No Effect Of MMR Withdrawal On The Incidence Of Autism
A Total Population Study Journal of Child Psychology and Psychiatry. February 2005
By Honda H, Shimizu Y and Rutter M
Japanese study is the strongest evidence yet for a link between MMR and autism. Andrew J Wakefield FRCS FRCPath and Carol M Stott PhD Honda and colleagues present a fascinating report on the cumulative incidence (numbers of new cases with time) of autistic spectrum disorders (ASDs) in the Kohoku Ward, Yokohama, Japan, for children born 1988 to 1996. The study seeks to examine the relationship between ASD and MMR vaccination. Japan is unique since MMR was introduced in 1989 and discontinued in April 1993. Honda et. al. see this as providing an ideal opportunity to test whether there is a causal association between MMR exposure and incidence of ASDs. They predict that, if MMR causes autism, stopping MMR should result in a subsequent decline in incidence.
This was not seen. In fact, there was a striking rise in the incidence
of ASDs in this population over time, with a marked rise postdating the
removal of MMR. The authors state that their finding ‘implies that MMR
could not cause a substantial proportion of cases of autism’.
In conducting a study of this kind it is important to consider the
background against which earlier hypotheses relating to the possible
association between measles containing vaccines such as MMR, bowel
disease and childhood developmental disorders were formulated, and
according to which any relevant data should be interpreted.
The above notwithstanding, the authors of the Japanese study are
confident in the completeness of ascertainment of ASD cases, the
accuracy and precision of their screening, and the quality of diagnostic
services for developmental disorders. Given this level of confidence in
the incidence figures, the data merit further scrutiny in light of
Japan’s unique experience with the vaccines of interest.
In 1998 one of us (AJW) made a recommendation in relation to how
parents might wish to protect their child from the relevant infections –
measles, mumps and rubella – by vaccination. This recommendation was
based upon published scientific studies from his own laboratory together
with an extensive examination of safety studies conducted in relation to
measles vaccine either given alone or in combination with the other
viral vaccines. The recommendations were that consideration should be
given to (i) having M, M and R separately as the individual component
vaccines and (ii) allowing an interval of one year between the vaccines.
The basis for these recommendations came from the following
First, that the safety studies of MMR vaccine were inadequate, a
conclusion subsequently endorsed by independent scientific review .
Second, that there was clear evidence from the early clinical
trials of MMR, of ‘interference’ between the component viruses in the
combined vaccine, an influence apparently mediated through an altered
immune response to the vaccines when given together . The safety
consequences of this ‘interference’ are completely unknown since they
have not been investigated as they should have been.
Third, that children that had experienced concurrent natural
measles (or single measles vaccine) and natural mumps infections within
the same year were at significantly greater risk of later inflammatory
bowel disease . The latter finding is consistent with a natural
‘interference’ phenomenon that potentially increases the risk of
long-term measles virus infection and delayed disease. It is quite
possible that this effect could operate for an interval of one-year or
more between exposure to two different viruses. Measles virus and
measles vaccines can suppress the immune system for a prolonged period
after exposure . This effect is exemplified by the excess mortality and
immunosuppression associated with potent measles vaccines, observed in
developing countries, which led to these vaccines being abandoned
(reviewed in 3).
Having established this background, one can examine the relevant
events in Japan.
Vaccination policy and policy change in Japan Monovalent measles
vaccine was introduced in Japan in 1978 and was recommended to be given
at 12 – 72 months of age. Rubella vaccine was introduced in 1977 and
was recommended for junior high school female students. An MMR
vaccination program was launched in April 1989 for children aged between
12 and 72 months with the majority receiving the vaccine by 18 months of
age. There was no mumps vaccine used in Japan before the introduction of
It is notable that various brands of MMR vaccine were licensed in
Japan, some of them containing the mumps Urabe AM9 strain. Due to
increasing public and professional concern about reported incidences of
meningitis following MMR, public confidence declined over the years
following its introduction and MMR vaccine uptake fell. Subsequent
studies confirmed that the Urabe AM9 mumps vaccine was causally
associated with meningitis. This resulted in the termination of the MMR
program in April 1993, and no child in the current study received MMR
from 1992 onwards. The Urabe AM9 mumps vaccine was discontinued and
replaced with a strain of mumps vaccine which did not cause meningitis.
Single measles, mumps, and rubella vaccines replaced the combined
vaccine in 1993 in a new immunization schedule, which was formalised the
following year. The recommendation was for Japanese children to receive
monovalent measles, mumps and rubella vaccines to be given to infants
spaced by a period of not less than four weeks.
Against the background of this changing vaccination policy the
cumulative incidence curve of ASD in this population is very interesting
(see Figure One).
