Baird, G., Charman, T., Pickles, A., Chandler, S., Loucas, T., Meldrum, D., Carcani-
Rathwell, I., Serkana, D., & Simonoff, E. (2008). Regression, developmental trajectory and
associated problems in disorders in the autism spectrum: the SNAP study. Journal of Autism and
Developmental Disorders, 38, 1827-1836.
Regression, developmental trajectory and associated problems in disorders in the autism
spectrum: the SNAP study
Gillian Baird1, Tony Charman2, Andrew Pickles3, Susie Chandler1, Tom Loucas4, David
Meldrum5, Iris Carcani-Rathwell6, Devanitha Serkana7, Emily Simonoff8
1 Guy’s & St Thomas’ NHS Foundation Trust
2 UCL Institute of Child Health, London
3 Biostatistics, Health Methodology Research Group, University of Manchester
4 School of Psychology and Clinical Language Sciences, University of Reading
5 Chatswood Assessment Centre
6 Greenwich Neurodevelopmental CAMHS, Oxleas NHS Trust
7 Cheyne Child Development Centre, Chelsea & Westminster Hospital
8 Department of Child and Adolescent Psychiatry, Institute of Psychiatry, King’s College
Running Head: Regression in ASD
Regression in ASD 2
We report rates of regression and associated findings in a population derived group of 255
children aged 9-14 years, participating in a prevalence study of autism spectrum disorders (ASD);
53 with narrowly defined autism, 105 with broader ASD and 97 with non-ASD
neurodevelopmental problems, drawn from those with special educational needs within a
population of 56,946 children. Language regression was reported in 30% with narrowly defined
autism, 8% with broader ASD and less than 3% with developmental problems without ASD. A
smaller group of children were identified who underwent a less clear setback. Regression was
associated with higher rates of autistic symptoms and a deviation in developmental trajectory.
Regression was not associated with epilepsy or gastrointestinal problems.
Key words: SNAP, Autism, Regression, Outcome, Epilepsy, Gastro-intestinal problems
Regression in ASD 3
Retrospective histories from parents and analysis of home videotapes have shown that for
most children with autism, abnormalities in development become clear prior to 2 years of age
(Adrien et al., 1991; Dahlgren & Gillberg, 1989; Losche, 1990; Osterling & Dawson, 1994;
Rogers & DiLalla, 1990; Werner et al., 2005; Werner & Dawson, 2005) and often within the first
year of life. However, a feature of autistic development that remains a puzzle is that some
children present with apparently normal development as perceived by parents followed by quite
marked cessation of skill acquisition and frequent loss of, or failure to use, acquired language and
social skills. Commonest in the second year of life, this has been termed ‘autistic regression’, and
occurs in 15-40% of children with autism (Fombonne & Chakrabarti 2001; Kurita, 1985; Lord et
al., 2004; Luyster et al., 2005; Prizant 1996; Richler et al., 2006; Tuchman & Rapin, 1997).
Some parents report that there is a very abrupt change in their child’s development and
behaviour, others report a much more gradual change lasting weeks. In some cases parents report
that development has been normal prior to the regression (although more detailed examination
and retrospective videos may indicate some subtle impairment), others report that although there
was some delay in acquiring skills they still note significant regression. Even in those children
where there is no obvious loss of skills, stasis of cognitive and social development may be
reported by parents and has now been found in studies of at risk infants-the siblings of children
with autism (Landa & Garrett-Mayer, 2006; Landa, Holman & Garrett-Mayer, 2007).
A common operational definition of regression is a loss of spoken language after the first
3-5 word stage of acquisition (Le Couteur et al, 1989; Lord et al., 1994; Lord et al., 2004;
Tuchman & Rapin, 1999). There is also usually loss of non-verbal communication (e.g. gestures
such as waving bye-bye) frequently decreased use of eye gaze to regulate social interaction, some
social withdrawal and lack of social interest, and sometimes a loss of play skills (Ozonoff et al.,
2005; Werner et al., 2005). Gross motor development is not usually affected although some
Regression in ASD 4
parents note a change in fine manipulative skills (Davidovitch et al., 2000). However, in children
who have not reached the 3-5 word stage of language development, parents may note regression
of babble and proto-words with or without regression in social interest, gestures etc. The period
of developmental stasis or loss of skill-use is usually followed by a regaining of skills but at
varying rates. Some children never regain lost skills (Evans-Jones & Rosenbloom, 1978; Lord et
Several reports have suggested that the eventual outcome in children with regression is
that of a lower language level, lower IQ and lower adaptive level compared with those who do
not regress (Kurita, 1985; Hoshino et al., 1987; Kobayashi & Murata, 1998; Rogers & DiLalla,
1990). However, other studies have found no difference in outcome (Chakrabarti & Fombonne
2001; Davidovitch et al., 2000; Lord et al., 2004) or mixed results with the regressed group
showing a bimodal outcome in verbal IQ and social reciprocity (Richler et al., 2006) which may
be a result of inclusion of differing groups within the autism spectrum.
Regression to autism in older children following a period of clearly normal development
up to the age of at least 2 years is classified separately in ICD-10 (WHO, 1993) and DSM-IV
(APA, 1994) as Developmental or Other Childhood Disintegrative Disorder (CDD). Previously
titled Heller’s syndrome, regression is in language, communication, social play, curiosity in the
environment, sometimes loss of bowel/bladder function and the onset of stereotyped skills. This
is a rare phenomenon. Fombonne (2002) estimates a prevalence of CDD of no more than
0.2/10,000. In several studies the CDD group have a poorer outcome in terms of cognitive and
functional skills (Malhotra & Gupta 2002; Volkmar & Cohen, 1989), though Kurita et al. (2004)
found no such difference.
