Although motor performance is not part of the diagnostic criteria for ASD, motor deficits are common (Fournier et al. 2010
), have been recognized as an associated feature since the earliest descriptions of the phenotype (Asperger 1944
; Kanner 1943
), and suggested as a cardinal ASD characteristic (Fournier et al. 2010
; Staples et al. 2012
; Hilton et al. 2012
). Motor signs, such as the attainment of motor milestones, may be more easily and reliably observed than core ASD symptoms. This has led researchers to study early motor delays as a potential pathway for early identification and intervention in ASD. Emerging research has documented differences between ASD and typically developing infants, with higher rates of parent reported concerns about motor development and later attainment of motor skills, including walking among children with ASD (West 2018
). Longitudinal data suggest these differences amplify with age (Landa and Garrett-Mayer 2006
), and that early motor difficulties may be a risk factor for impaired social communication and cognition, traits that are related to ASD (Leonard et al. 2014
). At present, early motor delays are considered to be a prodromal symptom of ASD (Bhat et al. 2012
; Harris 2017
), although with low specificity, as they are also associated with intellectual (Lemcke et al. 2013
) and other developmental disabilities (Zwaigenbaum et al. 2015
; Hatakenaka et al. 2016
Attainment of walking is reported to be later among children with ASD. Estimates vary from 1.1 to 2.5 month delay in mean AOW compared with samples of typically developing children (Ozonoff et al. 2008
), children at low risk for ASD (West et al. 2017
), and a national birth cohort (Lemcke et al. 2013
). Mean AOW has also been reported among different ASD subgroups (Matson et al. 2010
; Lemcke et al. 2013
; Ozonoff et al. 2008
), and for other non-ASD samples with atypical development (Ozonoff et al. 2008
; Bishop et al. 2016
), intellectual disability (ID) (Lemcke et al. 2013
) or language delay (West et al. 2017
). Notably, study design, assessment methods, sample sizes and clinical groups used for comparison varied between these studies, hampering comparability and generalization of results. A further methodological limitation has been the lack of normative data regarding AOW. However, this is available in Norway, where the use of both national and regional data (Storvold et al. 2013
), as well as comparisons with other countries (Onis 2006b
) are considered to increase the external validity and generalizability of the results.
Increased severity of ASD has been related to greater deficits in a multitude of areas. An as yet unanswered question is whether delays in AOW is associated with severity of ASD symptoms across diagnostic categories. Several studies have reported a pattern of slowed motor development across clinical groups (Matson et al. 2010
; Ozonoff et al. 2008
; Lemcke et al. 2013
), where children with ID or general developmental delays show the most delay, followed by ASD subtypes by decreasing severity. Motor skills have also been negatively correlated with symptom severity in autistic children (Hilton et al. 2012
) and found to predict autism severity scores in toddlers (MacDonald et al. 2014
) and school-age children with ASD (MacDonald et al. 2013
), suggesting that motor skills may be related to symptom severity and not just an ASD diagnosis. Because of the high comorbidity of ID in children with ASD, the possible influence of cognitive impairment on early motor delays has been discussed as a limitation of several previous studies. In their sample of 1185 individuals (ASD, n
= 903; non-ASD, n
= 282), Bishop et al. (2016
) found that lower IQ scores were associated with increased rates of late walking in both ASD and non-ASD groups, but children with low IQ without ASD were more likely to show delayed walking. Among individuals with ASD and nonverbal IQ (NVIQ) above 85, late walking (defined as at or after 16 months) occurred in 13%, against 31% in children with NVIQ less than 70. Female sex was found to heighten risk for delayed walking overall.
Although previous studies have provided useful information regarding AOW as a potential early marker for ASD, whether delays in AOW is associated with severity of ASD symptoms across diagnostic categories remains unclear. We investigated this relationship in a large clinical sample of Norwegian children assessed for suspected ASD by specialist health services, who varied in their severity of symptoms, cognitive abilities, and age at diagnosis. Specifically, we compared AOW, sex, age, NVIQ, and severity of autistic symptoms between children receiving an ASD diagnosis and children not meeting the criteria for diagnosis (non-ASD). Furthermore, we investigated the associations between AOW and symptom severity independent of ASD diagnosis. Finally, we investigated these questions separately for males and females. Available Norwegian population norms for AOW allowed for comparison with typically developing children.
