By Peter A. McCullough, MD, MPH
Since 2020, Silicon Valley technology executive Steve Kirsch has been a tireless advocate for collection of data, analyses, and public health policy based on those findings. He is the founder of the Vaccine Safety Research Foundation.
🧪 Focal Points: McCullough & Kirsch on the Vaccine–Autism Datasets
Steve Kirsch is a data guy who admits his own surveys are inherently biased but allows us to transparently evaluate and draw our own conclusions. That is far more that what our public health apparatus as done on this steadily growing childhood epidemic. To settle issues of personal bias, both McCullough and Kirsch themselves, spouses, and children are all fully immunized against childhood vaccine-preventable illnesses.
🎯 The Core Thesis
Kirsch’s central argument is temporal asymmetry as the fingerprint of causation:
If vaccines were irrelevant, autism onset should be evenly scattered across the calendar relative to shots.
Instead, onset clusters in the days after vaccination and almost never in the days before.
Coupled with a dose-response relationship (more shots → higher odds), he asserts every Bradford Hill criterion is met.
McCullough frames it with the encephalitis model: vaccine → fever → subclinical brain inflammation (encephalitis) → autistic child (post-encephalitic). That’s the mechanistic story the datasets are meant to support. This is consistent with the McCullough Foundation Report: The Determinants of Autism Spectrum Disorder.
🧪 Dataset 1 — The 2023 10,000-Parent-Child Survey
What it is: A parent survey of roughly 10,000 kids, split close to 50-50 vaccinated/unvaccinated, drawn from Kirsch’s own readership.
Dose-response finding: odds of autism rose in proportion to number of vaccines received — a straight line in log space.
Autoimmune disease: odds ratio reported at ~21.7.
Independent check: statistician Matt Briggs reportedly found adverse events dose-dependent across disease categories.
Kirsch’s inference: all Bradford Hill criteria satisfied — “Vaccines cause autism. Period.”
Kirsch concedes his sample is biased — his followers skew toward people with a vaccine-injury story. His defense is that his odds ratio matches the peer-reviewed literature, so the bias is probably small.
⏱️ Dataset 2 — The 2024 Rapid-Onset Autism Survey
What it is: A survey restricted to rapid-onset cases — which Kirsch argues is more than half of all autism.
First 62 responses, the headline numbers:
Group Within 7 days after shot Within 7 days before All vaccines 34 of 56 3 MMR specifically 19 of 33 0
Statistical inference: if onset were random, you’d expect an even split. He computes the probability of the observed clustering by chance at 8×10−228 \times 10^{-22}8×10−22 (MMR) and 1.5×10−381.5 \times 10^{-38}1.5×10−38 (all vaccines).
His key point: the asymmetry — many after, almost none before — is the signature of causation.
The asymmetry is the strongest single idea in the whole interview. A “before” window is a genuine negative control, and it’s the thing mainstream cohort studies never run. But the 10−3810^{-38}10−38 figure is only as good as the sampling frame underneath it, and that frame is parents who already suspect vaccines. Recall bias doesn’t just blur dates — it can manufacture the asymmetry entirely, because a parent primed to suspect the shot will anchor onset to it.
🔬 Dataset 3 — The 2025 Follow-Up Survey
What it is: A confirmation run showing the same high temporal correlation between shot date and onset date.
Notable claim: no autism study has ever examined temporal association — the analysis that would settle it is the one nobody runs.
Self-correction: Kirsch acknowledges survey bias and announces funding a second survey against an unbiased list, with questions pre-reviewed by a neutral epidemiologist.
AI analysis: he cites a Claude Opus 4.7 analysis concluding the signal can’t be explained by recall or selection bias.
Caution flag: AI language models can be led to a conclusion as easily as a person, and the transcript itself later has McCullough noting AI agents refusing to go further on this topic “because it would promote vaccine hesitancy.” If AI outputs are being cited as support here, they’re just as citable as refusal elsewhere.