The process underlying regression and stasis is unknown. There has been speculation that
the anatomical remodelling of the brain with synaptic growth and pruning during the second year
Regression in ASD 5
of life is impaired in autism due to gene-based mechanisms (Carper & Courchesne, 2005)
resulting in variable effect on function. Also a suggestion that regression constitutes a genetically
different disorder (Molloy et al., 2005), unconfirmed in the IMGSAC sample where in families in
which more than one child has autism, regression occurred in one sib and not another (Parr et al.,
2006 and in preparation), although newer genetic techniques may cast further light on possible
genetic contributions (Marshall et al., 2008; Weiss et al., 2008). Whether the regressive process is
influenced by environmental factors is also unknown. There has been concern that the number of
children with regressive autism has increased but recent reviews (Fombonne & Chakrabarti,
2001; Taylor et al., 2002) have shown no such increase. Considerable research has failed to
support an association between one suggested environmental factor, MMR immunisation,
postulated to be linked with enterocolitis and the risk of a regressive autistic disorder (Baird et al.
2008; DeStefano, 2002; Honda et al., 2005; Richler et al. 2006). Other suggestions have included
epilepsy as a causative factor in regression allied to Landau-Kleffner syndrome, an aphasia due to
localised peri-sylvian epilepsy, (Robinson et al., 2001), although most children with regression of
language who have autism do not have epileptic seizures and language regression with autism is
not more common in those with epilepsy than those without epilepsy (Shinnar et al., 2001;
Tuchman & Rapin, 1997).
As part of a prevalence study of autism and related pervasive developmental disorders
(commonly called ‘autism spectrum disorders’; ASDs) we assessed a group of 255 9-to-14 year
old children with and without ASD drawn from a geographically defined population rather than a
clinically referred group. A sample weighting procedure enabled us to estimate characteristics of
the total population of children with autism and ASD. This study provided us with the
opportunity to examine the following questions regarding the nature, timing, consequences and
associated features of regression in children with autism, other ASDs and children without ASD
Regression in ASD 6
with mental retardation, learning difficulties and behavioural disorders: a) Does regression affect
developmental trajectory and outcome?; b) Does regression occur in non ASD cases?; c) Is there
a greater prevalence of gastro-intestinal problems or epilepsy in the regressed versus non-
The population studied is a cohort of 56,946 children born between July 1st 1990 and December
31st 1991 from 12 districts in the South Thames region of the UK. Children with a statement of
special educational needs (SEN) (1733; 218 of whom had a local ASD diagnosis) or a local
diagnosis of ASD but no SEN statement (37) were screened using the Social Communication
Questionnaire (SCQ; Rutter et al., 2003). The mean (SD) age at SCQ screening in the whole
cohort was 10.3 (0.4) years.
A subset of children, stratified by local diagnosis and high, medium and low SCQ score
(255) received an in-depth diagnostic assessment (see Figure 1 in Baird et al., 2006). The
diagnostic assessment included standardized clinical observation (Autism Diagnostic Observation
Schedule – Generic (ADOS-G; Lord et al., 2000)) parent interview assessments of autism
symptoms (Autism Diagnostic Interview-Revised (ADI-R; Lord et al., 1994)) and assessment of
IQ. Children were classified using ICD-10 research criteria as ASD or no ASD by clinical
consensus using all sources of information. The ASD group was divided into a ‘broad ASD’
(105) and ‘narrow autism’ (53) group, the latter defined as meeting autism criteria on the ADI-R,
the ADOS-G and clinical consensus of ICD-10 childhood autism and the broad ASD defined as
all other cases meeting clinical consensus of any ASD. The total number of ICD-10 autism
symptoms was recorded. The ‘no ASD’ group (97) had a variety of diagnoses: intellectual
disability (mental retardation), specific language or literacy impairments, ADHD/ODD, cerebral
Regression in ASD 7
palsy, deafness and visual impairment. Collectively they form a control group for some of the
The ADI-R has specific questions (Items 11-15) about regression of language and other skills.
Regression was defined in two ways. The first adopted the ADI-R definition of strict language
regression as ‘loss of 5 words used communicatively for 3 months before loss’ with or without
loss of skills in other areas, a group called ‘definite language regression’. An additional group
was defined where the parent described stasis or loss of words or babble, but where the child had
not reached the 5-word stage or there was regression of skills other than language (Q20 in ADI
2000), called the ‘lower level regression’ group.
A systematic enquiry was additionally made of early development using items based on
the Diagnostic Interview for Social and Communication Disorders (with permission) (DISCO;
Wing et al., 2002) that supplemented the ADI items on language acquisition. The DISCO was
developed for systematic enquiry about a range of normal and abnormal behaviours but does not
have population ‘norms’. 17 questions about the normality of development of sucking, babble,
gesture, play and social responsiveness in the first year were used and scored as described by the
authors yielding a range of scores from 0 to 34 to give a single figure as proxy for normal early
development, a higher score indicating greater abnormality. Both the ADI-R and DISCO
questions rely on retrospective memory. Contemporaneous child health records were available in
the majority of cases (79%) and were systematically searched to look for age of concern, age of
referral, medical problems and any contemporaneous note of developmental problems or
regression to validate the parental history. No case met criteria for Rett syndrome or CDD.