In this study of AOW in a large sample of Norwegian children assessed for suspected ASD by specialist health services, we found that mean AOW was later among children with ASD compared to their typically developing peers, consistent with previous reports (Ozonoff et al. 2008
; Lemcke et al. 2013
; West et al. 2017
). AOW was associated with severity of core autistic symptoms, even after adjustment for potential confounders. Whereas AOW was significantly later in males with ASD compared with non-ASD diagnosis, females with autistic symptoms seem to have a liability toward later AOW, regardless of ASD diagnosis.
To our knowledge, this is the first study of AOW among children evaluated for suspected ASD, and directly aimed at investigating associations with symptom severity and possible sex differences. Applying a dimensional approach, we found that among children who displayed signs of ASD without meeting the criteria for diagnosis (non-ASD), AOW was significantly later compared with norms for typically developing children, but to a less extent than in children with ASD. Consistent with our results, Lane et al. (2012
) found that in a small sample (n = 30) of young children referred for possible ASD, those who received an ASD diagnosis tended to have greater delays in fine and gross motor domains, although not statistical significant, compared with children not diagnosed as ASD.
In the present study, symptom severity was higher in the ASD group compared with non-ASD, but with some overlap on all measures. Such overlap may be unavoidable, reflecting genetic relationships between ASD and other developmental disorders (Lichtenstein et al. 2010
; Lundstrom et al. 2011
). Our findings support the concept of autistic symptoms as quantitative traits transcending diagnostic categories (Frazier et al. 2015
). Further, a pattern emerged, where AOW seems to represent a continuum along which children with ASD show the most delay, followed by those with fewer autistic symptoms. This is in line with previous findings indicating that the more severe the autistic symptoms, the greater the likelihood of co-occurring conditions (Lundstrom et al. 2011
) and functional difficulties (Skuse et al. 2009
), including motor difficulties (Matson et al. 2010
; Green et al. 2009
; Hilton et al. 2012
; MacDonald et al. 2013
). Regarding AOW, a similar pattern of observed delay has been reported in retrospective (Ozonoff et al. 2008
) as well as prospective (Lemcke et al. 2013
) studies. The latter, a Danish national birth cohort study, reported increasing delay in AOW across different conditions, with the longest delay among children with ID and not ASD, followed by childhood autism and then any ASD diagnosis, including childhood autism. Extending previous studies, we included children with autistic symptoms without an ASD diagnosis. The lack of a control group was mitigated by using normative AOW data from the same population (Storvold et al. 2013
). While significant differences in mean AOW between groups and compared to norms was found, most children in both groups did attain walking within 16 months. The proportion of children characterized as “late walking” (i.e., AOW at or after 16 months) was smaller but considerable; 31% of the ASD and 25% of the non-ASD group. Our findings contrast somewhat with a recent study by Bishop et al. (2016
), in which 22% of 903 children with ASD were “late walking”, with mean AOW 14.00 (4.73) months.
Children with ASD are reported to have high frequencies of one or more co-occurring neurodevelopmental, psychiatric, and possibly causative medical diagnoses (Levy et al. 2010
; Lord et al. 2018
). Other diagnoses or symptoms may be present before all the symptoms of ASD are evident. In a prospective study of 30 children referred for early motor delays or abnormalities, including delayed walking (Hatakenaka et al. 2016
), the majority were found to have at least one NDD. Thirteen children were later diagnosed with ASD, of which 92% had two or more NDDs. Also in the present sample NDDs were common; 52% in the ASD and 42% in the non-ASD group had two or more NDDs. Moreno-De-Luca et al. (2013
) have argued that “neurodevelopmental disorders should be thought of as different patterns of symptoms or impairments of a common underlying neurodevelopmental continuum”. As such, the possibility that the observed common co-occurrence of NDDs in the present sample may represent a common etiology or underlying issues affecting also the motor domain, should be considered. In the present sample, 29% in the non-ASD and 11% in the ASD group were diagnosed with ‘motor disorder’. This category comprised ICD-10 diagnoses F82 (Specific developmental disorder of motor function) and F95 [Tic disorders, including Tourette’s disorder (F95.2)], see Table 1
, the majority of which were Tic disorders. The inclusion of motor disorder had a negligible effect on the main results.
