The 82% miscarriage claim: a denominator error, and why Canadians are now seeing it everywhere

One number, produced by dividing a study's miscarriages by the wrong group of women, has travelled from a debunked social media post in 2021 to sworn testimony before a US Senate committee. This week it acquired a fresh set of official-looking documents. Canadian parents are the audience it is now reaching, and nothing in Canada is set up to answer it.

A QuackeryWatch consumer warning · Posted 11 August 2026

Where the number came from

In June 2021 the New England Journal of Medicine published preliminary safety findings from the American v-safe pregnancy registry, led by Tom Shimabukuro of the US Centers for Disease Control and Prevention. The registry had enrolled just under 4,000 women who received a COVID-19 vaccine while pregnant. Because the paper appeared while most of those pregnancies were still ongoing, the authors could only analyse the 827 pregnancies that had actually finished. Of those, 104 ended in miscarriage — a rate of 12.6%, which sits inside the ordinarily expected range.

The 82% figure is what you get if you keep the same 104 miscarriages but shrink the denominator. Roughly 700 of the completed pregnancies belonged to women who were not vaccinated until the third trimester, long past the window in which first-trimester miscarriage occurs. Remove them and 127 women remain. Divide 104 by 127 and you get 82%.

The error is not subtle. The 700 women removed from the denominator were women whose pregnancies had already passed safely through the first trimester. Taking the survivors out of a group and then reporting the loss rate of what is left will produce an alarming number in any dataset, on any subject. It is the same manoeuvre that would let you prove that most hospital patients die by counting only the ones admitted to palliative care.

British fact-checking organisation Full Fact and the science-review site Health Feedback both published corrections in 2021, when the claim was first circulating on Instagram and Twitter. Health Feedback traced its early spread that July to an American gym owner with no relevant training.

What the follow-up data actually showed

The published paper was preliminary by design, and the obvious question — what happens to women vaccinated early in pregnancy — was answered directly the following month. In August 2021 the CDC published an analysis restricted to 2,456 people who received an mRNA vaccine either just before conception or before 20 weeks of pregnancy. The cumulative miscarriage risk in that group was 13%, against a previously published background range of 11% to 16%.

That is the study the 82% claim requires you not to know about. It uses precisely the population the claim says was hidden, and it finds nothing.

Nor has the broader literature moved since. In June 2025, when the US Department of Health and Human Services published a public FAQ that leaned on the same framing, FactCheck.org went through it with researchers who pointed out that calculating miscarriage rates without accounting for risk factors is inappropriate, since the vaccinated groups in these datasets carried more of the comorbidities that raise miscarriage risk in the first place. A second study cited in that same HHS document, examining outcomes in patients undergoing IVF, had likewise found no association between vaccination and pregnancy loss.

The step that changed the audience: RU-486

A miscalculated statistic is ordinary. What made this one durable was a rhetorical addition.

The 82% figure was carried into testimony before a US Senate committee by James Thorp, an American obstetrician and maternal-fetal medicine specialist, who told senators the raw data showed an 82% first-trimester miscarriage rate and that this figure mirrors the effect of chemical abortion drugs such as RU-486.

That comparison is the load-bearing element, and it is worth being clear about what it accomplishes. It converts a vaccine-safety claim into an abortion claim. In doing so it hands the material to an entirely separate constituency — one that was never part of the vaccine debate, that is highly organised, and that has every reason to find the framing compelling. This is why the posts you are seeing move so much faster than the ordinary run of anti-vaccine content, and why they reach people who would scroll past a claim about autism.

August 2026: the document release

On 10 August 2026, Senators Ron Johnson and Rand Paul released the first batch of text messages recovered from Anthony Fauci's former government telephone, which HHS had turned over to a Senate subcommittee. The senators said the device held more than 34,000 messages and 522 voicemails; they published screenshots of eleven of them. This followed Fauci's appearance before the Senate Homeland Security and Governmental Affairs Committee on 29 July 2026, at which he invoked his Fifth Amendment right more than a hundred times, and a subsequent committee vote to hold him in contempt of Congress.

The texts include a January 2021 exchange between Fauci, then-CDC director Rochelle Walensky and Vivek Murthy about vaccination at different stages of pregnancy. In it, Fauci initially said there was no data or theoretical reason to prefer early or late vaccination, then returned two hours later, after consulting colleagues, to raise the possibility that fever and cytokine reactions following a second dose could bear on first-trimester miscarriage. Days later he publicly said monitoring had turned up no red flags in pregnant women.

QuackeryWatch takes no position on the contempt proceedings or on the broader congressional investigation. But the content of the texts should be described accurately, because the posts circulating this week are not describing it accurately. What the messages show is officials privately discussing a theoretical mechanism during a period when, as Walensky said in the same exchange, there was essentially no data. They do not show anyone identifying an increased miscarriage risk, because at that point nobody had measured one. When it was measured — six months later, in the analysis described above — it was not there.

A private January 2021 conversation about a hypothesis and a public August 2021 dataset that tested the hypothesis are not in conflict. That is the ordinary shape of the thing working properly. Presented as a screenshot with no dates, it looks like a confession.

Why this matters in Canada specifically

None of the above is a Canadian story. That is exactly the problem.

The material arrives with no domestic counterweight. There is no Canadian committee generating rival documents, no Canadian hearing, no Canadian transcript. A Canadian reading about this sees only the American frame, carried on American accounts, complete with genuine government URLs. The Canadian position exists — it is on the Public Health Agency's own pages — but a guidance webpage is not written to rebut a congressional exhibit and will not be read by anyone the exhibit reaches.

It targets the earliest decision, not the last one. The claim that vaccines cause autism asks a parent to regret something already done. A miscarriage and fertility claim reaches people before they have children, and it lands on the prenatal and newborn decisions — pertussis, RSV, influenza, the newborn schedule — that come a year or more before the measles question is ever asked. A woman deterred at 28 weeks is a child under-immunized at two years old.

Canada has no margin left. Canada lost its measles elimination status on 10 November 2025 after more than twelve months of sustained transmission, with over 5,200 cases in 2025 and two infant deaths, and roughly a thousand further cases in 2026. MMR coverage among two-year-olds fell from 89.5% in 2019 to 80.4% in 2024. In British Columbia, a CBC analysis published in April 2026 found the share of seven-year-olds fully vaccinated against measles fell from 90% in 2015 to 70% in 2024. Roughly 87% of Canadian measles cases in 2026 have occurred in unvaccinated people. There is no slack in these numbers for a further shift in prenatal confidence.

It will not travel through the channels people are watching. This claim moves better through pregnancy, birth and natural-wellness communities than through openly political ones. The loud end is the organised advocacy groups. The quiet end is a prenatal wellness account posting a screenshot of a Senate exhibit between two posts about sleep training, and that is the version that changes decisions.

And no Canadian body can reach it. Our health regulators license practitioners, not platforms. A Canadian college can act against a registrant who repeats the 82% figure to a patient — and QuackeryWatch has documented at length how slowly that tends to happen. Nobody in Canada has any lever on the distribution itself.

What to do with it

• Check the denominator. Ask who was removed from the bottom of the fraction. In this case, roughly 700 women whose pregnancies had already passed the point of risk.
• Check the date range against the claim. A January 2021 message about what was unknown in January 2021 tells you nothing about what was found in August 2021.
• Ask what the person repeating it does when corrected. The 82% figure was publicly corrected in 2021. Continuing to present it in 2026 without addressing the correction is the relevant fact about the source.
• Take the actual decision to a clinician. Pregnancy immunization decisions have real trade-offs worth discussing with someone who knows your history. Those are conversations for a physician or a midwife.