Epilepsy was enquired about twice, once during the ADI-R and at subsequent interview
about medical conditions. Seizures were classified as febrile only, epileptic past or current (on
Regression in ASD 8
treatment) and non-epileptic (e.g. reflex anoxic seizures). Medical notes were checked for
corroboration of epilepsy. Gastrointestinal (GI) symptoms, reflecting the presentation of GI
symptoms in general clinical paediatric practice, for the last three months (current) and at any
point in the past, , were assessed using a questionnaire completed by the main caregiver (Circani-
Rathwell et al., in preparation). Current symptoms elicited were of vomiting (occurring at least
once per day or more than five times in a week); diarrhoea (defined as loose/watery stools three
or more times a day >14 days); persistent abdominal pain (three or more episodes severe enough
to interfere with activity) constipation (defined as a bowel action<three times per week); weight
loss; blood in stools and soiling. Past symptoms of vomiting, diarrhoea, abdominal pain (defined
as above) and stool withholding were also elicited. The four symptoms of vomiting, diarrhoea,
abdominal pain and constipation were summated to give a possible score of 0-4 either of past or
current GI symptoms. A ‘possible enterocolitis’ group was constructed from the presence of 2 or
more of the following 5 current gastro intestinal symptoms: persistent diarrhoea, persistent
vomiting, weight loss, persistent abdominal pain; blood in stool; plus past diarrhoea >14 days
duration and excluding current constipation. 87 children were screened for coeliac antibodies (in
whom a sufficient blood sample was obtained) and the single child (from the control group)
found to be positive, but asymptomatic, was excluded from the gastro-intestinal analysis together
with eight children with cerebral palsy who might be expected to have motility or upper GI
Measures used were IQ, adaptive behaviour on the Vineland Scales (Sparrow et al., 1984)
and severity of autism symptoms using an ICD-10 ASD symptom score (0-12). IQ was measured
using the Wechsler Intelligence Scale for Children (N=209; WISC-III- UK; Wechsler, 1992),
Raven’s Standard Progressive Matrices (SPM) or Coloured Progressive Matrices (CPM; Raven,
1990a,b), depending on the child’s ability. For the 35 cases where SPM or CPM but not WISC
Regression in ASD 9
full scale IQ's were available, imputed full-scale IQ's were obtained using the regression
relationship of full to SPM/CPM IQ within each diagnostic group (conversion to IQ from
Catherine Lord, personal communication February 2008). For the 11 cases where no direct
cognitive testing was possible all cases had Adaptive Behaviour Composite on the Vineland
Adaptive Behaviour Scales (Sparrow et al., 1984) below 20 and these cases were assigned an IQ
score of 19 to reflect their profound level of mental retardation.
The text includes data on the exact numbers of children with and without regression seen in the
study. All subsequent analyses presented in the tables and text are adjusted using stratification
weights. Stratification of the ASD/SEN sample was based on whether or not a child had a locally
recorded ASD diagnosis (yes/no) and 4 levels of SCQ score (low score (<8), moderately low
score (8-14), moderately high score (15-21), high score ( >22); see Baird et al., 2006,Figure 1,
p.212 for details). Use of weights allowed all statistics such as proportions, means, group
differences and screen performance measures to be presented as target population estimates,
taking account not only of the differences in sampling proportions according to SCQ score and
local ASD diagnosis, but also the differential response to the SCQ associated with a prior local
ASD diagnosis, health district and child’s sex. Standard errors of simple means and regression,
logistic regression and proportional hazards regression coefficients and contrasts, Wald test
statistics and p-values were calculated using the linearisation version of the robust parameter
covariance matrix as implemented by the svy procedures of Stata 9 (Stata, 2005). Confidence
intervals for the rate of regression were estimated at the 2.5th and 97.5th percentiles from 1000
Regression in ASD 10
In the 255 cases assessed, regression was reported in 38 cases, 28 with definite language
regression and 10 with lower level regression who are reported separately. Table 1 shows the
number of cases, weighted rates and 95% confidence intervals (CIs) according to regression
category and diagnosis.
The definite language regression group
Of the 28 children who met criteria for ‘definite language regression’, 26 had ASD and 2
did not. The rate of definite language regression was significantly higher (30.2%) in the narrow
autism group than in the broad ASD group (8%, p=.01) and the no ASD group (2.8%, p=.003).
The rate of definite language regression did not significantly differ between the (combined) ASD
group (12.6%) and the no ASD group (p=.08). 16.3% regressed in one or more area other than
language: 5.5% were reported to have lost purposive hand movements (but did not follow a
trajectory typical of Rett syndrome), 10.5% motor skills, 2.4% self-help skills, 19.4%
constructive or imaginative play and 19.4% were reported to have regressed in the area of social
engagement/responsiveness. Association with illness (regression reported by parents as occurring
within 7 days of an illness) was reported in 11 children: non-encephalopathic illness in 8 with
ASD (two parents reported that illness and regression followed within 2 weeks of the MMR); 1
ASD case presented aged 1 year with frequent epileptic seizures and had a left temporal tumour
subsequently removed. Of the 2 remaining non-ASD cases, one had a definite encephalitis, the
other was a child with Down’s syndrome who developed leukaemia.
Table 1 about here
Age at regression
Regression in ASD 11
The mean weighted age of regression was 25.0 (SE 1.5) months for the cases with definite
language regression. For the two non-ASD cases with clear language regression the age of loss
was 20 months and 48 months, respectively. The only other case with age of language loss
greater than 33 months was one case with autism whose parents, on the ADI-R, reported
plateauing of development at 24 months and then loss of language at 69 months of age, although
contemporaneous health records indicated parental report of language loss at 24 months of age
and they also reported that development prior to language loss was not normal The pattern of
development was thus not consistent with CDD.
Outcome at 9-14 years of definite language regression compared with no regression
Outcome was examined in terms of IQ, Vineland composite scores, and ICD-10 ASD symptom
score. Table 2 shows mean differences and pair-wise comparisons. The effects of regression
controlling for diagnosis were tested in multivariate regression models. With regard to IQ and
Vineland scores, there was no significant difference between regression and non regression once
diagnostic category was taken into account. The ICD-10 ASD symptom score was significantly
greater in the regression group than the no regression (non-standardized B =3.25, p<.001). The
effect remained significant when diagnosis (broad ASD or narrow autism) was accounted for.
