Abstract
Autism spectrum disorder (ASD) is a neurodevelopmental disability characterized by deficits in social communication and the presence of restricted and repetitive behaviors and interests (American Psychiatric Association [APA], 2013). Diagnosis of ASD is uncommon before the age of 3 yr because the focus of the diagnostic criteria is on higher level cognitive and language functions (Barbaro & Dissanayake, 2009). Consequently, only a handful of well-developed early interventions are in use with children younger than age 3 yr (Boyd, Odom, Humphreys, & Sam, 2010). The study of early signs of ASD is a priority as a means of identifying methods of earlier diagnosis and therapeutic targets for early intervention (U.S. Department of Health and Human Services, 2010; Zwaigenbaum, 2010).
Current research examining early signs of ASD has focused on language, social, and cognitive domains (Barbaro & Dissanayake, 2009; Landa, Holman, & Garrett-Mayer, 2007; Sullivan et al., 2007). Few differences in these higher level skill domains have been noted before age 7 mo between children who go on to develop ASD and those who do not (Bryson et al., 2007; Landa & Garrett-Mayer, 2006). However, differences have been noted in early infancy in lower level skill domains, such as temperament at 6 mo (Zwaigenbaum et al., 2005), sensory and motor development at 6 mo (Dawson, Osterling, Meltzoff, & Kuhl, 2000), and eye gaze at 2 and 6 mo (Jones & Klin, 2013).
Because it is difficult to detect clinically meaningful variation in the higher level skill domains in early infancy (Guinchat et al., 2012), motor differences may offer an effective window into ASD risk in the first year of life. Moreover, emerging evidence has shown that motor difficulties are pervasive in ASD (Fournier, Hass, Naik, Lodha, & Cauraugh, 2010), parents report motor anomalies early in life (Guinchat et al., 2012), and motor difficulties potentially mask the severity of the core deficits of ASD (Bhat, Landa, & Galloway, 2011). The identification of specific motor features in early infancy that characterize children who develop or are at risk of developing ASD may guide efforts for earlier diagnosis.
Previous studies investigating early motor signs of ASD in early childhood have used retrospective video analysis to code behaviors in children who have a confirmed diagnosis of ASD (Esposito, Venuti, Maestro, & Muratori, 2009; Loh et al., 2007; Nickel, Thatcher, Keller, Wozniak, & Iverson, 2013; Ozonoff et al., 2008; Teitelbaum, Teitelbaum, Nye, Fryman, & Maurer, 1998). The findings from these studies have been mixed. Significant differences between infants later diagnosed with ASD and controls (including typically developing and developmentally delayed infants) have been reported in static and dynamic symmetry of postures in supine in babies younger than age 5 mo (Esposito et al., 2009), rate of acquisition of sitting and standing postures and initiation of posture changes (Nickel et al., 2013), acquisition of gross motor milestones in the first year of life (Teitelbaum et al., 1998), and spontaneous arm waving in 12-mo-olds (Loh et al., 2007).
In a recent study, presence of head lag (an indicator of postural control) in babies aged 5–10 mo was significantly associated with ASD diagnosis at 3 yr and was more frequently observed in high- than in low-risk infants (Flanagan, Landa, Bhat, & Bauman, 2012). Ozonoff and colleagues (2008), however, found that maturity of gross motor skills up to age 2 yr could not be used to discriminate between ASD and developmental delay. Saint-Georges and colleagues (2010), who reviewed the home video literature examining all early signs of ASD, concurred with this finding, concluding that motor differences could not be used to discriminate between infants with ASD and infants with developmental delay.
The current literature is limited by several methodological concerns. First, although specific guidelines for editing home video were proposed in the studies, home video poses several problems for the research process. For example, it is nearly impossible to gain a standardized view of each participant when using home video, thus limiting the comparability of observations across participants. Second, many studies did not use an established protocol of motor behavior assessment in infants. Third, a variety of ages was examined in the studies (from 12 wk to 18 mo), further limiting the comparability of results across studies. Another weakness of the current literature is that less attention has been paid to the trajectory of motor skill acquisition (i.e., changes in skill over time) in early infancy, and reports have instead focused on motor milestone achievement at limited time points in the first year of life. Initial reports have suggested variances in the trajectory of fine and gross motor development in children with ASD compared with typically developing children (Landa & Garrett-Mayer, 2006; Ozonoff et al., 2008).
