Autism Rates After MMR Withdrawal: Honda et al. (2005) in Context
- Feb 6
- 7 min read
Updated: Feb 26
Honda et al. (2005)
No effect of MMR withdrawal on the incidence of autism: a total population study
Journal of Child Psychology and Psychiatry, 2005
What was this study about?
Honda and colleagues examined whether autism rates changed after the MMR vaccine was withdrawn in Japan.
Japan discontinued use of the combined MMR vaccine in the early 1990s. This created an opportunity to observe what happened to autism incidence when the vaccine was no longer given.
The study focused on children born in Yokohama, a large city with detailed health and developmental records.
What question were the researchers asking?
The researchers asked a straightforward question:
If MMR vaccination were a major cause of autism, would autism incidence decline after the vaccine was no longer used?
To answer this, they compared autism rates in children born:
• before MMR was introduced
• while MMR was in use
• after MMR had been fully withdrawn
How was the study conducted?
The study used a population-based design and included:
• all children born in the study area between 1988 and 1996
• follow-up of autism diagnoses up to age seven
• comparison of cumulative autism incidence across birth cohorts
MMR vaccination was used in Yokohama only for children born between 1989 and 1992. Children born after 1993 did not receive MMR.
What did the study find?
The researchers reported that:
• autism incidence increased steadily across successive birth cohorts
• this increase continued after MMR vaccination had ended
• children born after MMR withdrawal showed higher autism incidence than earlier cohorts
The removal of MMR did not correspond to a decline in autism incidence.
How did the authors interpret these findings?
The authors concluded that:
• withdrawal of the MMR vaccine did not lead to a reduction in autism incidence
• MMR vaccination was therefore unlikely to explain the rising autism rates observed in this population
Their interpretation focused on population-level trends rather than individual risk.
What kind of evidence does this study provide?
This study provides evidence about:
• autism incidence trends over time
• population-level effects following vaccine withdrawal
It does not examine:
• individual susceptibility
• biological mechanisms
• subgroup-specific responses
Those questions fall outside the scope of this study design.
In simple terms
This study looked at what happened to autism rates after Japan stopped using the MMR vaccine.
Even after the vaccine was no longer given, autism diagnoses continued to increase. The authors concluded that stopping MMR did not reduce autism incidence in the population they studied.
Critic 1
Changes in Recognition and Diagnosis
Honda et al. (2005)No effect of MMR withdrawal on the incidence of autism: a total population study
Journal of Child Psychology and Psychiatry, 2005
What is this critique about?
Honda et al. (2005) reported that autism diagnoses increased in children born in later years, including those born after the MMR vaccine was withdrawn in Japan.
A key question raised in the research literature is whether this increase reflects:
• a real rise in autism
• or changes in how autism was recognised and diagnosed during that period
This critique focuses on how diagnostic practices were evolving at the same time as the vaccine policy change.
How were autism definitions changing?
During the 1990s, diagnostic manuals such as DSM-IV and ICD-10 broadened the concept of autism.
These changes included recognising:
• milder forms of autism
• children with average intelligence
• children without intellectual disability
• presentations that may previously have been labelled differently
As definitions expanded, more children became eligible for diagnosis.
Improvements in recognition and services
Autism diagnosis depends not only on definitions, but also on recognition and access.
Rates can be influenced by:
• increased awareness among clinicians and educators
• better referral pathways
• expansion of specialist assessment services
• earlier identification in schools
When systems improve at identifying children, diagnosis rates can rise even if the underlying number of autistic children remains stable.
The IQ pattern in the Honda study
Honda et al. presented autism incidence by IQ level.
The largest increases were observed among:
• children with average IQ
• children with higher IQ
This pattern is consistent with improved recognition of higher-functioning autism over time, rather than a shift limited to children with intellectual disability.
Why this matters for interpreting MMR withdrawal
The withdrawal of MMR is often described as a natural experiment.
However, during the same years:
• autism definitions were evolving
• diagnostic practices were changing
• service access was expanding
Because these changes occurred at the same time as MMR withdrawal, it becomes difficult to attribute rising autism incidence to one single factor.
In simple terms
Honda et al. (2005) shows that autism diagnoses continued to increase after MMR was withdrawn.
This critique highlights that the rise took place during a period when autism was being recognised more widely and defined more broadly.
Understanding these changes helps clarify what the study’s trends can — and cannot — demonstrate about causation.
