A liver that turns on itself is not a modern mystery. Hepatologists have kept a running list for decades of things that can flip the organ’s immune tolerance and provoke autoimmune hepatitis: minocycline, nitrofurantoin, the interferons, a handful of older vaccines. The clinical picture barely changes from trigger to trigger. A person with a healthy liver develops jaundice and fatigue, their transaminases climb into the thousands, autoantibodies show up in the blood, and a biopsy needle pulls out tissue showing the immune system chewing through the boundary it is supposed to respect, the one between the portal tract and the liver cells around it. Pathologists call that pattern interface hepatitis, and it has meant the same thing for as long as anyone has been looking.

So when case reports of that exact pattern started surfacing after COVID-19 mRNA doses in 2021, the histology was not the surprise. A familiar door had been opened by a new key. The fight that followed, and it has quietly divided hepatology ever since, was never about whether the liver could do this, because we have known for years that it can. It was about whether a surveillance system stood up during the fastest vaccine rollout in history could tell rare apart from unreal. On one side sat the biopsy needle. On the other sat a disproportionality calculation, and the two did not agree.

What the biopsies actually showed

Start with the tissue, because the tissue is the least ambiguous part of this story. A 2022 review in Frontiers in Immunology pulled together 27 published cases of autoimmune hepatitis following COVID-19 vaccination: 22 women and 5 men, ages 27 to 82. The cases split across platforms, with Moderna implicated in 11, Pfizer in 9, AstraZeneca in 6, and a single inactivated CoronaVac dose in the last. Symptoms arrived anywhere from a day to about a month after injection, and nearly all of these patients turned yellow. On biopsy the finding was consistent and specific: interface hepatitis with a lymphoplasmacytic infiltrate, the histological handwriting of autoimmune liver injury. Every patient went on corticosteroids, most recovered, and two died.

Consider one of the index cases, documented in a 2021 report with the pointed title “Autoimmune hepatitis after COVID-19 vaccine, more than a coincidence.” A 65-year-old woman with normal baseline liver enzymes received her first Moderna dose. Two weeks later her ALT had gone from 24 to 1,092. Her antinuclear antibody came back positive in a speckled pattern, her biopsy showed severe interface hepatitis with confluent necrosis, and 60 milligrams of prednisolone a day brought everything back to normal. The dechallenge is what gives that case its weight. When you remove a suspected trigger, treat the immune reaction, and the injury resolves, you have done what clinicians call a dechallenge, and while it is not proof of causation on its own, at the bedside it is close to the strongest circumstantial evidence a single case can offer.

There is even a coherent mechanism on offer, which is more than most drug-induced liver injury can claim. The leading candidate is molecular mimicry. The vaccine instructs cells to manufacture SARS-CoV-2 spike protein, and antibodies raised against that spike appear to cross-react with human tissue proteins, a mechanism that could plausibly reach hepatic antigens the immune system is normally trained to ignore. The Journal of Hepatology weighed the causality question as early as 2021 and landed where careful hepatology tends to: the temporal link and the exclusion of other causes are suggestive, the mechanism is plausible, and the phenomenon warrants serious follow-up rather than a shrug.

What “no safety signal” was actually measuring

Now set that against the sentence that got repeated instead. A 2023 pharmacovigilance analysis in Frontiers in Pharmacology went to VAERS, the U.S. passive reporting system, and counted 53 autoimmune hepatitis reports filed between December 2020 and March 2022, 48 of them after mRNA vaccines. It ran a disproportionality calculation and produced a reporting odds ratio of 1.43, with a confidence interval of 0.52 to 3.96 that comfortably straddles the line of no effect. The authors concluded that the data did “not suggest a safety concern attributable to COVID-19 vaccine at this time.” That sentence traveled widely, while the reasoning behind it stayed home.

Look closely at what the number actually compares. Disproportionality analysis does not ask whether the reaction is happening. It asks whether autoimmune hepatitis is reported more often for COVID-19 vaccines than for the other vaccines already sitting in the same database. A ratio near one means the answer is no, they look about the same, which is a genuinely different claim from “this does not occur,” and the two got quietly collapsed into each other. Worse, the instrument itself is a sieve. VAERS is passive and voluntary, and the agencies that run it openly acknowledge it captures a fraction of real events. The study’s own reporting rate, 0.21 cases per million vaccinees, is exactly what you would expect from a system that counts only the reports that happen to reach it, not the injuries that happen in the world. When your detector undercounts by design, “no signal” is a statement about the detector, not the liver.

The demographics, at least, line up across both bodies of evidence in a way that is hard to wave off as noise. Roughly three-quarters of the VAERS cases were women, matching the 22-of-27 skew in the case literature and the long-standing observation that autoimmune hepatitis prefers female livers. Onset clustered early, a median of 16 days after an mRNA dose, with 45 percent of reports landing inside the first week. That is a biologically sensible latency for an immune reaction, not the flat, random scatter you would expect if these were coincidental diagnoses swept up by chance.

The topic is back in circulation because a new TrialSite News review is again examining autoimmune hepatitis after both vaccination and infection, and its framing, rare but real, is the honest one. The harder half of that story cuts in a direction worth sitting with. SARS-CoV-2 infection itself can trigger the same autoimmunity, including a seronegative case reported in 2025 after COVID-19 with no vaccine involved. Some will read that as exoneration for the shot. If the molecular-mimicry hypothesis is right and it is the spike protein driving this particular liver injury, it should not matter whether the spike arrives by virus or by injection, and the mRNA platform is built to deliver spike to the body’s own cells. The infection data does not clear the vaccine so much as sharpen the same mechanistic question: spike, from either source, is the thing worth watching.

None of this makes autoimmune hepatitis a common outcome of vaccination. It is rare, and the evidence says so plainly. But rare and unreal are not synonyms, and the institutional reflex during the rollout was to let the first quietly become the second. Twenty-seven biopsies and 53 filed reports are not an epidemic. They are a pattern a passive counting system was structurally incapable of confirming, then cited as though its silence were an all-clear.

The tool that could actually settle this already exists. The CDC runs the Vaccine Safety Datalink, an active surveillance network wired into real electronic health records that can test whether liver-autoimmunity diagnoses rose above their expected background rate after vaccination, no voluntary report required. I can find no published VSD analysis of hepatic autoimmunity. That is the query someone should run, and the one to watch for. Until it is run, “no safety signal” stays a description of the instrument, and the biopsy needle, which files no VAERS report and answers to no reporting odds ratio, will keep saying what it has always said.

Sources

  1. Frontiers in Immunology – Autoimmune hepatitis after COVID-19 vaccination (2022 review, 27 cases)
  2. Frontiers in Pharmacology – Real-world evidence of autoimmune hepatitis following COVID-19 vaccination: a VAERS pharmacovigilance analysis (2023)
  3. PMC – Autoimmune hepatitis after COVID-19 vaccine, more than a coincidence (index case, ALT 1,092)
  4. Journal of Hepatology – Autoimmune hepatitis following COVID-19 vaccination: True causality or mere association? (2021)
  5. TrialSite News – Rare but Real: review of autoimmune hepatitis after COVID-19 vaccination and infection (2026)
  6. Cureus – Seronegative autoimmune hepatitis following COVID-19 infection (2025)
  7. PMC – Potential autoimmunity from molecular mimicry between SARS-CoV-2 spike and human proteins