The purpose of our study was to further examine differences in trajectories of motor skill development and specific motor delay in early infancy (ages 2–6 mo) in children with and without familial risk for ASD. A secondary purpose was to identify potential salient motor features in early infancy of children who go on to have a confirmed ASD diagnosis at age 2–3 yr. Specifically, our objective was to compare upper-extremity midline, grasp, and retrieval skills and gross motor development in the first 6 mo of life in infant siblings of children who had a diagnosis of ASD and compare their performance with same-aged infants with no increased risk for ASD. Each of these motor features has been identified in the literature as atypical in either older children with ASD or in young children with risk factors for ASD. No specific hypotheses were proposed for this exploratory study. Given the paucity of previous work in this area and the variance in findings to date, we seek to provide additional information to the field of occupational therapy so specific hypotheses can be posed in future studies.
Method
Research Design
This study involved the secondary analysis of video data collected as part of a larger prospective study examining early nonmotor signs of ASD conducted by authors Robson and Young. Approval for the study was gained from the ethics and institutional review boards at The Ohio State and Flinders Universities. Participants were followed from age 2 mo to age 3 or 4 yr. Analysis of available video footage for infants at ages 2, 4, and 6 mo who completed early developmental and autism screening assessments formed the basis for observations made in the current retrospective analysis.
Participants
Thirty-nine infants participated in this study. Participants included 25 infants who were considered at heightened genetic risk (i.e., heightened risk [HR]) for ASD (22 had at least one older sibling diagnosed with ASD and 3 had a first cousin with a diagnosis of ASD). All diagnoses were made using Diagnostic and Statistical Manual of Mental Disorders (4th ed. [DSM–IV]; APA, 1994) criteria, the standard at the time of assessment. The remaining 14 infants had no known family history of ASD (i.e., low risk [LR]). Infants were excluded if they had serious medical conditions, had low birth weight, or were born before 36 wk gestation.
HR infants were recruited through a local ASD society, an early intervention program, and a private therapy clinic. LR infants were recruited using flyers, electronic communications, and word of mouth. Attempts were made to recruit participants before birth to allow assessments to commence within the first weeks of life. However, because of the limited availability of HR participants, some were recruited at an older age. Table 1 displays participant characteristics and sample size. The Mullen Scales of Early Learning (MSEL; Mullen, 1995) was used to characterize the cognitive function of the participants. The MSEL is commonly used to establish early IQ and developmental age.
Participant Characteristics
Note. M = mean; MSELC = Mullen Scales of Early Learning composite score; SD = standard deviation.
Procedures
We assessed motor development by analyzing video footage of structured play sessions during which developmental and early ASD screening evaluations were administered for all available participants at ages 2, 4, and 6 mo. Infants were assessed in a variety of positions (e.g., sitting, supine, standing) as appropriate to the task and developmental age. Video footage averaged 26 min in length. Two protocols, the upper-extremity observation protocol and the gross motor protocol, were used to code the video data. A randomly selected sample of the videos was coded by two raters for each section to establish an interrater reliability rate >85%. Raters were trained research assistants who were blinded to HR and LR status. Follow-up assessments were done to determine diagnosis of ASD by a psychologist.
Upper-Extremity Observation Protocol.
