Background
Autism spectrum disorder (ASD) is defined by the presence of characteristic behaviors present in early childhood [
1]. While it is now recognized that ASD is typically a life-long condition, there exist only a handful of systematic studies of middle-aged and older adults [
2‐
4]. According to the most recent estimate from the Centers for Disease Control (CDC), the prevalence of ASDs in US children is approximately 1.5 % [
5]. Given the well-documented aging of the population in western societies [
6], we and others have highlighted the need for increasing research efforts into all aspects of ASD in the aging population, to address what undoubtedly will be an important and increasing public health issue [
7].
To begin to examine ASD in older (i.e., relative to previous studies of ASD) individuals with ASD, several years ago we began a pilot, descriptive study in North Carolina (NC) of adults with ASD 50 years and older. This study focused broadly across a range of domains—including current autistic behaviors, medical/psychiatric/neurological problems, cognitive level, adaptive behavior, quality of life, family/community support, and available services, to inform ourselves about specific areas for more concentrated focus. As the project progressed, we observed what appeared to be high rates of parkinsonian signs in our ASD subjects (study I). We, therefore, sought to ascertain a second sample of adults with ASD in which to conduct a systematic assessment of parkinsonian signs (study II) and examine the hypothesis generated from study I that “older” individuals with ASD had elevated rates of parkinsonian signs. We herein report the results from the examination of samples in studies I and II.
While systematic studies of Parkinson’s disease and parkinsonism have not been conducted in adults with ASD, a number of indirect and direct links between autism and Parkinson’s disease have been suggested in the research literature. Hollander et al. [
8] suggested links between the two conditions based on overlapping phenomenology in the area of repetitive behaviors, with a common underlying involvement of the basal ganglia. Links between autism and the basal ganglia have been reported by a number of studies [
9‐
11]. Recently, motor system deficits have been suggested as having a core role in ASD, based on the observation of motor dysfunction in children, adolescents, and young adults [
12‐
17]. Motor deficits have been reported in infants prior to the typical time for diagnosis of ASD [
18‐
21], suggesting the possible primacy of the motor system in autism. Overlap has also been noted in genetics, where the Park2 gene, implicated in Parkinson’s disease, has also been found in association with ASD [
22]; Park2 copy number variant mutations have been identified in case reports of children with Asperger syndrome [
23]; and mutations have been identified in a gene-regulating dopamine metabolism and behavioral responses to dopaminergic drugs, also implicated in Parkinson’s disease [
24]. Establishing a precedent that genes linked to neurodevelopmental disorders also play a role in late-life neurodegenerative conditions, the link between the fragile X pre-mutation and parkinsonism, as expressed in the fragile X associated tremor and ataxia syndrome (FXTAS), is now well established [
25‐
27].
In the present report, we systematically examine the relationship between parkinsonian signs and ASD in middle-aged and older adults. In study I, we conduct an exploratory study in 19 adults with ASD, over 49 years of age. In study II, we follow up a hypothesis generated from study I, systematically assessing 37 adults with ASD, over 39 years of age, for evidence of parkinsonian signs.
Discussion
In this paper, high rates of parkinsonism are reported in a sample of 19 adults with ASD over 49 years of age, who were directly examined, characterized on a structured rating scale, and diagnosed according to the standardized diagnostic criteria. These findings were then replicated in an independent sample of 37 adults with ASD over age 39. After excluding subjects currently on neuroleptic medications, the frequency of occurrence of parkinsonism was 20 % in the combined sample and 17 and 25 %, in each sample, respectively. Four patients (7 % of the combined sample) were diagnosed with PD by community neurologists. In two patients, the diagnosis should be considered tentative—one was on levodopa and while they met the criteria for parkinsonism, they had no explicit diagnosis of PD in the medical record, whereas the second patient met the diagnostic criteria for parkinsonism and was diagnosed as having “lower limb parkinsonism.”
The findings from this study suggest high rates of parkinsonism among ASD individuals over 39 years of age. The rate of parkinsonism in the general population aged 65–70 has been estimated at 0.9 % [
36], as compared to 27 % in our combined total sample and 20 % when subjects on atypical neuroleptics were excluded. It is important to note that the de Rijk et al. [
36] study included individuals with drug-induced parkinsonism in their parkinsonism group, making the comparison even more conservative (as those subjects were excluded from the current study). Moreover, in the current study, parkinsonism was only diagnosed when it occurred together with bradykinesia (a required criterion), whereas in the de Rijk et al. [
36] study, parkinsonism was diagnosed if any two of the four cardinal symptoms for PD were present. Furthermore, the rates observed in the present study were considerably higher than the 0.1 % published rate for Parkinson’s disease in the general population 40 to 60 years of age [
37].
