Age Windows in 2.56 Million Records Show No MMR Autism Link

A U.S. study of 2,560,035 children found no higher autism risk after the first MMR dose given before age 2. Todd L. Burstain of UW Medicine and colleagues reported an adjusted hazard ratio of 0.97 in The Pediatric Infectious Disease Journal. The 99 percent confidence interval ran from 0.91 to 1.03, a range that includes no difference at all.

The paper’s work sits in the timing of the shot, not in another pile of unadjusted head counts. Age-anchored landmark analysis was built to stop clinic habits and a child’s age from pretending to be a vaccine effect.

The Method Built to Cancel Timing Bias

Observational vaccine studies have a built-in tangle. Families who keep well-child visits get shots on time and get developmental screens. Parents already worried about speech or eye contact may stall the measles, mumps, and rubella vaccine. If the math ignores those patterns, vaccinated and unvaccinated kids are not the same kids.

Burstain, Dale Jacques, and Jennifer M. Burstain pulled records from Cosmos, Epic’s multi-hospital network. Epic now lists 310 million patients and 2,358 hospitals, plus 52.5K clinics. The analytic file kept children with at least one routine primary care visit between 12 and 24 months, linked birth-parent data, and follow-up through age 8 or December 2024. Autism diagnosed at or before 11 months was dropped so the shot would come first.

By 24 months, 91.4% of the cohort had received a first MMR dose. Childhood autism, a narrower code than the full autism spectrum, sat at 2.55% in the sample.

HOW THE AGE WINDOWS WORKED

• The cut: Children were grouped in short age bands such as 11.5 to 12.5 months, then 12.5 to 13.5 months, out to 24 months.
• The comparison: In each band, kids who got their first dose in that window were set against kids still unvaccinated at the start of the window who stayed unvaccinated through follow-up.
• The covariates: Models adjusted for the mother’s age, race and ethnicity, gestational age, small-for-gestational-age birth, multiple birth, diabetes or genital or urinary tract infection in pregnancy, neighborhood disadvantage, and U.S. region.
• The stop rule: Factors that happen after the shot were left out, because those can distort a hazard ratio instead of cleaning it.

Directed acyclic graphs guided which pregnancy and birth variables belonged in the model. That is a causal-diagram habit, not a kitchen-sink dump of every column in the chart.

2.56 Million Children and a Hazard Ratio of 0.97

Across every window, first-dose MMR was not tied to a higher autism diagnosis rate. In the primary analysis, covering vaccination between 11.5 and 24 months, the adjusted hazard ratio was 0.97 (99 percent CI 0.91 to 1.03). A ratio of 0.97 is a 3% lower diagnosis rate among vaccinated children, a gap the interval cannot separate from zero.

Two later windows, 14.5 to 17.5 months and 17.5 to 24 months, came in slightly lower still, at 0.90 and 0.92. The authors treat those dips as leftover bias, not as a protective drug effect. Extra birth-related factors left the primary ratio at 0.97. Dropping the neighborhood score, which was missing for 53% of children, moved it to 0.96. Excluding kids with only one recorded autism diagnosis did not change the story.

THE ADJUSTED RATIOS SIDE BY SIDE

The pneumococcal conjugate vaccine booster, given around 12 to 15 months, has no theoretical path to autism. It is the checkup companion to first MMR. Its near-null ratio is the authors’ check that the method is not minting fake links out of well-child visits. “These results reinforce existing evidence supporting the safety of MMR vaccination,” the University of Washington team wrote.

CDC’s Autism Page Now Draws a Line Around MMR

On July 22, 2026, the Centers for Disease Control and Prevention updated its autism and vaccines page. The rewrite, posted under the Data Quality Act, walks back a blanket sentence that vaccines do not cause autism. HHS says it has opened a wider review of autism’s causes.

The claim “vaccines do not cause autism” is not an evidence-based claim because studies have not ruled out the possibility that infant vaccines cause autism.

Centers for Disease Control and Prevention, Autism and Vaccines page, July 22, 2026

The same page still treats MMR as the one childhood shot with a large observational file. Reviews from the Institute of Medicine and the Agency for Healthcare Research and Quality, it says, keep a high strength of evidence for no association with autism spectrum disorder, based on observational studies only. It then repeats older complaints: most older papers had serious method limits, foreign schedules differ from the U.S. calendar, and a retrospective cohort cannot prove causation or settle every vulnerable subgroup.

Burstain’s file is a U.S. schedule in U.S. clinics, with children compared at the same ages, pregnancy covariates chosen on a diagram, and a same-visit control shot. Jake Scott, an infectious diseases physician at Stanford, put the design in those terms: the work answered the standard criticisms of past research, on American kids, compared at identical ages, with a built-in test for hidden bias.

Why Vaccinated Kids Look Slightly Safer on Paper

Vaccinated children had much longer contact with the medical system. Median follow-up was 853 days for vaccinated children and 366 days for unvaccinated children, more than twice as long. More time in clinic usually means more chances to pick up a formal diagnosis. The team still did not find a higher autism rate after MMR.

That imbalance runs the other way from the fear that the shot causes autism. If extra visits inflate diagnoses among vaccinated kids, a true harm signal would be easier to see, not harder. The slight dip below 1.0 in later windows fits a different leak: families already watching a toddler’s development may delay the dose, so the still-unvaccinated group in those later bands is enriched for kids heading toward a diagnosis.

Cosmos also lacked family history of autism. If parents who already have an older autistic child skip or stall MMR for a younger sibling, that missing field would tend to make the vaccine look protective. The authors say so themselves. A records study also cannot test what a parent means by a regression after a 16-month visit. Codes capture what a clinician entered, not the scene in the waiting room.

