Background. We examined whether, in the UK, there is an increased risk of autism (AD) following exposures, in early life, to: (1) wild measles; (2) live attenuated measles, alone or in combination as MMR; and (3) the alteration of the mumps strain within MMR.

Method. We conducted time trend analyses of 2407 AD subjects born between 1959–93; and for comparison, 4640 Down's syndrome (DS) subjects born between 1966–93. Between 1968–86, we correlated variations in AD and DS births with wild measles incidence. Between 1959–93, we tested for abrupt changes in the long-term AD birth trend for the effects of introducing: (1) monovalent measles vaccines in 1968; (2) MMR immunization in 1988; and (3) the ‘overnight switch’ from mixed use of Urabe MMR to exclusive use of Jeryl–Lynn MMR in 1992. Incidence rate ratios (IRRs) were used as measures of association.

Results. We found no significant association between AD births and exposure (prenatal and post-natal up to 18 months age) to population rates of measles infections, and no ‘step-up’ increase in AD births associated with the introduction of monovalent measles and MMR vaccines, and changing mumps strain. An unexpected reduction in AD births of 21% (95% CI 6·9–33·3%; P=0·005) among the post-1987 birth cohorts was detected.

Conclusion. No increased risk of AD following exposures to wild measles and vaccinations with monovalent measles, and Urabe or Jeryl–Lynn variants of MMR was detected. The precise meaning of the detected AD births reduction is unclear. Our study cannot exclude rare complications of MMR, given its correlational design.