

Pfizer and BioNTech announced a 4,000-person randomized pregnancy trial. FDA’s label now records 346 maternal participants in the safety population, premature termination, no pre-24-week exposure data for major birth defects and miscarriage, unresolved preeclampsia and congenital-anomaly findings, inconclusive immunogenicity comparisons, and an efficacy analysis Pfizer called uninterpretable. Later observational studies provide meaningful reassurance for evaluated outcomes, including large first-trimester malformation analyses. That observational evidence matters. It does not become the trial by accumulation.

Kennedy’s department can remove recommendations, reclassify schedules, and invoke informed consent. None of that fixes the record.

The repair requires disclosure.

FDA, HHS, Pfizer, BioNTech, and EMA should publish a reconciliation table for C4591015 that includes the original sample-size plan, every protocol amendment, every SAP revision, the rationale for deleting hypothesis testing, planned versus achieved maternal and infant denominators, all endpoint-specific missingness, the full pregnancy and infant safety tables, the complete adverse-events-of-special-interest tables, all withdrawals and loss-to-follow-up records, all placebo-crossover timing data, and every regulatory communication supporting the October 2021 enrollment stop.

They should also separate three sentences in every public communication.

First: Pfizer’s randomized pregnancy trial failed to reach the evidentiary scale announced to the public.

Second: later observational evidence has not detected increased risk for major evaluated pregnancy and infant outcomes and provides real reassurance, especially against large common risks.

Third: observational reassurance does not retroactively complete a failed randomized trial, and recommendation status does not determine truth.

That is the Popular Rationalism position.

RFK removed the recommendation. Pfizer still owes the trial.

References

[1] HHS. Robert F. Kennedy, Jr. | HHS.gov. https://www.hhs.gov/about/leadership/robert-kennedy.html

[2] CDC. COVID-19 Vaccination for Women Who Are Pregnant or Breastfeeding. https://www.cdc.gov/covid/vaccines/pregnant-or-breastfeeding.html

[3] Popular Rationalism. Rebuttal: Pregnancy, COVID-19 Vaccines, and the Erosion of Medical Ethics when Pregnant Women Are Studied. https://popularrationalism.substack.com/p/rebuttal-pregnancy-covid-19-vaccines

[4] Popular Rationalism. Vaccine Liability is On the Table. https://popularrationalism.substack.com/p/vaccine-liability-is-on-the-table

[5] Pfizer. Pfizer and BioNTech Commence Global Clinical Trial to Evaluate COVID-19 Vaccine in Pregnant Women. https://www.pfizer.com/news/press-release/press-release-detail/pfizer-and-biontech-commence-global-clinical-trial-evaluate

[6] FDA. Clinical Review Memo, August 23, 2021 - COMIRNATY. https://www.fda.gov/media/152256/download

[7] Pfizer. C4591015 Final Public Disclosure Synopsis. https://cdn.pfizer.com/pfizercom/clinical%20trials/csr%20synopsis/C4591015%20Final%20Public%20Disclosure%20Synopsis.pdf?VersionId=eOzOBKJIDimi8wXmYFnCkZIw04R3qVVX

[8] Pfizer. Pfizer Upholds Commitment to Transparency and Shares Analysis of COVID-19 Vaccination in Pregnant Women. https://www.pfizer.com/news/announcements/pfizer-upholds-commitment-transparency-and-shares-analysis-covid-19-vaccination

[9] ClinicalTrials.gov / Pfizer. Protocol C4591015 Statistical Analysis Plan, Version 3, March 31, 2022. https://cdn.clinicaltrials.gov/large-docs/94/NCT04754594/SAP_001.pdf

[10] FDA. Package Insert and Patient Package Insert - COMIRNATY. https://www.fda.gov/media/151707/download

[11] JAMA Network Open. Bernard C, et al. First-Trimester mRNA COVID-19 Vaccination and Risk of Major Congenital Malformations. DOI: 10.1001/jamanetworkopen.2025.38039. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2840133

[12] CDC Newsroom. CDC Acts on Presidential Memorandum to Update Childhood Immunization Schedule. https://www.cdc.gov/media/releases/2026/2026-cdc-acts-on-presidential-memorandum-to-update-childhood-immunization-schedule.html

[13] KFF. The New Federal Vaccine Schedule for Children: What Changed and What Are the Implications?. https://www.kff.org/other-health/the-new-federal-vaccine-schedule-what-changed/

[14] Johns Hopkins Bloomberg School of Public Health. HHS Changes Its Pediatric Vaccine Recommendations: What’s Different, What Remains, and What It Means for American Health. https://publichealth.jhu.edu/2026/hhss-abridged-vaccine-recommendations

[15] CIDRAP. Federal judge blocks Kennedy’s changes to childhood vaccine policy. https://www.cidrap.umn.edu/childhood-vaccines/federal-judge-blocks-kennedy-s-changes-childhood-vaccine-policy

[16] AP News. Judge blocks US government from slimming down vaccine recommendations. https://apnews.com/article/kennedy-acip-vaccines-cdc-fc758951019f41d2f5e81e4e2faa22d3

[17] FDA. FDA Issues Guidance to Improve Collection of Pregnancy Safety Data for Drugs and Biologics. https://www.fda.gov/news-events/press-announcements/fda-issues-guidance-improve-collection-pregnancy-safety-data-drugs-and-biologics

[18] HHS Office for Human Research Protections. Subpart B - Additional Protections for Pregnant Women, Human Fetuses and Neonates Involved in Research. https://www.hhs.gov/ohrp/regulations-and-policy/regulations/45-cfr-46/common-rule-subpart-b/index.html

[19] FDA. Pregnant Women: Scientific and Ethical Considerations for Inclusion in Clinical Trials. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/pregnant-women-scientific-and-ethical-considerations-inclusion-clinical-trials