Observations of grasp, midline, and retrieval behaviors were collected using an upper-extremity observation protocol developed for this study by authors Heathcock and Lane when the infant was in the supine position. As part of the MSEL, infants were provided with opportunities to explore the same three or four objects, which varied in size and shape. Objects were presented at eye level and midline. The number of opportunities for object exploration was recorded and used to calculate percentage. Raters reviewed videos multiple times at slow speeds to determine accurate start and end times for each behavior. Video was recorded at 30 Hz or 30 frames per second, and raters could observe the infant frame by frame. The same two independent raters, blinded to HR or LR status, collected data from the videos by recording the frequency and duration of three behaviors: (1) number of times a participant brought his or her hands together in midline and how long he or she kept them there (midline proportion), (2) number of object grasps per opportunity (object retrieval rate), and (3) how long the participant held an object when it was successfully grasped (grasp duration).
Gross Motor Protocol.
Gross motor development was assessed from the video footage using the Alberta Infant Motor Scale (AIMS; Piper & Darrah, 1994). The entire video was used to score the AIMS. The AIMS is a norm-referenced evaluation tool for at-risk infants ages 0–18 mo. Items follow common milestone skill development in supine, standing, prone, and sitting with observed and elicited items (Piper & Darrah, 1994). Points are given for completion of each item, and a higher score implies a more mature performance. The AIMS total score is calculated by summing the scores from 58 items. The infant’s score can then be converted to a percentile and compared with a normative sample.
Validity for the Bayley Scales of Infant Development (Bayley, 2006) and interrater reliability values for the AIMS are well established, with r values ranging from .86 to .99 (Jeng, Yau, Chen, & Hsiao, 2000; Silva, Maia, Lopes, & Cardoso, 2013). For some participants, a certain position (supine, prone, standing, or sitting) was not recorded or the view was obstructed, so the AIMS could not be scored. In these cases, the infant was dropped from analysis. Head lag during the pull-to-sit item was analyzed separately as an indicator of postural control. Head lag was specifically chosen because it has been reported as a specific motor skill that is delayed in infants with heightened ASD risk (Flanagan et al., 2012).
Diagnosis of Autism Spectrum Disorder.
Follow-up diagnostic assessments were conducted for all participants at age 24 mo and between ages 36 and 48 mo. Assessments involved the Modified Checklist for Autism in Toddlers, Revised With Follow-Up (Robins, Fein, & Barton, 2009) with the primary caregiver, followed by an assessment of ASD symptoms using DSM–IV criteria and standard tools by authors Robson and Young when screening tests or parent concerns warranted. Independent assessments were recommended for children who displayed atypical development, and children thought to have ASD were referred for a second opinion. Of the 25 HR infants who participated in the study, 5 received an ASD diagnosis, meaning 1 in 5 of our HR group had ASD, above the estimates for ASD in siblings (Constantino, Zhang, Frazier, Abbacchi, & Law, 2010).
Analysis
To examine developmental trajectories, we used repeated-measures analyses of change over time for each of our variables of interest—midline proportion, retrieval rate, grasp duration—and AIMS percentile score. Because of the young age of the participants in our study, analysis of the trajectory of head lag was not attempted. In typical development, head lag persists through 4 mo. Nonparametric (Friedman’s) analysis for variables without normal distribution and parametric analysis (repeated-measures analysis of variance [ANOVA]) were applied to assess changes in scores over 2-, 4- and 6-mo time points. Friedman’s analysis does not allow for between-groups analysis within a repeated-measures design, so it was not attempted for midline proportion, retrieval rate, and grasp duration. Instead, Mann–Whitney U tests were used for between-groups (HR vs. LR) differences on specific motor skills at individual time points after analysis and visual examination of trajectory.
Between-groups comparisons were included in the repeated-measures ANOVA, examining change over time on the AIMS percentile score. Bonferroni corrections in all analyses were used to account for multiple comparisons. Missing data points were accounted for using group means or described in detail for each variable.
Because only 5 study participants went on to receive an ASD diagnosis, we were unable to perform reliable inferential statistics to compare early infancy motor performance in these children with that of children who did not receive an ASD diagnosis. Instead, we report mean plots of each of these infants, a technique that has precedent in identifying features of ASD and ASD risk in a similar sample size (Sheinkopf, Iverson, Rinaldi, & Lester, 2012). Because only 1 participant with ASD was recruited to the study at 2 mo, we report data only from the 4-mo (n = 3 with ASD) and 6-mo (n = 5 with ASD) time points.