Although it is possible to make a diagnosis of ASD before 24 months age in some cases, the majority of children with ASD in northern Europe are diagnosed by early school age (Lord et al. 2018
). In the present sample, mean age at ASD diagnosis was 9.3 years. Our results are consistent with Suren et al. (2012
) who used nationwide Norwegian register data and found that the proportions with ASD from 2008 to 2010 increased by age and was 0.7% in 11-year-olds. This suggests that ASD is often not diagnosed until late childhood or early adolescence in Norway. Later diagnoses are reported to occur in the context of co-occurring problems and other factors (e.g. female sex, more advanced language) that might have either exacerbated or masked the ASD (Lord et al. 2018
). The present study included children from both child habilitation services and child and adolescent mental health services evaluated for suspected ASD. This enabled the inclusion of individuals with a broad range of autistic symptoms and cognitive abilities. We consider this to strengthen the representativity of our results for the broader population of individuals assessed for suspected ASD in the health care system.
Taken together, the relatively high number of females, individuals with ASD subtypes without language delay (36% had Asperger syndrome) and the high proportion with co-occurring NDDs may have contributed to the relatively late age at ASD diagnosis in our sample. In terms of cognitive functioning, individuals with ASD display a wide range of abilities, from severe ID to superior intelligence, with prevalence rates for ID in different studies between 15 and 65% (Lord et al. 2018
). In our sample, 15.1% of ASD and 9.8% of non-ASD individuals were diagnosed with ID, further indicating a more ‘high functioning’ sample. Applying a dimensional approach, we included children with a broad range of autistic symptoms despite having other co-occurring disorders. In our sample, 27 children with ASD and seven in the non-ASD group had known genetic conditions, some of which may have contributed to later AOW in both groups, and later AOW compared to other ASD samples with more strict exclusion criteria. Further, Norwegian children are on average
older at AOW, compared with other countries (Storvold et al. 2013
; Onis 2006b
Our finding of mean AOW at 14.7 months in the ASD group is later compared with some earlier reports (Lemcke et al. 2013
; Bishop et al. 2016
). The magnitude of delay, however—children with ASD walking on average
almost 2 months later compared with typically developing children—is comparable to previous studies (Lemcke et al. 2013
; Ozonoff et al. 2008
; West et al. 2017
). This highlights the need to assess AOW in relation to autistic symptoms. The strongest association between AOW and symptom severity was found for ADI-R, with AOW making a unique contribution in explaining ADI-R total score. This held after adjusting for potential confounders. The association between AOW and SCQ was lost following adjustment in available case analyses, but remained significant after adjustment in the MI sample, which is considered less biased and to strengthen our results. A weaker association between AOW and SCQ may be reasonable, however, given that SCQ is a short parent-report questionnaire allowing only yes/no answers, whereas ADI-R is a semi-structured interview requiring trained examiners, which may perform better in eliciting parental concerns and capturing current and historical ASD symptoms. Further, the SCQ is found to be more similar to the ADI-R total score in differentiating ASD from non-ASD in the older (8-10, > 11) than younger age groups (Corsello et al. 2007
). Contrary to our finding that AOW was associated with symptom severity, as measured by the ADI-R and SCQ, and previous reports of correlations between SRS and motor skills (Hilton et al. 2007
), we found no significant association between AOW and SRS. This may indicate that SRS captures other aspects of social impairment that are not as strongly associated with AOW, compared with measures of core autistic symptoms.
