Is it safer to delay a child’s first MMR shot?
No. In 2.5 million US children, getting the first MMR vaccine between 11.5 and 24 months was not linked to a higher chance of an autism diagnosis. The risk came out about 3% lower than in children who had not yet been vaccinated at the same age, which in statistical terms means no real difference (adjusted hazard ratio 0.97).
The MMR vaccine protects against measles, mumps, and rubella. Worry about a link to autism has stuck around for decades, even though study after study has failed to find one. This new paper went after a narrower version of the question that many parents actually ask: not “is the shot safe at all,” but “is it riskier to give it earlier rather than later?”
That question is harder to answer than it sounds. Children who get vaccinated on time also tend to see doctors more often, and children whose parents already have concerns about development sometimes get vaccinated later. Both patterns can make a vaccine look falsely risky. The researchers built their study to strip those patterns out.
What the data show
The main analysis looked at children whose first MMR dose fell between 11.5 and 24 months. Their autism risk was about 3% lower than the comparison group, and the true figure is very likely somewhere between 9% lower and 3% higher (adjusted hazard ratio 0.97; 99% confidence interval 0.91 to 1.03). Because that range comfortably includes “no change at all,” the honest reading is that early MMR neither raised nor lowered autism risk.
The researchers repeated the analysis at every age band they had data for, and the answer kept coming back the same. As the authors put it, the hazard ratios “remained near the null” across all of the age-specific groups.
They also ran a clever check. They tested a completely different shot, the pneumococcal booster given at 12 to 15 months, which nobody claims causes autism. If their method were picking up some hidden bias rather than a real effect, that shot should have looked suspicious too. It did not. The pneumococcal booster came out about 1% higher, very likely between 1% lower and 4% higher (adjusted hazard ratio 1.01; 99% confidence interval 0.99 to 1.04). Two different vaccines, two flat results.
Dr. Kumar’s Take
I like this study for a reason that has nothing to do with the headline. The headline was never in doubt for me. What I find useful is that the authors took the specific fear seriously and designed around it instead of dismissing it.
The worry most parents carry is rarely “vaccines cause autism.” It is usually softer than that. It sounds like “could we just wait a few months, to be safe?” That version of the question deserves a real answer, and until now the answer has often been a general reassurance rather than a direct test. This paper tests it directly, at a scale of 2.5 million children, and finds nothing.
The negative control is the part I would point to if I had one thing to show a parent. Any method can produce a clean-looking result by accident. Running the same method on a vaccine that no one suspects, and getting the same flat answer, tells you the method is not manufacturing reassurance. It is measuring something real.
How the study was done
The team used Cosmos, a large electronic health record database, and pulled 2,560,035 US children whose records were linked to their birth parents’ records. They followed those children through age 8, which is old enough that most autism diagnoses have already been made.
The core method was a “landmark analysis,” which is simpler than it sounds. Instead of comparing vaccinated children to unvaccinated children over a whole childhood, you pick a specific age, look at who has been vaccinated by that exact point, and count what happens afterward. This stops a child from being counted as unvaccinated during months when they were simply too young for the shot, an error that has distorted earlier work on vaccine timing.
The analyses were prespecified, meaning the plan was locked in before the results were seen. The adjustments they made were chosen using directed acyclic graphs, a formal way of mapping out which other factors could muddy the comparison before deciding which ones to account for.
How strong is the evidence?
This is observational research, not a randomized trial, and it never could be one. You cannot ethically assign children at random to delay a measles vaccine. That means some unmeasured factor could in theory still be lurking behind the result, and the authors are careful to say the negative control “supports” rather than proves that residual bias is not driving things.
The size helps a lot, though. At 2.5 million children, a real effect of any meaningful size would have been hard to miss. The narrow certainty range around the main result is the tell: the data are precise enough to rule out anything but a very small effect in either direction, and even that would be an effect toward lower risk, not higher.
The findings also apply to a specific window. This study asks about MMR given before 24 months, plus the single pneumococcal booster it used as a negative control, and it does not speak to any other schedule or any other outcome.
Practical Takeaways
- If you have been considering pushing your child’s first MMR shot back a few months out of caution, this study looked at exactly that question in 2.5 million children and found no benefit to waiting.
- Delaying the shot is not a neutral choice, because it leaves a young child unprotected against measles during months when they could otherwise be covered.
- Bring the timing question to your pediatrician directly rather than working it out alone.
- Treat the negative control result as part of the evidence, not a technical footnote, because it is what separates a well-built study from one that merely reaches a comforting conclusion.
Related Studies and Research
- Less sugar before age 2 was linked to less dementia later
- Bigger baby food portions linked to later childhood weight gain
- Status of oral rehydration therapy in Bangladesh: how widely is it used?
- Small changes in sleep, exercise, and diet linked to 9 extra years of life
FAQs
Why do researchers keep studying the MMR vaccine and autism?
The question keeps getting studied because the worry keeps being raised, and because each new study can test a slightly different version of it. Older work mostly asked whether vaccinated children had more autism than unvaccinated children. This one asks something narrower, which is whether the age at which the first dose is given matters. Answering the narrow version requires methods that were not widely used before, and it also requires enormous datasets. Researchers are not repeating themselves so much as closing progressively smaller gaps.
What is a negative control, and why does it matter here?
A negative control is a test you run where you already know the answer should be nothing. Here the researchers used the pneumococcal booster given at 12 to 15 months, a shot nobody has linked to autism. If their method had a hidden flaw that made vaccines look risky or protective for the wrong reasons, that flaw should have shown up on the pneumococcal shot too. It came back flat, at about 1% higher with a range from 1% lower to 4% higher. That flat result is evidence about the method itself, not about the vaccine.
Does this study say anything about vaccines given at other ages?
No, and that limit is worth being clear about. The analysis was anchored to first MMR doses given between 11.5 and 24 months, and every conclusion is tied to that window. It says nothing about the second MMR dose, about vaccines given in the first year of life, or about the childhood schedule as a whole. Those questions have their own body of research behind them, but this particular paper was not designed to answer them.
Bottom Line
Across 2,560,035 US children followed to age 8, receiving a first MMR vaccine between 11.5 and 24 months was not associated with any increase in autism risk, with a result that sat essentially at zero difference (adjusted hazard ratio 0.97; 99% confidence interval 0.91 to 1.03). A negative control vaccine tested with the same method produced the same flat answer, which argues that hidden bias is not what is producing the reassurance. For a parent weighing whether to delay the shot, the data point in one direction: waiting buys nothing, and it costs months of protection against measles.