Results
Upper-Extremity Observation Protocol
Midline Proportion.
A significant difference was noted over time for midline proportion for the LR group (p = .008) but not the HR group after Bonferroni adjustments. Visual examination of the mean scores indicated that for LR babies, the proportion of time spent in midline play increased markedly between 2 and 4 mo and then decreased between 4 and 6 mo, returning to near 2-mo levels. This developmental change appears reduced in HR babies, particularly at 4 mo (Figure 1a). Moreover, HR infants in our sample spent almost no time in midline at 2 mo compared with LR infants, who averaged 6.7% of their time in midline.
(A) Percentage in midline play by risk category shows significant change over time for LR. (B) Retrieval rate by risk category shows significant change over time for LR. (C) Grasp duration by risk category shows no significant change over time for either group.
Retrieval Rate and Grasp Duration.
Conversely, a significant difference was noted over time for retrieval rate in the HR group (p < .001) but not the LR group after Bonferroni adjustments. Retrieval rate increased between 2 and 6 mo for both the HR and LR groups (Figure 1b); however, HR babies appeared to make the most change in retrieval rate, showing fewer retrieval attempts and successes than their LR counterparts at 2 mo but strongly increasing these attempts from 4 to 6 mo. The overall pattern of development from 2 to 6 mo in retrieval rate, however, appeared similar between groups. No significant differences between HR and LR groups were detected for grasp duration (Figure 1c).
Specific Motor Delays.
On the basis of the results of our initial analysis of motor development trajectories, it was hypothesized that specific motor delays would be observed in HR babies for midline proportion at 4 mo (where midline play appears attenuated compared with LR infants) and for retrieval rate at 2 mo. Mann–Whitney U tests were used to test these hypotheses, but neither difference was found to be significant (midline at 4 mo, p = .29; retrieval rate at 2 mo, p = .19).
Salient Features of Autism Spectrum Disorder.
At 4 mo, 2 out of 3 infants with ASD spent no time in midline and scored below the mean for the HR group for midline proportion. At 6 mo, 3 out of 5 infants with ASD spent no time in midline and scored below the mean of the HR group. Two infants with ASD spent more time in midline than the LR group at 6 mo. One infant with ASD followed a more typical trajectory, with more midline behavior at 4 mo and less midline behavior at 6 mo.
At 4 mo, 3 out of 3 infants with ASD scored above the mean for the HR group for retrieval rate. At 6 mo, 4 out of 5 infants with ASD scored below the mean and 1 infant above the mean of the HR group for retrieval rate.
At 4 mo, 2 out of 3 infants with ASD scored below the mean for the HR group for grasp duration and 1 infant scored 2 standard deviations above the mean of the HR group. At 6 mo, 4 out of 5 infants with ASD scored below the mean of the HR group for grasp duration.
Gross Motor Protocol
Motor Developmental Trajectory.
Of the 39 participants in this study, comprehensive longitudinal assessment of the AIMS at all three time points was available and scored for 15 participants (HR, n = 9; LR, n = 6). A 2 (group) × 3 (time) repeated-measures ANOVA revealed a significant main effect for time, F(2, 26) = 5.96, p = .007, and group, F (1, 13) = 17.16, p = .001, meaning both groups demonstrated an increase in AIMS score over time and the LR group outperformed the HR group. No interaction effect was observed.
Specific Motor Delays.
Each of the 39 participants had at least one video that allowed scoring of the entire AIMS at one time point. At 2 mo, there were 10 HR and 7 LR participants; at 4 mo, there were 19 HR and 13 LR participants; and at 6 mo, there were 23 HR and 14 LR participants. Means for the HR group were higher at each time point compared with the LR group (Figure 2). Independent t test with Bonferroni adjustment revealed a significant difference at 4 mo, t (30) = –3.18, p = .003.
Mean percentile AIMS score by risk group.
