In this study we found the prevalence of PDD was at least 82.8 per 10.000 children which is even higher than previously reported. We consider this to be the minimum prevalence rather than the actual prevalence as a quarter of the schools did not respond. This relatively high prevalence may be accounted for by the fact that we concentrated on school-age children rather than on pre-school children like Chakrabarti and Fombonne , who reported a prevalence of 62.6 per 10.000. Fourteen children in our study population had been diagnosed before school entry, which would equal a prevalence of 55.2 per 10.000 at that point in time. However, it is noteworthy that the National Autistic Society (NAS) recently estimated the total prevalence of PDD to be 91 per 10,000 .
Observations of other authors support the impression that the incidence – and thereby also prevalence – of PDD has increased over the last decade [11–14]. Powell et al.  found that the incidence of autistic spectrum disorders in their population appeared to have increased by 55% per year over a five year period (1991–1996). The authors of this paper came to the conclusion that clinicians were becoming increasingly able and willing to make this diagnosis in pre-school children, rather than the data reflecting a "true" rise in number of cases. This theory has since been supported by a number of recent publications [14–16]. Interestingly, a recent epidemiological study in the U.K. has suggested that the incidence of PDD appears to have stabilised in the last few years .
In our study the prevalence of 23.7 per 10.000 for childhood autism was also higher than previously reported by other authors [5, 17, 18]. However, two recently published studies from groups in the United States have reported similar figures. Croen et al. reported a prevalence of 14.9 per 10.000 in Californian children , while the prevalence reported by Bertrand et al. in children from New Jersey was 40.0 per 10.000 .
Our prevalence for Asperger's syndrome (11.8 per 10.000) was lower than reported by Ehlers and Gillberg  (36 per 10.000), but higher than recently published by Chakrabati and Fombonne . Again the explanation for the higher incidence of Asperger's syndrome in comparison to the latter group is likely to be due to the difference in age of our populations. Two of our three patients in this subgroup were only diagnosed after school entry.
The male preponderance we found in our study population of children with PDD was similar to observations published by other authors [5, 17, 21, 22].
It was reassuring to find that all children in our study were assessed and treated by the Speech and Language Therapy team and received additional support at school, with the majority (90.5%) having been statemented for special educational needs. A tendency for improvements regarding statementing and special provisions at school for children with PDD has previously been reported by Smith et al. .
For a considerable proportion of children with PDD (36.8%), no areas of concern were noted at the 2-year child health surveillance check. In contrast to this in nearly all cases of PDD (94.1%) abnormalities had been noted at the 3.5-year check. Abnormal development had been documented in all cases of childhood autism at the 2-year check, but a significant amount of "milder" cases of PDD – including all 3 patients later diagnosed with Asperger's syndrome – had not been detected at that stage.
We found that at the 3.5-year check concerns about speech and language development had been raised in about 3/4 of the PDD cases (vs. only 57.9% at 2-year check) and that behavioural concerns were noted far more frequently (47.1% vs. 5.3%). These results suggest that the 3.5-year check – and to a lesser extent the 2-year check – is a valuable contributing factor in the early detection of PDD.
Another important finding was that 68.4% of the children with PDD were referred for further assessment as a direct result of the routine child health surveillance checks. This illustrates that abnormalities were not only noted and documented at the checks, but that action was taken in a large proportion of children to address these issues appropriately.
Recently the Joint Working Party on Child Health Surveillance (U.K.) has published a new edition of "Health For All Children" . The authors of this document are proposing that no formal universal screening should be carried out at 2 years (i.e. 18–24 months) and 3.5 years (i.e. 39–42 months). Instead they recommend a strategy whereby the health visitor negotiates the nature of subsequent reviews at the age of four months with the parents in the light of individual needs. However, at this age characteristic features of PDD would not have developed. This approach would rely mainly on the parents to recognise autistic features or abnormal development and subsequently seek assistance from health care professionals. Awareness training of play group and nursery staff as suggested by the National Initiative for Autism: Screening and Assessment (NIASA) may be an important tool to overcome this problem .
The guidelines that have been proposed by NIASA conclude that primary screening for PDD by the use of tests applied to the whole population at specific ages cannot be recommended, based on previous reports (e.g. by Baird et al. , who reported a low sensitivity of 38% for the CHAT / Checklist for Autism in Toddlers). However, the group of experts endorsed the recommendation that at specific times in the pre-school years (8–12 months, 2–3 years and 3–5 years) there should be a focus on development by a health professional with the parents or carer.
Early diagnosis of PDD is essential to facilitate early intervention. Various forms of intervention have been shown to be beneficial and improve overall outcome in non-randomised studies [7, 27, 28]. More recently, two small randomised controlled trials have demonstrated that early, intensive therapy can improve the intellectual development, language and visual-spatial skills of the affected child as well as significantly increase parent satisfaction [29, 30]
The mainstays of therapy continue to be speech and language therapy and behavioural therapy. One approach is a TEACCH-based (Treatment and Education of Autistic and related Communication handicapped CHildren) home intervention programme, which has repeatedly been shown to be effective. Ozonoff and Cathcart  were able to demonstrate that children participating in this programme improved significantly more than the autistic children in the control group in the areas of fine motor, gross motor and non-verbal conceptual skills. Another promising approach is the NAS EarlyBird Programme. There is already supporting evidence of its efficacy, although long-term data requires further evaluation . The early intensive behavioural intervention method (EIBI) is another behavioural approach which has been reported to be an effective intervention which may produce long-lasting and significant gains for young children with autism [32, 33]. However, the assessment and evaluation of outcomes of interventions in complex disorders such as PDD is difficult and further studies are needed to evaluate the long-term impact of these methods.
We are aware of the limitations of this study, most notably that standardised assessment tools such as the widely used ADI (Autism Diagnostic Interview) were not employed in the initial diagnostic assessment. The diagnoses are based on assessments mainly conducted by experienced Consultants in Community Paediatrics, who are specifically trained for this role, but are observer dependent. This might have led to over-diagnosing or over-inclusion, especially at the "milder" end of the spectrum of PDD. The second major shortfall of this study was that we did not receive a response from all schools – which may have affected our prevalence. However, our epidemiological findings are consistent with observations reported by other authors, who have employed different methods to conduct their surveys.