Denmark and the Sibling Studies Already Looked Like This

The new U.S. file is large, not first. Anders Hviid and colleagues at Statens Serum Institut followed 657,461 children born in Denmark from 1999 through 2010. During 5,025,754 person-years, 6,517 children were diagnosed with autism, an incidence of 129.7 per 100,000 person-years. The fully adjusted hazard ratio for MMR was 0.93 (95 percent CI 0.85 to 1.02). Subgroups with a sibling history of autism, other childhood vaccines, or a high autism risk score did not show a rise after MMR.

In 2015, Anjali Jain and colleagues studied 95,727 privately insured U.S. children who had older siblings. ASD was diagnosed in 994 children (1.04%). An older sibling had ASD in 1,929 cases (2.01%). Among those with an affected older sibling, 6.9% were diagnosed, against 0.9% with unaffected siblings. First-dose MMR coverage at age 2 was 73% when an older sibling had ASD and 84% when the older sibling did not, an 11-point gap. At age 2, the adjusted relative risk after one MMR dose in the affected-sibling group was 0.76 (95 percent CI 0.49 to 1.18). A 2002 Danish cohort in the New England Journal of Medicine had already reported an adjusted relative risk of 0.92 (95 percent CI 0.68 to 1.24) for autistic disorder.

THREE COHORTS, THE SAME DIRECTION

The fear itself traces to a 1998 case series on 12 children that was later retracted for scientific misconduct. Large cohorts have not reproduced that claim. The Burstain paper’s addition is less the null than the age lock: U.S. kids, U.S. timing, a control vaccine that should have rung if clinic traffic were the whole machine.

What the Authors Say the Records Still Miss

Linked birth-parent records cut a large share of otherwise eligible children, so the cohort is not every child in Cosmos. Neighborhood disadvantage was missing for more than half. Autism was the childhood-autism code, not the broader spectrum used in education-linked surveillance. No chart review confirmed those codes. Vaccine status came from the electronic record, not a second audit of paper cards.

WHERE EXPERTS DISAGREE

• The comparison group: Peter A. McCullough, a cardiologist, argued the paper had no truly unvaccinated control, no clinical autism exam, no parent interview, and a narrow ICD definition. The authors’ unvaccinated arm is children still unshot at the start of each age window who stayed that way, not a lifetime zero-dose census.
• The leftover dip: Burstain’s team reads hazard ratios of 0.90 and 0.92 in later windows as bias. A critic who wants a never-vaccinated, exam-confirmed cohort will not take that reading as settled.
• The federal split: CDC’s July 22, 2026 page still cites high-strength observational evidence against an MMR-autism link, then says those studies cannot close the file on infant shots as a group or on every subgroup.

Future work, the authors say, could fold in outside data on family history and test the broader autism spectrum diagnosis. Until those fields sit in the chart, the small tilt toward fewer diagnoses in vaccinated children remains a bias they refuse to sell as protection. The ratio that the design was built to test, first MMR before age 2, stayed at 0.97.

Frequently Asked Questions

Does the First MMR Dose Raise Autism Risk?

In this cohort the adjusted hazard ratio for first MMR between 11.5 and 24 months was 0.97, and the 99 percent confidence interval included 1.0, so the result is compatible with no change in risk. U.S. routine timing for that first dose is 12 to 15 months of age, which sits inside the study’s earliest windows; the paper did not test later catch-up doses given after age 2 as a separate causal question.

Why Is the Study’s Autism Rate Lower Than the CDC Figure?

CDC’s Autism and Developmental Disabilities Monitoring Network identified ASD in about 1 in 31 eight-year-olds (3.2%) across 16 communities in 2022, using health and education records, with site rates from 1 in 103 in Texas (Laredo) to 1 in 19 in California, and boys more than 3 times as likely as girls to be identified. Burstain’s 2.55% used a narrower childhood-autism diagnostic code, dropped diagnoses at or before 11 months, and did not add school records, so the two percents are not the same case definition.

What Is a Negative Control in the Burstain Analysis?

A negative control is an exposure that should not cause the outcome if the method is clean. The team used the pneumococcal conjugate vaccine booster given at about 12 to 15 months, a shot with no proposed autism mechanism that often lands at the same well-child visit as MMR; its adjusted hazard ratio of 1.01 (99 percent CI 0.99 to 1.04) is the check that routine checkups were not manufacturing a false MMR signal.

Did the Study Include Children With Autistic Siblings?

Family history of autism was not in the Cosmos file, which the authors flag as a limit that would tend to make MMR look protective if hesitant families already living with autism skip the dose. Jain’s 2015 claims study of 95,727 privately insured children with older siblings did carry that field and still found no increased ASD risk after MMR, including among younger siblings whose older sibling had ASD, where coverage at age 2 was 73% against 84% in families without an older diagnosis.

Who Funded the 2.56 Million-Child MMR Study?

On the preprint record the authors stated that no external funding was used and that none had a conflict of interest to disclose. The published journal report is an unfunded University of Washington analysis of existing Cosmos records, not an industry trial, which still leaves the usual limits of any observational EHR study in place.

Disclaimer: This article is news reporting on a published observational study and on public-health documents. It is informational only and is not medical advice, a vaccination order, or a diagnosis of any child. Parents and caregivers should talk through immunization timing and developmental concerns with a qualified pediatrician or other licensed clinician who knows the child before making a care decision. Figures, diagnostic codes, and federal webpage language reflect the sources as dated in 2025 and 2026 and may be revised as records and agency pages change.