At 4 mo, 3 out of 3 infants with ASD scored below the mean for the HR group on the AIMS. At 6 mo, 4 out of 5 infants with ASD scored below the mean of the HR group and 1 out of 5 infants scored above the mean of the LR group. The individual infant with the high score at 6 mo was not tested at 4 mo.
Head Lag.
Presence or absence of head lag was coded at ages 4 and 6 mo because it is anticipated that more than 50% of a typical infant population will have head lag before age 3.25 mo (Piper & Darrah, 1994). At 4 mo, 82% of HR infants and 69% of LR infants had head lag. At 6 mo, 12% of HR infants and 33% of LR had head lag. Using a Fisher exact test because frequency counts are less than 5 in some cells, we observed no significant differences at either time point.
At 4 mo, none of the 3 infants with a later diagnosis of ASD had head lag. At 6 mo, none of the 4 infants with a later diagnosis of ASD had head lag (1 infant with ASD was not tested on this item). A summary of all results is presented in Table 2.
Summary of Results
Note. AIMS = Alberta Infant Motor Scale; ASD = autism spectrum disorder; HR = high-risk group; LR = low-risk group; SD = standard deviation.
Discussion
Our study aimed to explore trajectories of motor development and specific motor delays in young infants with and without risk factors for ASD. A secondary aim was to observe potential salient motor features of the 5 participants in the study who have a confirmed ASD diagnosis. A paucity of information is available on motor behaviors of children with ASD when they are infants because of the 2- to 3-yr time lag between infancy and an ASD diagnosis, making such data hard to collect. Using a systematic retrospective video analysis, we were able to capture some early behaviors of LR and HR infants and infants with ASD.
The results of our preliminary study suggest that there are observable differences in motor development trajectories between ages 2 and 6 mo in HR and LR infants on some of the variables measured. Moreover, we identified specific motor delays at 4 mo in gross motor skills and in upper-extremity skills that may require further study using a prospective study design with a larger sample size. Our study did not find evidence to support previous reports of head lag differences in young infants with risk factors for ASD (Flanagan et al., 2012). Our observations of salient features associated with ASD are preliminary but point to specific motor skills that warrant further study.
Delays in performing upper-extremity movements, such as bringing the hands to midline, may indicate the presence of a developmental disability such as ASD (Sherick, Greenman, & Legg, 1976). Our work joins that of others in suggesting that early upper-extremity skill differences may be a feature of infants at risk for ASD. In this study, infants with genetic risk of ASD (HR) had less change in midline behaviors from ages 2 to 6 mo. Moreover, some infants who had confirmed diagnoses of ASD in later childhood spent no time in midline during this time period.
Because early upper-extremity skills are foundational for more complex movements with the arms and hands such as pincer grasp, self-feeding, self-soothing, and gestural communication, it is possible that HR infants have different everyday experiences moving their arms, keeping their hands in midline, and exploring objects than do LR infants, contributing to continual poor performance in reaching and gesturing as they mature into toddlers. Midline play could be an important precursor to these other skills for several reasons. First, it is an antigravity movement, so it strengthens musculature of the shoulder, providing the proximal control for more precise distal movements with the hand. Second, when the hands are in midline, the location of the hands are in the immediate visual field of the infant, allowing for development of visual bimanual hand exploration and hand–eye coordination. Third, both hands are near the mouth, allowing for self-soothing, hand-to-mouth, and toy-to-mouth exploration. The results of this study suggest that the amount of time with the hands in midline might be an important behavior to consider for screening, assessment, and treatment of HR infants.
Poor motor skill development is a common characteristic of developmental delay and may be a feature of ASD (Mari, Castiello, Marks, Marraffa, & Prior, 2003; Rinehart & McGinley, 2010). The AIMS is a common evaluation tool for motor skills used in clinical settings and has good validity and predictability of disability for preterm infants, especially when infants are tested between ages 3 and 9 mo (van Haastert, de Vries, Helders, & Jongmans, 2006). In this study, all infants (in both groups) scored below the 50th percentile on the AIMS.
