Is there an autism epidemic?
There is a widespread belief about an 'epidemic' of autism. Much has been made
of a graph produced by the California Department of Developmental Services that
depicted an apparent meteoric increase in number of persons with autism registered
for services. The proponents of the 'epidemic' theory point to the 272.9% increase
in number of individuals with typical autism registered for services in the state
program between 1987 and 1998. In a commentary in the current issue of PEDIATRICS
(February 2001, page 411), Dr. Eric Fombonne provides a critical analysis of the
flawed data and conclusions that have been cited by others as evidence for the
so-called epidemic. Some of the important points are as follows:
The oft-cited California data are neither incidence nor prevalence data. Rather,
it was a simple count of the number of people in the system. The graph reproduced
in Dr. Fonbonne's commentary shows the number of persons by year of birth. It is
not a rate-based statistic (either prevalence or incidence). As expected, younger
persons are more numerous, in part reflecting more accurate diagnosis in recent years.
The California data do not take into account the population growth for the state,
between 1987 and 1999 (+25.8% for 0-14 years old). Another confounding variable
is selective migration into the state from adjoining states and elsewhere by families
with the perception of an increased availability of services for autism in California.
During the period covered by the California data, there was a succession of different
diagnostic systems with a tendency to broaden the criteria for inclusion under the
general heading of pervasive developmental disorders. This more inclusive definition
inevitably resulted in a higher rate of identification in recent years compared to
earlier years, as reflected in the upward trend of the curve for younger individuals.
In recent years there has been a trend to diagnosis at an earlier age, pulling more
cases into the service system each year.
The California report concluded that the number of new persons of all ages diagnosed
with autism exceeded that expected from one annual birth cohort. However, the cited
"cohort" is a single cross-sectional slice of all ages during 1998 and cannot be
reasonably compared to a birth cohort of uniform age. It is simply an illogical
"apples and oranges" comparison. In fact, the report underestimated the expected
number of persons with autism (only 13,054 persons registered as of July 2000)
compared to Fonbonne's estimate (using a prevalence estimate of about 22 per 10,000)
showing 73,334 (all ages) and 21,940 (age 0-19 years).
Finally, other chronic conditions, such as cerebral palsy, epilepsy, or mental
retardation, also showed an upward trend, pointing to the nonspecificity of the
data. There is no evidence that these other conditions are actually increasing
in incidence.
Figure 1 in Fombonne's commentary is a graph from the California report showing
the number of persons with autism in the system in 1991 plotted by birth year.
It is not a prospective cohort study.
A proper epidemiological study would require several years. It would begin with
a random sample, starting at birth, followed over a period of time and at the end
of which the probability is high that most individuals with autism would have been
identified (by age 5 years, for example). Each succeeding year a new birth cohort
would come into the study. At the time the first birth cohort reaches the age of
5 years, the last birth cohort of newly-born would be entering the study, so that
5 years later the last cohort would have been followed and observed for the emergence
of autism. It is also essential during the course of such a prospective cohort study
that unchanging diagnostic criteria be used to identify individuals with autism.
There is no shortcut to a well-designed epidemiological study. The above scheme
would extend over 10 years if each cohort were followed for 5 years. However,
if the study were confined to autistic disorder, which by definition is diagnosable
by age 3 years, then it could be completed in 6 years. The population study might
comprise a random sample or could include the entire childhood population for a city
or state if all children are under observation and most remain available for followup.