Pfizer's Only Placebo-Controlled Trial of Its COVID Shot in Pregnancy Never Dosed a Woman Before 24 Weeks

From 1940 to 1971, American doctors prescribed a synthetic estrogen called diethylstilbestrol to pregnant women in the belief that it prevented miscarriage and premature labor. According to the National Cancer Institute, an estimated 5 to 10 million Americans were exposed, counting the mothers and the children they carried. The finding that ended it came in 1971, when researchers tied prenatal DES exposure to a rare clear cell cancer of the vagina and cervix in a small group of daughters who were still very young women. The children had been born and had grown up, and the harm surfaced only because someone went on following them long after the pregnancy was over.

Hold the COVID vaccine to that standard and you get to the trial Pfizer and BioNTech ran in pregnant women, the only placebo-controlled test of their shot in pregnancy, and it fails almost every part of it.

Four thousand, then 348

On February 18, 2021, the companies announced that their pregnancy study would enroll approximately 4,000 healthy pregnant women across nine countries (Pfizer and BioNTech launch announcement). The ClinicalTrials.gov record NCT04754594 shows that 348 women were randomized, 174 to the vaccine and 174 to saline placebo, and the record’s summary now describes a planned study of about 350, as if the larger target had never existed. It gives no reason for the cut. Epidemiologist Tracy Beth Høeg was asking in November 2023 whether Pfizer had stopped recruiting, and the registry still has no answer. The trial finished in July 2022. Its results went up in December 2024, nearly two and a half years later.

The timing of the doses matters even more than the headcount. Every woman in the trial was vaccinated between 24 and 34 weeks of gestation, according to the ClinicalTrials.gov record NCT04754594. Nobody was dosed in the first trimester, when the heart and other organs form, or in the first half of the second. The professional advice that followed covers all of it: as Immunize.org summarizes, the American College of Obstetricians and Gynecologists recommends COVID vaccination during pregnancy and allows it in any trimester. Whatever observational data the college relies on, the randomized, placebo-controlled evidence the manufacturer produced for this product starts at 24 weeks, and the recommendation reaches back to conception.

Then there’s how long anyone watched. Babies were followed to 6 months of age and no further. Of the 167 infants born to vaccinated mothers, 15 did not complete follow-up, while 29 of the 168 born to placebo mothers dropped out, mostly by withdrawal, and the record offers no explanation for why one arm lost nearly twice as many.

In the infant safety analysis, serious adverse events struck 21 of 156 babies in the vaccine arm and 24 of 159 in the placebo arm, which points in neither direction. Adverse events of special interest, counted from birth to six months, ran 5.1 percent in the vaccine arm against 1.3 percent on placebo. The confidence intervals show how little a trial this size could settle: 2.2 to 9.9 percent for the vaccine arm and 0.2 to 4.5 for placebo. They overlap, so the trial recorded close to a fourfold gap and still cannot say whether it was chance. That cuts both ways, and neither way flatters the sponsor. A study too small to confirm a signal is also too small to clear one.

Birth defects follow the same pattern. The registry’s birth-outcome table records congenital malformations in 6.0 percent of vaccine-arm infants and 3.6 percent on placebo. Several of the vaccine-arm diagnoses, among them trisomy 21, DiGeorge syndrome and mucopolysaccharidosis, are genetic and fixed at conception, and a dose given at 24 weeks arrives after the heart and limbs are built. So the lopsided count cannot be pinned on the shot, and that is the trial’s deeper failure, because a study that begins dosing at 24 weeks has no way to detect harm from exposure during organ formation, which is exactly the exposure the recommendation now permits.

Where the instructions went

The case for vaccinating in pregnancy borrows its logic from older shots. A mother makes antibodies, the placenta actively pumps one class of them, IgG, into the fetal bloodstream, and the newborn arrives with protection on loan for the months before its own immune system can respond. That is how the whooping-cough shot in pregnancy works: the baby gets the antibody, never the vaccine itself. An mRNA vaccine is different in kind, because what goes into the arm is a set of genetic instructions wrapped in lipid nanoparticles, telling cells to build the coronavirus spike protein, and the obvious question is whether those instructions stay with the mother.

