A multimodal approach can identify specific motor profiles in autism and attention‐deficit/hyperactivity disorder

It is still unclear whether and to what extent the motor difficulties are specific to autism. This study aimed to determine whether a multimodal assessment of motor skills could accurately discriminate autistic children from attention‐deficit/hyperactivity disorder (ADHD) and typically developing (TD) peers. Seventy‐five children, aged 7–13, equally divided into three groups, were assessed with the developmental coordination disorder questionnaire (DCDQ), the movement assessment battery for children 2 (MABC2), the sensorimotor subtests of NEPSY‐II, and the kinematic analysis of a reach‐to‐drop task. Principal component analysis (PCA) on DCDQ subscales revealed one factor—Caregiver Report—, whereas MABC2/NEPSY‐II scores identified three factors—namely, Object Interception and Balance, Motor Imitation, and Fine‐Motor Skills—. Lastly, PCA on kinematic variables identified four factors: PC1, loaded by the parameters of velocity and acceleration throughout the task, PC2 and PC3 involved the temporal parameters of the two submovements, and PC4 accounted for the wrist inclination at ball drop. When comparing autistic and TD children, Caregiver Report and Motor Imitation factors predicted membership with 87.2% of accuracy. In the model comparing ADHD and TD groups, Caregiver Report and Fine‐Motor Skills predicted membership with an accuracy of 73.5%. In the last model, the Object Interception and Balance factor differentiated autistic children from ADHD with an accuracy of 73.5%. In line with our previous findings, kinematics did not differentiate school‐aged autistic children from ADHD and TD peers. The present findings show that specific motor profiles in autism and ADHD can be isolated with a multimodal investigation of motor skills.


INTRODUCTION
Beyond the core characteristics of the condition, autistic individuals often experience delays and difficulties in the gross motor domain. A recent meta-analysis (Wang et al., 2022) indicated autism is associated with vast differences in motor skills, which have a remarkable effect size of 1.04. Furthermore, in a second parallel metaanalysis, Wang and colleagues also outlined a moderate but significant correlation between gross motor and social abilities, suggesting that motor challenges are tied to autism core features (Wang et al., 2022). With this respect, two recent reviews (Lidstone & Mostofsky, 2021;Wang et al., 2022) proposed that visual-motor integration and, more in general, perception-action coupling represent particularly critical mechanisms for autistic individuals, which may specifically contribute to socialskill development. These observations, combined with the high prevalence of motor atypicality reported in the condition (up to 86.9% of autistic children/adolescents for Bhat, 2020), have recently prompted an ongoing debate about the potential of adding a motor specifier to the autism definition in future revisions of diagnostic manuals (Belmonte, 2022;Bhat, 2022;Bishop et al., 2022;Crippa, 2022;Ketcheson et al., 2022;Licari et al., 2022). Whether and to what extent the motor difficulties are specific to autism remains a matter of contention in this discussion.
Previous research suggests that motor difficulties could be common to other neurodevelopmental conditions, but those observed in autism are more challenging and persistent (Bhat, 2023;Fournier et al., 2010). Most studies have used cross-condition designs to verify the specificity of these motor peculiarities, often including children with either developmental coordination disorder (DCD) or attention-deficit/hyperactivity disorder (ADHD) as clinical comparison groups. While DCD is, by definition, characterized by primary motor skill disturbances, ADHD's core features are difficulty maintaining attention and hyperkinesia (American Psychiatric Association, 2013). However, up to 60% of ADHD may also present fine and gross motor peculiarities (Goulardins et al., 2017). Earlier direct comparisons of autism and ADHD revealed that autistic children frequently achieved lower scores on standardized measures of motor abilities (Dewey et al., 2007;Pan et al., 2009), although Biscaldi et al. (2015) reported similar difficulties in both autistic and ADHD children compared to neurotypical peers on the Zurich Neuromotor Assessment. In contrast, some studies provided initial evidence for a specific motor profile in autism and ADHD. Biscaldi et al. (2015) indicated that autistic children might struggle with imitation. In contrast, they documented alterations in static balance concerning ADHD symptoms, independently of the group status (either autism or ADHD). Ament et al. (2015), using a logistic regression model to predict group membership based on the Movement Assessment Battery for Children 2 (MABC2), suggested that catching and static balance items could be particularly associated with autism. Intriguingly, these group dissimilarities reflect different motor learning patterns. Indeed, when learning to compensate for a perturbation, autistic children rely more on proprioceptive feedback relative to visual feedback, while this bias toward reliance on proprioception is not present in both typically developing (TD) and ADHD children (Izawa et al., 2012).
To date, no studies have compared motor skills in autism and ADHD through the concurrent use of different measures, such as standardized neuropsychological tests, caregiver reports, or by using kinematic motion capture. Different sources of information could be more effective in detecting subtle motor atypicality in autism, as recently demonstrated for sex-based differences in motor performance (Crippa et al., 2021). Furthermore, the modality of the assessments itself can moderate the magnitude of the observed motor difficulties, with standardized clinical assessment having more significant effects than kinetic/kinematic measures and parent reports (Wang et al., 2022), also depending on age (Crippa, 2022). Given these premises, the hypothesis examined in this study is whether a multimethod approach can identify more fine-grained motor profiles specific to each condition. To this end, we evaluated motor abilities by means of the Developmental Coordination Disorder Questionnaire (DCDQ), a motor screener filled by caregivers, using the MABC2, a widely used clinical motor test, and with a reach-to-drop task that we have previously used to successfully distinguish autistic from non-autistic toddlers (Crippa et al., 2015;Forti et al., 2011). Our sample included autistic children, ADHD children, and TD children aged 7-13 years. By using this design, we aimed at better substantiating evidence for motor challenges specificity among different groups of school-aged children. To understand whether motor measures and kinematic variables could correctly discriminate among participants, we used a supervised machine-learning method, that is, logistic regression, to directly contrast the three groups of participants. Based on previous literature, we expected that the task of motor imitation and visual feedback would best differentiate autistic children from TD and ADHD peers, respectively.