Age at regression was not significantly associated with outcome IQ, Vineland scores or ICD-10
ASD symptom score.
Table 2 about here
Validation of parental history of regression
Contemporaneous casenote information was available for 16 of the 28 cases with clear
regression. Loss of skills or stasis/plateau was documented by paediatricians (from parental
Regression in ASD 12
report) in 11/16 cases (69%). For these 11 cases the age of regression recorded in casenotes was
25.1 (SE=1.9) months compared to 28.4 (4.7) reported by parents in the ADI-R, a difference that
was not significant (paired t-test; t=0.65, p>.10). The discrepancy between the reported ages was
12 months or less in N=8 cases and greater than 12 months in N=3 cases, including the case
where contemporaneous casenotes indicated a parent-reported loss of language from more than
20 words at age 24 months but loss was subsequently reported at age 69 months in the ADI-R.
Language development and regression
Age of first words and age of first phrases (weighted) as reported by parents during the ADI-R
are shown in Table 3. At the time of our assessment, amongst those with autism or ASD, only 1
child with an ASD had not attained single words. The age of acquiring first words in the definite
language loss group is significantly younger than the no regression group (B=-30.6, p=.03); these
findings remained significant when diagnosis was accounted for. There was no significant
difference between the definite language regression and no regression groups in age of acquiring
phrase speech, either on its own, or with diagnosis added as a covariate.
We examined whether the failure to achieve phrase speech varied according to regression
group. Twelve children with a ASD had not achieved phrase speech by the time of the
assessment, of whom 2 had a ASD diagnosis and 10 an autism diagnosis representing 8% of the
non-regression group, and 9% of the definite language regression group. A Cox proportional
hazards model of the time to phrase speech that took into account the censored times from those
who had not achieved phrase speech, confirmed the absence of significant regression group
differences, particularly when controlling for delay in phrase acquisition due to diagnosis
(hazard ratio 0.61, p=.03).
Early developmental skills, developmental trajectory and regression
Regression in ASD 13
The definite language regression group had lower DISCO total scores (indicating less
abnormality) than the no regression group (B=-3.7, p=.003) and remained so after diagnosis was
accounted for (B=-4.4, p<.001).
Table 3 about here
The relationship between outcome (symptom severity), regression and early development,
was then explored by predicting in a linear regression model symptom severity from early
development score, level of regression and interaction between the two independent variable A
significant interaction (t=-2.17, p= 0.03) was found, such that the early development score was
significantly positively related to later symptom severity in the no regression and lower level
regression groups but unrelated in the definite regression group (Wald test F(1,150)=1.12, p=
‘Lower level’ regression in ASD
Parents of 10 children reported symptoms that met the criteria for ‘lower level’ regression. The
rate of lower-level regression was significantly higher (8.4%) in the narrow autism group than in
the broad ASD group (2.6%, p=.04) and the no ASD group (0.4%, p=.002). The (combined) ASD
group was significantly more likely to show lower level regression (p=.02). Of the 10 children
with lower-level regression, 9 had ASD. Regression was not associated with illness in the 9 with
ASD; the one child without ASD regressed in motor skills only having had a cerebrovascular
event complicating an ear infection and resulting in cerebral palsy. Of the 9 with ASD, two
(20%) lost babble or words, 6 (60%) lost social engagement or play skills, 1 lost hand and self-
help skills. Contemporaneous casenote information was available on 7/10 of the lower-level
regression cases. Five of these 7 (71%) have a note of some type of regression or stasis. The age
Regression in ASD 14
at regression was 25.0 (SE 3.3) months for cases with lower level regression (not different to
definite language regression).
Prior to regression, the score for DISCO items in the lower level regression group (mean
10.6) was similar to that in the no regression group (mean 10.3). Hence, there was greater
developmental impairment than in the definite regression group (mean 6.6), although this
difference was not significant due to small sample size in the lower level regression group. The
lower level regression group had a significantly higher ICD-10 ASD symptom score than the no
regression group. (B=2.06, p=.003) and the definite language regression group (B=2.31, p=.01).
36% of the low-level regression group had not developed phrase speech at the time of assessment
(compared with 8% and 9 % of the no regression and language regression groups, respectively).
Both regression groups combined
The outcome of both definite language and lower level regression combined is similar to each in
that the main effect is on increase in ICD10 symptoms and thus a diagnosis of autism rather than
Association of regression with epilepsy or bowel problems
The weighted rates of epilepsy are shown in Table 4. 18% had a past or current history of
epilepsy. 8% have current epilepsy. There was no evidence suggestive of differential rates of
febrile seizures, past or current epilepsy when comparing the definite and no regression groups.
Although past and current epilepsy are highest in the group with lower level regression, the
difference is not statistically significant due to the small sample size in this cell. .
Table 4 about here
Regression in ASD 15
Table 5 shows the weighted mean symptom count for current and past gastrointestinal problems.
Current symptoms varied across regression groups (F(2,122)=11.96, p=.001), but the rate was
higher in the no-regression group than the lower (F(1,122)=7.09, p=0.0007) or definite regression
(F(1,122)=4.70, p=0.03) groups . No significant group difference was found in past gastro-
intestinal symptoms (F(2,121)=2.84, p=0.6). ‘Possible enterocolitis’, as defined above was
reported in one child who did not have ASD, also did not regress. No child had a previous
diagnosis of inflammatory bowel disorder.
Table 5 about here
Regression in non ASD SEN cases
Clear language regression occurred in 2 non-ASD cases at ages 20 months and 48 months, one
had encephalitis, the other was a child with Down’s syndrome severely ill with leukaemia . The
one non-ASD case with lower level regression had the cerebral incident at age 9 months.