The estimates in the current study are conservative in that, in order to maximize accuracy in classifying affected individuals, we used a cutoff of ≥2 when rating the UPDRS and MDS-UPDRS and ratings of 1 were collapsed with ratings of “0” (i.e., rated as “absent”). We view this as a conservative approach that may have led to several false negative classifications (see
Appendixes 1 and 2 for clinical examples and further clarification). In addition, rates of parkinsonism were based on the entire sample of subjects in each study group, although a substantial portion of subjects in both groups were unable to be rated on signs that required active participation (e.g., finger tapping to rate bradykinesia and postural instability among non-verbal subjects).
The observation that the frequency of motor signs did not differ between those subjects currently taking neuroleptic medications versus those not on such medications, as well as the observation that parkinsonism was undiminished in several subjects on successively lowered doses of atypical neuroleptics, suggests the possibility that some cases of parkinsonism, observed in individuals on neuroleptics, may have been misclassified as being secondary to medications, when in fact they may have had a primary parkinsonism. In the
Appendixes 1 and 2, we describe motor signs in a subject whose motor abnormalities were not improved on decreasing and ultimately discontinued (for 4 months as of the time of this submission) doses of olanzapine. In conclusion, it is highly likely that we may have under-estimated the true frequency of both PD and parkinsonism in ASD, which may be higher than herewith reported.
The process of diagnosing Parkinson’s disease in ASD is complex for several reasons: (1) As we found in this study, many adults with ASD are or have been on neuroleptic medication. (2) ASD has its own motor idiosyncrasies, such as motor stereotypies, generic “clumsiness,” and dyspraxia, which complicate the neurological exam. (3) Non-verbal ASD subjects are difficult to assess with the UPDRS/MDS-UPDRS, given that some of them are unable to follow commands. (4) Given that many adults with ASD live in-group homes or other supported accommodations, with a high turnover of caregivers, information about whether parkinsonian signs are progressive, with persistent asymmetry or had a unilateral onset, cannot always be ascertained. This limitation makes it difficult to use step 3 criteria of the UK Parkinson’s disease Society Brain Bank Clinical Diagnostic Criteria for a definitive diagnosis of Parkinson’s disease. We provide clinical vignettes in the
Appendixes 1 and 2 that illustrate several of the complexities we encountered in our efforts to classify these individuals relative to the presence or absence of Parkinson’s disease.
An important limitation of this set of studies was the absence of local control groups for direct comparison of rates of parkinsonism and Parkinson’s disease. The lack of a systematic ascertainment scheme in study I precluded the identification of a suitable comparison group. While ascertainment in study II was done more systematically, lack of sufficient funding for that study precluded the addition of a comparably ascertained and assessed contrast group. Nevertheless, we believe that there is no evidence that either study was biased towards recruiting subjects with movement disorders, and the comparable rates observed in the two independent samples examined raises confidence in the validity of the findings. In addition, there was no evidence that potential confounding factors (e.g., low IQ or living in a long-term residential care facility) explained the rates of parkinsonism observed, as rates in the high and low IQ groups and between those living in residential care and others did not differ significantly.
The proportion of subjects with low IQ in both studies was considerably higher than current population estimates based on epidemiologically ascertained samples of school-aged children with autism [
5]. A likely reason for this is that the rates of low IQ in the current sample are a reflection of ascertainment characteristics in childhood autism samples of the 1960s and 1970s, when many of the subjects in this study were diagnosed. It is possible that many higher functioning children with autism in the 1960s and 1970s were either undetected or misdiagnosed and continue to be under-recognized as adults, resulting in the difficulties we encountered in finding adults with ASD over 50 in the North Carolina sample and the higher proportion of subjects with low IQ ascertained in both study groups. This possibility is supported by findings from Mandell et al. [
38], where 10 % of adults with autism in a state psychiatric hospital were found to be misdiagnosed with schizophrenia and other psychiatric conditions. Finally, accurate calculation of IQ scores in adults with ASD was an additional problem encountered in this study. IQ was estimated in some cases, from the Vineland adaptive behavior composite. Given that most of these participants live in-group homes with restrictive policies regarding activities (e.g. cooking, traveling independently), individuals with higher capacity and skills may have been inadvertently penalized with relatively lower scores.