We think several reasons could explain these lower than initially anticipated scores. First, we applied the AIMS to retrospective videos of infants involved in a larger study. Because the experimenters in the videos were not testing the AIMS in real time, it is possible that some items on the AIMS were not tried, potentially deflating our scores. Second, the AIMS was standardized using infant norms from Canadian babies 20 yr ago (Piper & Darrah, 1994). The introduction of the Back to Sleep program and evidence of ethnic and cultural variation have suggested that norm references need to be established in each country. For example, some groups of typically developing infants from Taiwan and the Netherlands scored much lower on the AIMS than the norm-referenced group from Canada (Fleuren, Smit, Stijnen, & Hartman, 2007; Jeng et al., 2000). Therefore, it is possible that Canadian norms are not norms for Australian infants, again deflating our percentiles. Last, our results show clear group differences between HR and LR infants in the direction hypothesized. It is important that all infants in our study were videotaped in a similar manner and scored by blinded raters. There were no group differences regarding length of the videos or amount of time in any given position. It is likely that even if the normative scores do differ among countries, the ability of the AIMS to distinguish between typical and atypical groups is strong.
Poor postural control has also been identified as a possible early motor feature of ASD, and poor postural stability has been noted in older people with confirmed ASD (Minshew, Sung, Jones, & Furman, 2004). One specific skill, called head lag, tests infants’ ability to keep their ears in line with their shoulders when they are being pulled up by their hands to sit from a supine position. Young infants, and those with low tone, do not have the anticipatory control, tone, or flexor muscle strength to complete this task, and their head lags behind them while their neck arches.
One study, also using retrospective video analysis, found that head lag from ages 5 to 10 mo was more common in HR infants than in LR infants in the United States (Flanagan et al., 2012). Norms for this item on the AIMS suggest that 90% of infants do not show head lag at age 4.75 mo (Piper & Darrah, 1994). We observed no differences in head lag at 4 and 6 mo between HR and LR infants, in contrast to the results of Flanagan et al. (2012). However, the infants in our sample were younger. Although not intuitive, it is possible that head lag is a skill that is not initially delayed but later regresses in infants who go on to have ASD. The inclusion criteria for the HR group in Flanagan et al. and in this study were very similar (a sibling or first cousin with ASD), but theirs was a U.S. study and ours is Australian.
Infants with genetic risk for ASD are also at a higher risk for other developmental disabilities, including language and motor delays (Constantino et al., 2010). In our study, they showed poor development of motor skills on the AIMS compared with LR infants. Age 4 mo was a time point at which group differences may have peaked. At 4 mo, typically developing infants are beginning to be able to independently control their arms and legs to touch a toy and can control their head in an upright position for several minutes. From a screening and assessment point of view, these results suggest that age 4 mo might be when HR infants differ the most from infants with typical motor development.
Limitations
This retrospective study on early motor signs in ASD has several limitations. First, the sample size of infants who go on to have an ASD diagnosis was small, and their performance on most variables demonstrated variability. However, the number of infants who had an ASD diagnosis was higher than expected for a group of infant siblings, and our sample size for a group of siblings is common in the current literature. Second, our trajectory analysis did not allow inclusion of participants with missing data points, reducing sample size further. Third, it is likely that cognition differences exist between the groups given the expectation that the HR group is at genetic risk for ASD and language and developmental delay.
Implications for Occupational Therapy Practice
Developmental delays in early infancy of infants at genetic risk for ASD are apparent from this work and others. The clinical and research implications for occupational therapy practice include the following:
Close monitoring of early motor development in infants with risk factors for ASD and developmental delay is warranted.
Interventions targeting the development of gross motor skills and midline play may support the acquisition of later cognitive, language, and gestural skills in infants at high risk for ASD.
Refined assessment protocols of motor skills in the first year of life may improve our ability to isolate salient features of ASD and enhance early diagnostic efforts and could be a focus for future research.
Footnotes
Acknowledgments
We acknowledge all participants and families who participated in this project.