In this study, regression is confirmed as a feature of ASD development. It is rare in children who
do not have ASD and in these cases if it occurs is likely to be in association with a neurological
illness. We have found that the main effect of a history of regression in autism is an outcome of
increased symptom score and more severe autism as shown by diagnostic category. This is true
for both definite language regression and ‘lower level’ regression. To investigate the important
question of whether regression as a feature of autism presentation exerts an additional effect on
potential developmental outcome, we have used a measure of development in the first year as a
proxy for developmental competence and compared the predicted trajectory from the DISCO
score in non regressed and regressed groups. The accuracy of parental recall in the DISCO items
Regression in ASD 16
is unknown. However, the early DISCO score does predict outcome in the non-regressed group
and our results suggest that there is an expected continuity in development which is displaced by
regression. Thus, despite the definite language regression group showing more typical
development in infancy evidenced by earlier first words and less abnormal social communication
in the first year (lower DISCO scores), that early trajectory is not maintained (cf. Landa et al.,
2007; Pickles et al., under review). The neurodevelopmental abnormality that underlies this
deviant developmental trajectory remains to the determined.
This study has used two levels of regression, one clearly defined, the other looser but
based on clear parental report of stasis and loss of babble or 1-2 words plus or minus loss in other
areas. It remains unclear whether the aetiological or pathological process differs between definite
language regression and ‘lower level’ regression. Although reported here separately to enable
comparison with other studies, our results show that the two regression groups show common
trends in association with diagnostic group and effects on outcome.
Recent studies reporting the development of siblings of children with autism who go on to
develop autism suggests that there is stasis and plateauing of the rate of development in the
second year (Landa & Garrett-Mayer, 2006; Landa et al., 2007). Thus, overt regression may lie
on a continuum of no arrest in developmental progress through plateauing to frank regression and
the manifestation of the regressive process appears to depend on the stage of brain maturation and
of development the child has reached rather than their chronological age (Pickles et al., under
review). No case in this study met full criteria for Rett syndrome or CDD. The one case that by
parental report on the ADI-R regressed at 69 months was not totally normal prior to language
loss, plateauing was reported at 24 months and indeed examination of contemporaneous health
records indicated parent report of loss of language at 2 years of age.
Regression in ASD 17
Fombonne and Chakrabarti (2001), using a similar definition of `clear language
regression’, found no differences in outcome between regressed and non-regressed. Richler et al.
(2006) did find lower VIQ scores and higher (more impaired) social reciprocity ADI-R scores
following regression but also found a bimodal distribution of VIQ scores in their regression
group. That study had much larger numbers than ours (163 with regression and 188 without) and
hence greater power to detect differences. However, there were also differences in methods:
Richler et al (2006) used a less stringent diagnostic criterion for autism and a broader definition
of regression. Neither of these studies attempted to predict developmental trajectory from
reported developmental status prior to regression.
We defined gastro-intestinal problems in a standard way. Several studies have found high
reported rates of gastro-intestinal symptoms in ASD (CPEA study, Richler et al., 2006,
Valicentini-McDermott et al., 2006). The choice of 14 days for diarrhoea symptoms chosen in
this study may be regarded as too short and reflective of acute illness rather than chronic GI
disorder, however we found no evidence of more current or past gastrointestinal problems in
regressed versus non-regressed groups. This finding is in contrast with the larger CPEA study
although there are differences in the questions asked. For example, the CPEA study enquired
about ‘change in stool frequency and consistency’ rather than the specific stool frequency
indicating constipation as in the present study. The conclusions in their paper point out that had
corrections been made for multiple comparisons of data, the differences would have no longer
We found no evidence for excess epilepsy in regressed versus non-regressed groups past
or current, which is consistent with Tuchman (1999). Neither did regression signal increased
epilepsy with age (to 11-14 years). No child had a diagnosis of Landau-Kleffner syndrome. The
Regression in ASD 18
reported age of language regression in this study (25 months) is slightly later than in some other
studies. There is variation from a mean 16 months in Lord et al. (2004) and Ozonoff et al. (2005)
to around 20 months in many studies. Tuchman (1999) noted the age of regression to be 12 to 48
months. Some variation in age of reported regression may be accounted for by age at reporting.
The sample of Lord et al. (2004) was 4-5 years old at their most recent ADI-R) but the large
CPEA study (Luyster et al., 2005) were interviewed at a mean of 9 years and mean age of
reported word loss was 20 months.
Importantly, regression of social, language or motor development rarely occurred in
children with non-ASD neurodevelopmental problems (see also Pickles et al., under review who
show that regression is rare in children with language disorder), and then in association with
Strengths and limitations of the study
The present study reported on a carefully ascertained and assessed population derived sample and
thus free of the bias usually associated with a clinically referred sample. It included non ASD as
well as narrow autism and broader ASD cases. The use of a statistical weighting procedure
enables generalisation of the findings to the unselected population, though at the expense of
precision (note the wide confidence intervals for some estimates). Another strength of the present
study was to be able to corroborate parental reporting of regression and medical problems from
contemporaneous health records although with only 57% of records available, and only 69% of
those documenting regression, positive corroboration exists for <40% of the reported regression
cases. The limitations are the relatively small number of cases with regression and in common
with most other studies investigating regression in ASD, the reliance on parental report regarding
the nature and timing of regression as well as pre-regression development.
Regression in ASD 19
In young children presenting to child health services with concerns about development, the
spectrum of autistic disorders are among the commonest of the developmental disorders. One
feature of the history that is particularly important to take note of is regression. In the absence of
an acute neurological event or neurological signs including epilepsy, regression in a child of 1- 3
years should be a ‘red alert’ for assessment of autism and signals an altered trajectory of
development (Filipek et al., 2000).