When assessing relationships between ASD symptoms and other behavioral or neurobiological variables, taking into account phenotypic characteristics, such as age, IQ or co-occurring difficulties is important. ASD symptom measures such as the SRS and ADI-R are reported to capture more than symptoms of ASD, with elevating scores potentially reflecting impairments in dimensions other than the core characteristics of ASD (Havdahl et al. 2016
). The possibility that early motor delays are more general signs of compromised neurocognitive development, rather than specific to ASD, has also been discussed (Bolton et al. 2012
; Ozonoff et al. 2008
). Of the covariates included in the regression model in the present sample, NVIQ was making the strongest contribution to attenuating the relation between AOW and severity of core ASD symptoms. Significant associations remained, however, as did the difference in mean AOW between the ASD and non-ASD groups after adjusting for NVIQ. Thus, in our sample AOW was related to ASD symptom severity, even after adjusting for NVIQ. In order to examine whether AOW predicts ASD symptom severity over and above general motor ability, results from broader measures of motor functioning would have been useful. Unfortunately, such a measure was not included in the present study.
Because of potential typical sex differences, it is important to compare how males and females with ASD differ from typically developing males and females (Lai et al. 2015
). The WHO Multicentre Growth Reference Study (MGRS) found no significant, consistent sex differences in motor milestone achievement ages among typically developing children (Onis 2006a
). However, “girls in the MGRS tended to achieve milestones at earlier ages than did boys” (p. 71). Contrary to this, but in line with previous reports (Bishop et al. 2016
; Arabameri and Sotoodeh 2015
), we found that females with autistic symptoms (regardless of ASD diagnosis) are more liable to delayed
walking compared with males. Findings from screening-negative infants later diagnosed with ASD (Oien et al. 2018
) have highlighted the discrepancy between categorical criteria for ASD and developmental signs of an emerging or subthreshold autism phenotype (Oien et al. 2018
). Specifically, girls had less advanced early gross motor skills compared with boys. Along with a recent report from a large population study that autistic social traits in females tend to increase towards adolescence (Mandy et al. 2018
), these results may indicate a different phenotype or emerging pattern of symptoms in females with ASD.
Strengths of our study include the sample size and inclusion of individuals with a broad range of autistic symptoms and cognitive abilities. In addition, we used validated instruments, and the nature of data collection allowed adjustment for covariates and potential confounding factors. Study limitations are the retrospective nature of some of the information collected in the study, varying measures of autistic symptoms, and missing data (se Appendix for further discussion). We used clinical diagnoses obtained from different clinics, which may have introduced variation. Misclassification (in both directions) is possible, but not very probable for ASD and the non-ASD disorders, and is unlikely to be related to AOW assessment. ASD diagnoses assigned by Norwegian specialist health services have previously shown high overall validity (Suren et al. 2012
). Further, the relatively high number of females in our sample may indicate that referral and ascertainment bias leading to under recognition of ASD in females (Lai et al. 2015
) was low. For some analyses regarding sex differences our sample may have been underpowered. Otherwise, we do not consider type I errors to be likely. Nevertheless, these findings should be replicated in independent samples. Finally, the lack of control group was overcome by using normative AOW data from the Norwegian study by Storvold et al. (2013
). In that study, information on AOW was collected by parent report when children were 18 months of age, whereas we used retrospective information on AOW collected at inclusion (age from 4 to 18 years), introducing the possibility of recall bias. The quality of information about developmental milestones from caregivers has been examined by Hus et al. (2011
), who found AOW to be one of the most reliable parent report measures (Hus et al. 2011
). Although the precision of information regarding AOW may have varied, it is unlikely to have systematically biased our results. Further, the pattern and magnitude of delay observed is in accordance with results from previous studies.
Children with ASD share common features with children with other developmental delays, which may contribute to difficulties of accurate diagnosis. Although delayed onset of walking is not unique to ASD, the present study supports previous reports that it occurs commonly in ASD, and further demonstrate associations with severity of symptoms in other diagnostic criterion domains that characterize ASD. Recognizing that autistic symptoms may be difficult to interpret at an early age, assessing early motor delays and specifically AOW may have the potential to improve earlier identification of some cases with ASD, and perhaps particularly in females. Considering the possibility of ASD in infants with motor delays may not only enhance the potential for earlier diagnosis, but also improve the chance of targeting and addressing these delays in treatment programs and facilitate better prognostic outcomes.