Sources
• Honda et al. (1996) — Cumulative incidence and improved detection in Yokohamahttps://pubmed.ncbi.nlm.nih.gov/8871801/
• Fombonne (2009) — Epidemiology of pervasive developmental disorders
• Institute of Medicine (2004) — Immunization Safety Reviewhttps://nap.nationalacademies.org/catalog/10997/immunization-safety-review-vaccines-and-autism
Critic 2
Critic 2
Incidence Measurement and Follow-Up Window
What is this critique about?
Honda et al. (2005) calculated cumulative autism incidence up to age seven for children born between 1988 and 1996.
This means the study counted how many children had received an autism diagnosis by their seventh birthday.
This approach is standard in epidemiology, but it introduces specific interpretive considerations.
Why age seven matters
Autism is often diagnosed in early childhood, but not all children are identified by age seven.
Some children:
• are diagnosed later in childhood
• are recognised after changes in services
• are identified only when social demands increase
Stopping follow-up at age seven may not capture later diagnoses.
Diagnosis timing can change over time
If children in later birth years are diagnosed earlier than children in earlier birth years, cumulative incidence up to age seven may increase.
This can occur even if the total number of autistic individuals remains stable.
Shifts in age at diagnosis can influence incidence curves.
Differences in follow-up between birth years
Earlier birth years had more time to reach age seven before the study was conducted.
Later birth years had less long-term follow-up beyond early childhood at the time of analysis.
Differences in observation time can affect comparisons across birth years.
Incidence versus longer-term patterns
Cumulative incidence up to age seven captures early childhood diagnosis.
It does not capture:
• adult diagnosis
• diagnoses made after age seven
• lifetime prevalence
As a result, conclusions must be limited to early childhood incidence within the study window.
In simple terms
Honda et al. (2005) measured how many children had been diagnosed with autism by age seven.
This is a clear and structured approach, but it depends on when diagnoses are made and how long children are followed.
Understanding these timing limits helps clarify what the study’s incidence trends can — and cannot — show.
Sources
• Honda et al. (1996) — Cumulative incidence and improved detection in Yokohama
• Fombonne (2009) — Epidemiology of pervasive developmental disorders
Critic 3
Interpreting Natural Experiments and Population-Level Evidence
What is this critique about?
Honda et al. (2005) is often described as a natural experiment. When the MMR vaccine was withdrawn in Japan, researchers were able to observe autism incidence before and after the policy change.
Natural experiments are influential because they allow real-world comparison without controlled laboratory conditions.
However, they also have specific limits.
Strengths of natural experiments
Natural experiments can provide valuable information because:
• exposure changes clearly at a defined point in time
• large populations are affected
• trends can be tracked across birth years
In this case, MMR withdrawal created a distinct change in vaccine exposure.
Limits of natural experiments
Unlike randomised trials, natural experiments cannot control for other changes happening at the same time.
During the period studied:
• autism definitions were evolving
• diagnostic practices were shifting
• service access was expanding
• awareness was increasing
These concurrent changes complicate interpretation of time trends.
Population-level patterns
Honda et al. examined autism incidence across an entire population.
Population-level studies are well suited to detecting large widespread effects.
They are less suited to identifying:
• rare vulnerabilities
• small subgroup effects
• individual biological mechanisms
This distinction is important when interpreting “no decline” findings.
How this fits into the wider evidence base
Natural experiments are interpreted alongside:
• individual-level cohort studies
• case–control analyses
• systematic reviews
No single study design is definitive on its own.
Honda et al. contributes population-level evidence, but its findings must be understood within the broader research landscape.
In simple terms
Honda et al. (2005) looked at autism rates after MMR was withdrawn.
The study did not observe a decline in autism incidence.
This provides useful population-level evidence, but natural experiments cannot isolate every possible explanation or test every individual-level scenario.
Understanding the strengths and limits of this design helps clarify how the findings are interpreted.
Overall Assessment
Honda et al. (2005) is frequently cited as a natural experiment following the withdrawal of the MMR vaccine in Japan. The study reported that autism incidence continued to rise in children born after MMR was discontinued, suggesting that removal of the vaccine did not correspond to a decline in diagnoses at the population level.
At the same time, the study took place during a period of evolving diagnostic definitions, expanding recognition of higher-functioning autism, and changes in service access. It measured cumulative incidence up to age seven within a specific local system, meaning its findings reflect both biological and diagnostic realities operating at that time.
Taken together, Honda et al. contributes important population-level evidence. However, like all observational studies, its findings must be interpreted within the broader context of changing diagnostic practices, follow-up windows, and the limits of natural experiments. Its role in the literature is significant — but not standalone — forming part of a wider body of evidence examined across multiple study designs and time periods.