Regression in ASD 20
American Psychiatric Association. (2000). Diagnostic and Statistical Manual of Mental
Disorders 4th Edn. – Test Revision (DSM-IV-TR). Washington, DC: American Psychiatric
Adrien, J. L., Faure, M., Perrot, A., Hameury, L., Garreau, B., Barthelemy, C. et al. (1991).
Autism and family home movies - preliminary findings. Journal of Autism and
Developmental Disorders, 21, 43-49.
Baird, G., Charman, T., Baron-Cohen, S., Cox, A., Swettenham, J., Wheelwright, S. & Drew, A.
(2000). A screening instrument for autism at 18 month of age: A six-year follow-up study.
Journal of the American Academy of Child and Adolescent Psychiatry, 39, 694-702.
Baird, G., Robinson, R., Boyd, S. & Charman, T. (2006). Regression in young children with
autism: should a sleep EEG be a recommended investigation? Developmental Medicine
and Child Neurology, 48, 604-608.
Baird, G., Simonoff, E., Pickles, A., Chandler, S., Loucas, T., Meldrum, D. & Charman, T.
(2006). Prevalence of pervasive developmental disorders in a population cohort of children
in South East Thames: The Special Needs and Autism Project (SNAP). The Lancet, 368,
Baird, G., Pickles, A., Simonoff, E., Charman, T., Chandler, S., Loucas, T., Meldrum, D., Afzal,
M., Thomas, B., Jin, L. & Brown, D. (2008). The measles virus and antibody response to
infection and vaccination in autistic spectrum disorder: a virological case control study.
Archives of Disease in Childhood. doi:10.1136/adc.2007.122937
Berument SK, Rutter M, Lord C, Pickles A, Bailey A. (1999). Autism screening questionnaire:
diagnostic validity. British Journal of Psychiatry, 175, 444-451.
Regression in ASD 21
Carper, R.A., & Courchesne, E. (2005). Localized enlargement of the frontal cortex in early
autism. Biological Psychiatry, 57, 126-33.
Chakrabarti, S. & Fombonne, E. (2001). Pervasive developmental disorders in preschool
children. Journal of the American Medical Association, 285, 3093-3099.
Chakrabarti, S. & Fombonne, E. (2005). Pervasive developmental disorders in preschool
children: Confirmation of high prevalence. American Journal of Psychiatry, 162, 1133-
Dahlgren, S. O. & Gillberg, C. (1989). Symptoms in the 1St 2 Years of Life - A Preliminary
Population Study of Infantile-Autism. European Archives of Psychiatry and Clinical
Davidovitch, M., Glick, L., Holtzman, G., Tirosh, E., & Safir, M. P. (2000). Developmental
regression in autism: Maternal perception. Journal of Autism and Developmental
Disorders, 30, 113-119.
DeStefano, F. (2002). MMR vaccine and autism: a review of the evidence for a causal
association. Molecular Psychiatry, 7, S51-S52.
Evans-Jones, L.G., & Rosenbloom, L. (1978). Disintegrative psychosis in childhood.
Developmental Medicine and Child Neurology, 20, 462-70.
Filipek, P. A., Accardo, P. L., Ashwal, S., Baranek, G. T., Cook, E. H., et al. (2000). Practice
Parameters: Screening and diagnosis of autism. Neurology, 55, 468-479.
Fombonne, E. & Chakrabarti, S. (2001). No evidence for a new variant of measles-mumps-
rubella-induced autism. Pediatrics, 108, e58.
Green, H., McGinnity, A., Meltzer, H., Ford,T., & Goodman, R. (2005). Mental Health of
Children and Young People in Great Britain, 2004. ONS, HMSO: London.
Regression in ASD 22
Honda, H., Shimizu, Y., & Rutter, M. (2005). No effect of MMR withdrawal on the incidence of
autism: a total population study. Journal of Child Psychology and Psychiatry, 46, 572-579.
Hoshino, Y., Kanako, M., Yashima, Y., Kumashiro, H., Volkmar, F., & Cohen, D. (1987).
Clinical features of autistic children with setback course in their infancy. Japanese Journal
of Psychiatry and Neurology, 41, 237-245.
Kobayashi, R., & Murata, T. (1998). Setback phenomenon in autism and long-term prognosis.
Acta Psychiatrica Scandinavica, 98, 296-303.
Kurita, H. (1985). Infantile-autism with speech loss before the age of 30 months. Journal of the
American Academy of Child and Adolescent Psychiatry, 24,191-196
Landa, R., & Garrett-Mayer, E. (2006). Development in infants with autism spectrum disorders: a
prospective study. Journal of Child Psychology and Psychiatry, 47, 629-638.
Landa, R.J., Holman, K.C., & Garrett-Mayer, E. (2007). Social and communication development
in toddlers with early and later diagnosis of autism spectrum disorders. Archives of General
Psychiatry, 64, 853-864.
LeCouteur, A., Rutter, M., Lord, C., Rios, P., Robertson, S., Holdgrafer, M. et al. (1989). Autism
Diagnostic Interview - A standardized investigator-based instrument. Journal of Autism
and Developmental Disorders, 19, 363-387.
Lingam, R., Simmons, A., Andrews, N., Miller, E., Stowe, J., & Taylor, B. (2003). Prevalence of
autism and parentally reported triggers in a north east London population. Archives of
Disease in Childhood, 88, 666-670.
Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism Diagnostic Interview-Revised - A revised
version of a diagnostic interview for caregivers of individuals with possible pervasive
developmental disorders. Journal of Autism and Developmental Disorders, 24, 659-685.
Regression in ASD 23
Lord, C., Risi, S., Lambrecht, L., Cook, E. H., Leventhal, B. L., DiLavore, P. C. et al. (2000).