A team at NYU Long Island School of Medicine went looking. In a 2024 letter in the American Journal of Obstetrics and Gynecology, Xinhua Lin, Nazeeh Hanna and colleagues examined placentas from two women vaccinated during pregnancy and found vaccine mRNA in both, in the maternal-facing decidua and in the fetal villi, along with spike protein in placental tissue from both. In one woman they also detected vaccine mRNA in umbilical cord blood, at about 13 percent integrity compared with about 85 percent in her own blood. A heavily fragmented message is less likely to be read by a cell, though it is still there, and two patients cannot say how often this happens. The work was funded by the NIH’s child-health institute, and the authors reported no conflicts.

The same group had already looked at breast milk. Their JAMA Pediatrics research letter, 2022 tested 131 milk samples from 11 vaccinated lactating women and found vaccine mRNA in 7 samples from 5 of them, at times up to 45 hours after the shot, with a top whole-milk level of 11.7 picograms per milliliter and higher concentrations in the milk’s extracellular vesicles. The authors judged breastfeeding safe, but they also wrote that “caution is warranted about breastfeeding children younger than 6 months in the first 48 hours after maternal vaccination until more safety studies are conducted.” That is the authors’ recommendation, not an infant outcome they measured, and the safety studies it calls for are still owed.

The strongest evidence on the benefit side comes from Norway, and it is large. A study led by Helena Niemi Eide of the University of Oslo, published in Pediatrics in March and reported by NPR, followed 146,031 children born between March 2020 and December 2023 through national medical records for up to two years, with funding from the university and a Scandinavian government agency rather than any drugmaker. Babies of mothers vaccinated in pregnancy were about half as likely to have a hospital visit for COVID in their first two months. At 3 to 5 months that reduction had shrunk to 24 percent, and it was gone after six months of age. Exposed babies were no more likely to visit the hospital for infections of any kind, though they were about 5 percent more likely to see a doctor for an infection, which the researchers attribute to the doctor-going habits of mothers who choose vaccination.

This is an association in an observational cohort, careful and independently funded but not randomized. On its own terms it supports one thing: lower rates of infant COVID hospital visits, concentrated in the first two months and fading by mid-infancy. It says nothing about the developmental outcomes that Pfizer’s trial was too late, too small and too short to see. A review published October 28 in Vaccines carries a title that concedes as much: Maternal mRNA Vaccination: Immunological Principles and Evidence Gaps in Infant Protection.

Who holds the next card

Washington has already stepped back. Immunize.org reports that the CDC’s September 2026 clinical considerations send pregnancy questions to the product package inserts, and that the pregnancy safety content the agency once published has been removed from its website since 2025. Immunize.org’s answer is that “no evidence supports a change in previous conclusions on the safety and effectiveness of the vaccine during pregnancy.” That holds only in the narrow sense that the evidence capable of changing those conclusions was never collected. ACOG has kept its every-trimester recommendation regardless.

The children exposed in the womb during the 2021 rollout are now turning four and five, and two institutions can find out how they are doing. Pfizer holds the records on the 335 infants born to its 348 trial mothers. It owes the public an on-the-record account of why a 4,000-woman study became 348, and independent researchers deserve access to those infants’ data. Eide’s group in Oslo already links 146,031 births to their medical records, and extending that analysis to developmental, autoimmune and neurological diagnoses at school age would give the first large-scale look at what the trial never measured. ACOG has the standing to demand both, since it is the body asking women to take the shot in the weeks Pfizer never tested. Whether either dataset gets opened before these children start kindergarten will tell us more about the vaccine’s safety in pregnancy than another review of the gaps.

Sources

1. ClinicalTrials.gov record NCT04754594
2. Pfizer and BioNTech launch announcement
3. American Journal of Obstetrics and Gynecology: Lin, Botros, Hanna et al., vaccine mRNA in placenta, maternal and cord blood (2024)
4. JAMA Pediatrics research letter, 2022
5. NPR via GBH: Norwegian Pediatrics study of 146,031 children and maternal COVID vaccination (March 2026)
6. Immunize.org: How ACOG’s 2026-27 COVID-19 recommendations differ from CDC’s clinical considerations
7. Vaccines (MDPI): Maternal mRNA Vaccination: Immunological Principles and Evidence Gaps in Infant Protection (October 2026)
8. Tracy Beth Høeg: Did Pfizer stop recruiting for their pregnancy trial? (November 2023)
9. National Cancer Institute: Diethylstilbestrol (DES) exposure and cancer fact sheet