The Japanese study does not tell us anything about the incidence
of ASD prior to 1988; prevalence data are used as an estimate of the
upper limit (Figure 1). Following the introduction of MMR there was a
rise in annual incidence of ASDs to 85.9 for children born in 1990. The
incidence subsequently declined to 55.8 for children born in 1991.
The incidence then rose again sharply, to a level of 161
(121.8-200.8) in 1994. During this time the single vaccine option gained
further acceptance as public and professional confidence was restored
following the removal of the Urabe mumps vaccine. The authors note that
beyond 1994 the Kohuku Ward was redistricted but claim no effect of this
on interpretation of the data. It is interesting to note, however, that
the confidence intervals on the point estimates of ASD incidence
increase in parallel with this demographic change. A result of this is
that the precision of the point estimates appears to have been
compromised after this time. ASD incidence beyond 1994 is, therefore, is
not as accurate as preceding years.
The multiphasic shape of the incidence curve is strikingly
different from that seen in the UK (Figure 2) and the US (Figure 3)
where distributions are primarily monophasic (i.e. a continuous rise).
The shape of the Japanese graph would be consistent with an influence of
an additional factor(s) on the evolution of an environmentally induced disease where, overall, exposure to the cause was increasing over time.
In light of the biological nature of viral interactions (‘interference’) and the protracted effects on the immune system of measles exposure in particular (either as natural infection or vaccination) it is evident that, although MMR vaccine itself was discontinued in this infant population beyond 1993, for all practical purposes children vaccinated according to the recommended schedule were still receiving ‘M-M-R’ at age one. In other words the administration of the separate vaccines in close temporal proximity amounts, in biological terms, to overlapping exposure. Such close proximity of exposure is clearly atypical and something that would have been very rare with
natural infection to measles, mumps and rubella viruses. The Japanese
data are therefore not at odds with the original interpretation and the
subsequent recommendations referred to earlier. They are entirely
consistent with what is known about the behavior of these viruses. The
authors of the Japanese study make the error of examining MMR as the
single exposure of interest without giving any consideration to the
arguments that have been put forward or the data upon which those
arguments were based.
In light of these observations the data could be interpreted as
indicating a major influence of the pattern of exposure to these vaccine
viruses on ASD incidence in this Japanese population. Moreover, it
suggests a possible re-challenge effect of close temporal exposure to
these vaccine viruses on ASD incidence at the population level, whereby
the exposure (MMR) has been introduced, removed (voluntarily through
lack of public confidence) and then re-introduced (as M, M and R close
together). Nonetheless the interpretation by Public Health authorities
that this is the ‘last word on the subject’ and that these data prove
that MMR is safe is misleading and suggests a very limited perspective
of the issues and a misunderstanding of the previously published
concerns that have guided the research of those involved with examining
the safety of measles vaccines. Enthusiasm to exonerate the MMR vaccine
is no excuse for misrepresenting the published basis for the safety
Regressive autism: methodological flaws
It is also worth commenting on one major methodological flaw in
the paper. The original description by Wakefield et al and subsequent
studies indicate that any potentially causal relationship between MMR
and ASD relates to a regressive form of autism, in which the child
developed normally prior to exposure.
In the study of Honda et al, children underwent routine
developmental assessment at 3 months and 18 months of age, while the
recommended schedule for MMR vaccination was 12 months of age. The
authors define regression as demonstrable loss of skills after 18 months
of age. Therefore children who have developed normally for the first
year of life, who then receive an MMR at 12 months of age and who
subsequently regress over the course of the next 6 months, will be
misclassified as non-regressive cases when in fact quite the opposite
may be the case. Misclassification of the children’s autism in this way
will render meaningless, the authors sub-analysis comparing regression
and non-regression. This is supported to by the fact that the shape of
the respective incidence curves in the regression and non-regression
sub-groups is similar. The regression data, therefore, do not merit
The authors conclusion that their ‘.findings indicate that simply
terminating MMR vaccination programs will not lead to a reduction in the
incidence of ASD’ is self-evident. The original recommendation however
made no such naÃ¯ve claim. The recommendations were that the vaccines
should be given separately and spaced by one year; this was based on
empirical data, which indicated a serious adverse effect of close
temporal exposure to two or more of these vaccines. The Japanese data
give no reason to change theses recommendations.
Legend to Figure 1.
* The published prevalence of ASD did not exceed 25 per 10,000 at
any time in Japan before the introduction of MMR. This prevalence figure
is therefore an overestimate of the incidence figure in this population.
M-M-R = separate measles, mumps and rubella vaccines.
For graphics: http://www.sarnet.org/img/awart.gif