The Autism Diagnostic Observation Schedule-Generic: A standard measure of social and
communication deficits associated with the spectrum of autism. Journal of Autism and
Developmental Disorders, 30, 205-223.
Lord, C., Shulman, C., & DiLavore, P. (2004). Regression and word loss in autistic spectrum
disorders. Journal of Child Psychology and Psychiatry, 45, 936-955.
Losche, G. (1990). Sensorimotor and action development in autistic children from infancy to
early childhood. Journal of Child Psychology and Psychiatry, 31, 749-61.
Luyster, R., Richler, J., Risi, S., Hsu, W. L., Dawson, G., Bernier, R. et al. (2005). Early
regression in social communication in autism spectrum disorders: A CPEA study.
Developmental Neuropsychology, 27, 311-336.
Maestro, S., Muratori, F., Cesari, A., Pecini, C., Apicella, F., & Stern, D. (2006) A view to
regressive autism through home movies. Is early development really normal? Acta
Psychiatria Scandanavica, 113, 68-72.
Malhotra, S., & Gupta, N. (2002). Childhood disintegrative disorder - Re-examination of the
current concept. European Child and Adolescent Psychiatry, 11,108-114.
Marshall, C. R., Noor, A., Vincent, J. B., Lionel, A. C., Feuk, L. et al. (2008). Structural variation
of chromosomes in autism spectrum disorder. American Journal of Human Genetics, 82,
Molloy, C. A., Keddache, M., & Martin, L. J. (2005). Evidence for linkage on 21q and 7q in a
subset of autism characterized by developmental regression. Molecular Psychiatry, 10,
Regression in ASD 24
Mouridsen, S.E., Rich, B., & Isager, T. (1999). Epilepsy in disintegrative psychosis and infantile
autism: a long-term validation study. Developmental Medicine and Child Neurology, 41,
Osterling, J., & Dawson, G. (1994). Early Recognition of Children with Autism - A Study of 1St
Birthday Home Videotapes. Journal of Autism and Developmental Disorders, 24, 247-257.
Ozonoff, S., Williams, B. J., & Landa, R. (2005). Parental report of the early development of
children with regressive autism - The delays-plus-regression phenotype. Autism, 9, 461-
Parr, J.R., Lamb, J.A., Bailey, A.J., Monaco, A.P., & The International Molecular Genetic Study
of Autism Consortium (IMGSAC) (2006). Response to paper by Molloy et al: Linkage on
21q and 7q in autism subset with regression. Molecular Psychiatry, 11, 617-619
Pickles, A., Simonoff, E., Conti-Ramsden, G., Falcaro, M., Simkin, Z., Charman, T., Chandler,
S., Loucas, T., & Baird, G. (under review). Loss of language in early development of
autism and specific language impairment.
Prizant, B.M. (1996). Brief report: communication, language, social, and emotional development.
Journal of Autism and Developmental Disorders, 26, 173-178.
Raven, J.C., Court, J.H., & Raven, J. Coloured Progressive Matrices. Oxford: Oxford University
Raven, J.C., Court, J.H., & Raven, J. Standard Progressive Matrices. Oxford: Oxford University
Richler J, Luyster, R, Risi S, Hsu WL, Dawson G, Bernier R et al. (2006). Is there a ‘Regressive
Phenotype’ of autism spectrum disorder associated with the measles-mumps-rubella
vaccine? A CPEA Study. Journal of Autism and Developmental Disorders, 36, 299-316.
Regression in ASD 25
Robinson, R. O., Baird, G., Robinson, G. & Simonoff, E. (2001). Landau-Kleffner syndrome:
course and correlates with outcome. Developmental Medicine and Child Neurology, 48,
Rogers, S.J., & DiLalla, D. (1990) Age of symptom onset in young children with pervasive
developmental disorders. Journal of the American Academy of Child and Adolescent
Psychiatry, 29, 863-872.
Shinnar, S., Rapin, I., Arnold, S., Tuchman, R. F., Shulman, L., Ballaban-Gil, K. et al. (2001).
Language regression in childhood. Pediatric Neurology, 24, 183-189.
Sparrow S, Balla D, Cichetti D. Vineland Adaptive Behaviour Scales. Circle Pines, Minnesota:
American Guidance Services; 1984.
Stata Statistical Software Release 9.0: Survey Data Manual College Station, TX: Stata
Taylor, B., Miller, E., Lingam, R., Andrews, N., Simmons, A., & Stowe, J. (2002). Measles,
mumps, and rubella vaccination and bowel problems or developmental regression in
children with autism: Population study. British Medical Journal, 324, 393-396.
Tuchman, R.F., & Rapin, I. (1997). Regression in pervasive developmental disorders: Seizures
and epileptiform electroencephalogram correlates. Pediatrics, 99, 560-566.
Valicenti-McDermott, M., McVicar, K., Rapin, I., Wershil, B.K., Cohen, H., & Shinnar, S.
(2006). Frequency of gastrointestinal symptoms in children with autistic spectrum disorders
and association with family history of autoimmune disease. Journal of Developmental
Behavior and Pediatrics, 27, S128-36.
Volkmar, F. R. & Cohen, D. J. (1989). Disintegrative disorder or late onset autism. Journal of
Child Psychology and Psychiatry, 30, 717-724.
Wechsler, D. Wechsler Intelligence Scale for Children (III-UK Edition). London: The
Regression in ASD 26
Psychological Corporation; 1992.
Weiss, L. A, Shen, Y., Korn, J. M., Arking, D. E., Miller, D. T., Fossdal, R., et al. (2008).
Association between microdeletion and microduplication at 16p11.2 and autism. New
England Journal of Medicine, eprint Jan 9 2008.