📋 Dataset 4 — The 2026 Sudden-Onset Autism Survey (New Data)
What it is: An August 2026 recruitment call for parents whose normally-developing child had a sudden, lasting behavioral change later diagnosed as autism. ~3-minute survey, answers promised publicly.
Preliminary results Kirsch calls “stunning”:
30% of onsets on the same day as vaccination — after, never before.
40% within 2 days of vaccination.
Onset age peaks at 18 months, aligning with the CDC schedule — then drops by a factor of 12 at month 19.
Onsets least likely on Sunday — his line: “Genes don’t take weekends off. But pediatricians do.”
Supporting cases cited:
Dr. Doug Hulstedt (California pediatrician): 47 cases of sudden-onset autism within 2 weeks after vaccination, zero in the 2 weeks before; license revoked after writing a single vaccine exemption.
A non-vaccinating pediatric practice: 6,526 unvaccinated kids, 0 autism over 25 years, versus roughly 1 in 15 in adjacent CDC-schedule clinics.
🔗 Dataset 5 — The 2026 Sequence Survey
What it is: Parents of kids who became suddenly autistic after vaccination report a tightly linked chain with no gaps.
The reported chain: normal child → vaccination → fever → Tylenol → autistic child.
The gap question: he asked whether any normalcy gap existed between the vaccine reaction subsiding and onset. Answer: a gap appeared in only 2 of 39 reports, both under 24 hours.
The logic: if onset were random relative to vaccination, there’d be a random number of normal days in between. There isn’t.
Goldberg et al. (2003): neutral recruitment, 67.6% of parents reported immunization concurrent with loss of skills — Kirsch argues coincidence would predict under 20%.
Facebook autism group survey: 75% of regressions in the first 14 days after a shot.
The absence of “normal days” after the shots frame is genuinely sharper than the raw time-to-onset number, because it kills the “they just remembered it as after” objection — a parent who watched a kid check out over 18 lethargic days knows there were no normal days. But it’s still retrospective, self-reported, from motivated samples. The NIH and CDC has never performed a prospective study collecting the same data because of their central concern—promotion of vaccine hesitancy.
⚖️ Overall Assessment
Where the interview is strongest:
The temporal asymmetry argument (after vs. before) is a legitimate causal-signature test that mainstream cohorts genuinely don’t run.
The dose-response relationship, if it holds under a clean design, is one of the more robust Bradford Hill criteria.
The lack of any normal days after the battery of vaccines. This suggests subclinical encephalitis and autism developing as a permanent post-encephalitic complication.
Where it’s weakest:
Every dataset except the promised “unbiased list” run is self-selected from vaccine-skeptical populations. Kirsch says so himself. “Bias is probably small” is a hope, not a measurement.
AI analyses are being used — that could be a category error, as McCullough pointed out that AI, specifically Copilot, refuses to engage on the same topic stating it will not promote “vaccine hesitancy.”
The smoking–lung cancer analogy is emotionally persuasive but cuts both ways: it took decades and large prospective cohorts to settle that one. Kirsch’s surveys are the cross-sectional stage, not the prospective stage.
Attribution of motive: the “autism support societies funded by vaccine interests” explains why these groups would join an Academy of Pediatrics lawsuit against HHS to restore the full vaccine schedule. Why would an autism society take this 180-degree stance unless it was funded to do so.
Bottom line: This is a coherent, internally consistent case built on the right kind of question — the before/after asymmetry — but resting on samples that can’t yet distinguish causation from recall. McCullough has advised Kirsch to publish each dataset as a separate report with a citable date-of-issue (DOI). The study that settles it — a prospective cohort with pre-recorded onset dates and clean vaccination records with expert examination of the child and confirmation of autism — has never been proposed, funded, or performed by any government agency.
📝Please subscribe to FOCAL POINTS as a paying ($5 monthly) or founder member so we can continue to bring you the truth. AlterAI may be used to assist in searches, synthesis, and review.
Peter A. McCullough, MD, MPH
President, McCullough Foundation
FOCAL POINTS has partnered with AlterAI to support independent research and open inquiry. Subscribe to AlterAI today and get a discount on unbiased and accurate AI!
References
