Les publications des taux de prévalence de l’autisme et des TED font état de résultats certes divergents mais globalement en augmentation. Les changements dans les protocoles des études réalisées depuis 1966 et d’autres hypothèses explicatives sont exposées et discutées.

Depuis 1943, la définition de l’autisme a d’abord évolué selon les auteurs jusqu’à un accord sur la validité descriptive de l’autisme infantile (CIM 10) ou du trouble autistique (DSM IV). Mais, parallèlement, d’autres catégories diagnostiques, aux contours parfois plus flous, ont élargi le concept de troubles envahissants du développement (TED) — et les enquêtes plus récentes intègrent en proportion significative des formes « légères » ou des formes plus graves de TED, notamment les enfants et adultes avec retard mental qui ont le plus de troubles du comportement et de la relation.

Aussi, les écarts de prévalence dans les travaux et enquêtes épidémiologiques (par exemple, Inserm 2001/FFP-HAS 2005) restent difficiles à interpréter ; ils montrent qu’il convient de lire avec prudence ces chiffres comme indicateurs d’une progression réelle ou apparente de l’incidence du trouble.

Estimates of the prevalence of autism and pervasive developmental disorders (PDD) are discordant and are moving towards an apparent increase in rates.

The studies carried out since 1966 illustrate the variability of the protocols used and explanatory hypotheses put forward. These investigations are difficult, sparse, but still growing at the same time that a debate develops on the possible increase in actual prevalence. Indeed, the rate initially admitted for classic autism was 5/10,000, then 1/1000 with an expanded definition to the forms, but the current figures are very different (almost 0.7% for all PDD), and this increase raises questions. The arguments in favour of an apparent increase are primarily methodological. Several biases are encountered when one compares the recent publications with those of previous years. First, autism is better known and recognized than 30 or 40 years ago. Then, the diagnostic criteria used over time are changing variables, and comparisons difficult. Recent studies using the criteria of a broader definition of autism, polyhandicap with severe retardation and autism signs of lighter forms. The fact that children with autism are diagnosed more frequently in the younger age could also occasionally lead to an artificial increase in the number of cases identified in new surveys in populations of young children. Other factors are cited to explain the current increase. There could be higher rates of autism (and mental retardation) among children of migrants from distant countries, with the aetiological hypothesis of maternal infections, more frequent due to immune deficiency against infectious agents depending on the environment, metabolic decompensations also related to changes in surroundings, or more births from unions among migrant mothers and men with Asperger syndrome (with increased risk of paternity of a child with autism). Other theories relate to pollution, vaccinations, a growing number of premature babies; all assumptions that appear, for the time being, insufficiently explored and documented.

The issue is also one of the motivations underlying these steps, and setting a parallel prevalence actually increased with this or that factor has presently been scientifically validated.

Finally, if a careful reading of recent publications indicates that autism has become more frequent; assumptions that describe an increase in “artificial”, based on methodological arguments, seem to be more consistent.

Today, the recruitment of individuals with autism in a population far exceeds the initial criteria of Kanner in the 1970’s. It includes clinical forms with associated pathologies, or lighter and probably more frequent clinical forms. Other assumptions arouse interest, but also controversy regarding their relevance. The enumeration of cases of PDD in a population is actually at its beginning. In the 1970’s, “childhood psychoses” (the term then used) seemed rare. The identification of cases was probably the main reason. Long available figures remain scarce, and their rate increases gradually from the 1990s, but is, in fact, a problem of inflation. What is the part played in this flight of changing diagnostic criteria and substitutions, or other methodological effects? Or even opportunistic effects, if we speak of an epidemic to undermine a variety of factors. The evidence provided so far is the improved identification of cases, enlargement of the concept, and better shared diagnostic criteria. However, the validity and limitations of clinical forms are still vague and unresolved.

How to study epidemiology in the future – to move forward, studies should be designed with partners’ medical history and medicosocial studies, based on a better consensual methodology, epidemiology, statistics and diagnosis, with a definition of the thresholds for inclusion, and arbitration procedures. On this basis, a study must also be coordinated with those concerning mental retardation, learning disorders, etc, otherwise the same topics will be counted twice or even three times. As for the addition of syndromic forms of PDD (those with known aetiology), their number is still below a proportion sufficient to be an appeal. Moreover, another problem exists: the degree of membership of each of these syndromes, or individual cases, or autistic spectrum disorders (internal variability phenotypes). For the moment, we could design two studies included better: developmental disorders and associated pathologies. Regarding the “ethic” dimension, a more regular diagnosis of PDD (preferred to that of mental retardation or learning disorder) will lead to shared practices and set limits for greater recognition.