Werner, E. & Dawson, G. (2005). Validation of the phenomenon of autistic regression using
home videotapes. Archives of General Psychiatry, 62, 889-893.
Werner, E., Dawson, G., Munson, J., & Osterling, J. (2005). Variation in early developmental
course in autism and its relation with behavioral outcome at 3-4 years of age. Journal of
Autism and Developmental Disorders, 35, 337-350.
Wing, L., Leekam, S. R., Libby, S. J., Gould, J., & Larcombe, M. (2002). The Diagnostic
Interview for Social and Communication Disorders: background, inter-rater reliability and
clinical use. Journal of Child Psychology and Psychiatry, 43, 307-325.
World Health Organisation. Mental Disorders: A Glossary and Guide to their Classification in
Accordance with the 10th Revision of the International Classification of Diseases:
Research Diagnostic Criteria (ICD-10). Geneva: WHO; 1993.
Regression in ASD 27
Gillian Baird, Guy’s & St Thomas’ NHS Foundation Trust, London, UK; Tony Charman,
Behavioural & Brain Sciences Unit, UCL Institute of Child Health, London, UK; Andrew
Pickles, Biostatistics, Health Methodology Research Group, University of Manchester, UK; Susie
Chandler, Guy’s & St Thomas’ NHS Foundation Trust, London, UK; Tom Loucas, School of
Psychology and Clinical Language Sciences, University of Reading, UK; David Meldrum,
Chatswood Assessment Centre, Sydney, New South Wales, Australia; Iris Carcani-Rathwell,
Greenwich Neurodevelopmental CAMHS, Oxleas NHS Trust, London, UK; Devanitha Sirkana,
Cheyne Child Development Centre, Chelsea & Westminster Hospital, London, UK; Emily
Simonoff, Department of Child and Adolescent Psychiatry, Institute of Psychiatry, King’s
College London, UK.
Susie Chandler is now at Behavioural & Brain Sciences Unit, UCL Institute of Child Health,
Acknowledgments: The study was funded by the Wellcome Trust and the Department of Health.
We thank the expert group, Patrick Bolton, Antony Cox, Ann Gilchrist, Rebecca Landa, Ann Le
Couteur, Catherine Lord, Lennart Pedersen and Michael Rutter. Thanks also to Greg Pasco,
Samantha Ross, Vicky Slonims, Emma Rowley and Martha Turner for their help with
Correspondence concerning this article should be addressed to: Dr G Baird, Newcomen Centre,
Guy’s & St Thomas’ NHS Foundation Trust, St. Thomas’s Street, London SE1 9RT, UK. Tel:
0207 188 9662; email: firstname.lastname@example.org
Regression in ASD 28
Presence of Regression according to Diagnostic Category
No ASDBroad ASD Narrow Autism
Rate .97 .89 .61
95% CIs.91 – 1.00 .79 - .97.45 - .76
95% CIs0 - .02.01 - .06.02 - .17
95% CIs0 - .09.02 - .18 .18 - .45
Rates and confidence intervals (CIs) are weighted. Rates are given for proportion of each
regression classification within each diagnostic group
Regression in ASD 29
Outcome in IQ, Vineland Adaptive Behaviour Scale and ICD-10 ASD Symptom Severity
according to Regression and Diagnostic Classification
Mean Score (95% confidence Intervals)
No Regression Lower-Level
65.0 IQ Combined ASD70.360.6
(63.1 – 77.5)(41.9 – 79.3) (57.6 – 70.3)
Broad ASD72.8 72.264.1
(64.1- 81.6) (58.5 – 86.0) (55.5 – 72.8)
Narrow autism55.2 43.963.8
(50.6 – 61.8) (14.6 – 73.0) (54.7 – 73.0)
VinelandCombined ASD 46.737.542.3
(42.0 – 51.4)(25.9 – 49.2)(34.7 – 49.8)
(44.1 – 54.1)33.8 – 56.4(39.2 – 56.0)
(26.8 – 40.2)(13.8 – 39.3)31.0 – 44.4
Combined ASD a,b
(6.41 – 7.52)
(9.14 – 11.17)
(5.53 – 10.61)
5.69Broad ASD a,b,c
(5.86 – 6.76)(8.16 – 11.24)(3.19 – 8.19)
Narrow autism10.55 10.8110.12
(9.99 – 11.11)(10.46 – 11.16)(9.23 – 11.01)
F tests for significant differences within diagnostic category across regression groups, p< .05
a overall test between all three groups; b no regression vs. lower level regression; c lower-level
regression vs. definite language regression
Regression in ASD 30
Early Development according to Regression Classification
Diagnoses combinedMean Score, (95% Confidence Intervals)
No regression Lower level
Age of first wordsa,b,c
(22.0 – 30.2)
(17.5 – 75.8)
(13.6 – 19.0)
Age of phrase speech43.5
(37.2 – 49.7)
(30.0 - 73.2)
(39.4 – 70.0)
Early development problemsa,b
(8.6 – 12.1)
(4.8 – 16.5)
(4.8 – 8.3)
F tests for significant differences within diagnostic category across regression groups, p< .05
a overall test between all three groups; b definite language regression vs. no regression; c definite
language regression vs. lower-level regression
Regression in ASD 31
Rate of Febrile Convulsions and Epilepsy according to Regression Group
Rate (95% confidence intervals) and
(0 - .31)
(0 - .25)
(.01 - .07)
(0 - .71)
(0 - .75)
(0 - .07)
(05 - .21)
(.02 - .15)
Regression in ASD 32 Download full-text
Mean Number of Gastrointestinal Symptoms1
Mean number of symptoms (95% Confidence Intervals) Total
No regressionLower level
(321- .7890) N/A
(.27- .80)(0 - .61)
(0 - .341)
(.51 – 1.11)
1 Mean scores range from 0-4