While the findings from these two studies suggest a relationship between ASD and Parkinson’s disease, we can only speculate on the underlying explanation for this observed relationship. Several possibilities exist. First, given the relatively small size and the lack of a population-based ascertainment of both study groups, an unknown ascertainment bias must be considered as a possible explanation. Second, while current medication use did not appear to explain the rates of parkinsonism in either study, it is possible that long-term neuroleptic use and an underlying increased sensitivity to those medications may have had the effect of increasing the risk for parkinsonism later in life (even in those now off such medications). Thus, future studies should assess lifetime exposure to neuroleptic medication and examine the possibility that individuals with ASD, exposed to neuroleptics during their lifetime, may be at increased risk of developing Parkinson’s disease in mid- and later-life. A third possibility is that neuroleptic use may have “unmasked” a pre-clinical stage of Parkinson’s disease, perhaps lowering the threshold for dopamine loss in the development of Parkinson’s disease [
39]. This hypothesis warrants examination in longitudinal studies of younger populations, as well as in older populations where better historical information is available.
Of greatest interest is a fourth possibility, that ASD and Parkinson’s disease share a common underlying pathogenetic mechanism or mechanisms that have both early and later manifestations. This idea is reminiscent of the early effects of the expanded triplet repeat in the FMR1 gene in causing both fragile X syndrome in children and the later effects of the FMR1 pre-mutation in causing fragile X-associated tremor and ataxia in carrier grandfathers in their fifth and sixth decades of life [
25]. Parkinson’s disease may itself be a syndrome with multiple etiologies, some of which may be shared with ASD.
The findings from these studies, linking ASD and PD, have potential implications for understanding the neurobiology of both conditions. Links between the two have been made previously based on overlapping phenomenology and neurobiology [
8]. Motor stereotypies are a well-known feature of ASD in a subset of individuals with autism [
40], and an emerging literature suggests that a variety of other types of motor deficits (e.g., postural control, precision grip, and motor learning) are present at high rates in individuals with autism [
12‐
17,
41‐
43]). Making observations of motor impairment in siblings and twins concordant and discordant for autism, Hilton et al. [
44] have suggested that motor impairment constitutes a core characteristic of autism and that it is influenced by genetic factors. Also of interest is the relatively recent observation in high risk infant studies that gross motor deficits in the first year of life appear to precede the appearance of social deficits in the latter part of the first and second years, in high risk infants who go on to develop ASD [
19‐
21]. Motor deficits in infancy have also been observed in retrospective, videotape analyses of individuals with autism [
18]. Together these findings suggest a more primary and fundamental role of motor dysfunction in the pathogenesis of autism.
Common biological underpinnings between autism and Parkinson’s disease have also been suggested. Consistent with the behavioral observations of deficits in motor function in autism, brain abnormalities have been reported in individuals with autism and mouse models of autism, in regions and structures linked to the motor system. Structural imaging studies have noted abnormalities in the striatum [
9‐
11,
45] and resting state functional abnormalities in primary motor cortex [
46]. The cerebellum has also been implicated as playing a role in Parkinson’s disease pathophysiology [
47], is known to be interconnected to the basal ganglia [
48], and has been noted to be abnormal in individuals with autism on postmortem examination [
49,
50]; brain imaging [
51,
52], rapid precision grip [
53], and in mouse models of the disorder [
54,
55]. Genetic links between PD and autism have also been suggested by molecular studies identifying a mutation in Parkinson disease-associated, G-protein-coupled receptor in a number of unrelated autistic individuals [
24]. Finally, the finding of high rates of parkinsonism in pre-mutation carriers of the fragile X mutation who manifest signs of FXTAS [
25‐
27] suggests a possible common underlying biology in the subset of individuals with autism who go on to have Parkinson’s disease in later life. The molecular signaling pathways involved in fragile X (e.g., phosphoinositide 3-kinase or PI3-kinase and mammalian target of rapamycin or mTOR) lay at the confluence of molecular signaling pathways underlying several other genetically defined autistic syndromes (e.g., Rett Syndrome, PTEN, and tuberous sclerosis) and have been implicated in Parkinson’s disease [
56‐
58]. While specific biological mechanisms underlying the possible relationship between Parkinson’s disease and autism remain to be explored, identifying those individuals with parkinsonism in adulthood is a reasonable a strategy to explore in identifying a more etiologically homogenous subgroup with autism than those with idiopathic autism.
The possibility that individuals with autism are at increased risk for Parkinson’s disease as adults has important implications for detection and assessment, clinical practice, systems of care, training, and public policy. While research on ASD is now underway in young adult populations, the striking absence of research on autism in “older” adults has been noted in the literature [
7]. This dearth of knowledge about the problems faced by this population is likely to have a serious impact on future quality of care, with the increasing demands for care associated with the dramatic aging of western societies. The finding of high rates of Parkinson’s disease in older adults with autism is likely to be only one of many potential new findings to be discovered, as the research community turns its attention to this relatively neglected population. Such findings will have important implications for building capacity to care for such individuals, as it is likely that there will be unique aspects of their care that will need to be appreciated. If the findings of the present study are confirmed, there will be a need to understand the underlying pathogenesis of parkinsonism and Parkinson’s disease in ASD, to examine co-occurring behavioral and other features of the condition and to explore potential differences in treatment from the traditional approaches to Parkinson’s disease in individuals without ASD. Our finding of a high frequency of parkinsonism among middle-age and older autistic individuals on atypical neuroleptics emphasizes the urgent need to clarify whether there is a subset of individuals with ASD at a high risk for developing movement disorders, as well as exploring the possibility that neuroleptic medications may unmask or precipitate true Parkinson’s disease in older individuals with ASD. As a more general observation, the vast majority of the subjects we detected who had parkinsonism and were on neuroleptic medications, had motor deficits not appreciated by their care providers or noted in the medical records. Failure to recognize drug-induced parkinsonism in the elderly is well described [
59], and it is possible that recognition may be even lower in elderly with ASD. Co-morbid dementia may complicate a diagnosis of Parkinson’s disease, but none of our subjects had a history of functional decline or stepwise deterioration.
The lack of controls in both studies and ascertainment scheme in study I have been noted above as limitations of this work. There were also differences between studies in level of education, IQ, and living arrangements that may have had an unknown impact on the findings. Another limitation was that only 22 % of patients in study 2 were neuroleptic free, as compared to 63 % in study 1. These findings suggest that a relatively high percentage of adults with ASD will be on neuroleptic medication and/or will have a history of neuroleptic intake. This will be an important confounder for future studies aiming at ascertaining representative samples of adults with ASD off neuroleptics. The safety of withdrawal of neuroleptics among ASD individuals with well-controlled challenging behaviors remains unknown. A better option could be using DaTscan studies, a functional technique of imaging the dopamine transporter that helps to determine presynaptic dopamine neuronal degeneration [
60]. DaTscan studies may help distinguishing neurodegenerative from drug-induced parkinsonism, although false-positive results in drug-induced parkinsonism has been reported [
61]. Other limitations included the lack of establishment of inter-rater reliability within and across studies, although good reliability of the UPDRS and MDS-UPDRS has been demonstrated [
33,
62], limited information on life-long medication history and, for a number of subjects, limited information about early childhood behavior (to provide more certainty of diagnosis). Our relatively small sample size may account for the lack of significant differences in frequency of parkinsonism between ASD individuals on vs. off neuroleptics, as well as the lack of significant association between parkinsonism and clinical variables such as IQ. In addition, while it seems unlikely to be an explanation of the findings from the present studies, there is a possibility that some affected subjects had fragile X syndrome co-occurring with their ASD, as the presence of the fragile X mutation is known to account for a small portion of cases of ASD, and the FMR1 pre-mutation has been linked to risk for Parkinsonian motor deficits [
27,
63]. Future studies of Parkinson’s disease and parkinsonism in older adults with ASD should rule out the presence of the fragile X mutation in their study subjects.
Authors’ contributions
JP and SS were responsible for conceptualizing the study and all aspects of oversight of the project including design, analysis, and writing. MP participated in data collection, instrument selection, analysis and writing. LP and SG contributed to instrument selection and data collection. All authors approved the final manuscript.