METHODS Participants
A total of 75 children aged 7-13 years participated in the study, including 25 autistic children, 25 ADHD children, and 25 TD children. Each group consisted of 21 boys and four girls. Inclusion criteria were as follows: aged between 7 and 13 years, having Italian as a first language, and having a full-scale intelligence quotient (FSIQ) greater than 80 as measured by the Wechsler Intelligence Scale for Children-IV (WISC-IV, Wechsler, 2012). Three children (two autistic and one ADHD) did not meet the above intellectual functioning criterion but were still included as having a Perceptual Reasoning Index on WISC-IV greater than 80. Exclusion criteria were using any stimulant or non-stimulant medication affecting the central nervous system, having a well-defined genetic disorder, vision or hearing problems, and suffering from chronic or acute medical illness.
Participants in the autism and ADHD group were recruited from in and outpatient populations of the Child Psychopathology Unit of Scientific Institute, IRCCS Eugenio Medea (Bosisio Parini, Italy). At admittance, autistic children were diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5) criteria by a medical doctor specialized in child neuropsychiatry with experience in autism. The diagnoses were confirmed using the Module 3 of the Autism Diagnostic Observation Schedule-Second Edition (ADOS-2; Lord et al., 2012). All autistic participants in the present sample had no ADHD comorbidity. ADHD children had also been diagnosed according to the DSM-5 criteria by a child neuropsychiatrist with experience in ADHD. A child psychologist who was blinded to neuropsychiatrist's diagnosis independently confirmed the diagnosis through direct clinical observation and administering the semi-structured Development and Well-Being Assessment (DAWBA; Goodman et al., 2000) interview with parents. According to the clinical assessment, 19 children met the criteria for the ADHD combined subtype, 5 fulfilled the criteria for the ADHD inattentive subtype, and one had the hyperactive-impulsive subtype. No ADHD children presented with co-occurring autism. Lastly, children in the TD group were recruited from the general population by local pediatricians or schools near the institute. They had no previous history of social/communicative disorder, other developmental concerns, or family history of any neurodevelopmental condition. Autistic features in ADHD and TD children were checked using the Social Communication Questionnaire-Lifetime (SCQ, Rutter et al., 2003).
The Ethics Committee of our institute approved the present study, "Comitato Etico IRCCS E. Medea-Sezione Scientifica Associazione La Nostra Famiglia" (Prot. N.33/18-CE). It was therefore performed following the ethical standards outlined in the 1964 Declaration of Helsinki and later amendments. All the parents or caregivers of children signed a written informed consent form, and all procedures were conducted according to good clinical practice.

Measures
Parents/caregivers filled out the Social Responsiveness Scales (SRS; Constantino & Gruber, 2005) to rate the degree of autistic traits across participants. In the autistic group, ADOS-2 total raw scores were converted into calibrated severity scores because the latter are less influenced by participants' characteristics and allow comparisons across individuals with different developmental levels (Gotham et al., 2009;Hus et al., 2014). Familial socioeconomic status (SES) was coded according to the Hollingshead scale for parental employment (Hollingshead, 1975).
The MABC2 is clinically used to describe the motor domain of children aged 3-16 years. The MABC2 includes eight subtests grouped into three areas investigating different components of motor performance: manual dexterity, ball skills (such as aiming and catching), and static and dynamic balance. The tasks are agespecific, and in the present study, participants completed tasks designed for age band two (7-10 years) or three (11-16 years). Raw scores per area were converted into the three component standard scores, recorded as dependent measures. The internal consistency and the testretest reliability are excellent at 0.90 and 0.97, respectively (Wuang et al., 2012).
NEPSY-II is a battery of neuropsychological tests for evaluating children between the ages of 3 and 16 in six cognitive domains (Korkman et al., 2007). The present study used the Italian version of NEPSY-II (Urgesi et al., 2011), which can be administered to children aged 5-16 years. In particular, the four subtests of the sensorimotor domain were administered to participants: Fingertip Tapping, Imitating Hand Positions, Manual Motor Sequences, and Visuomotor Precision. Raw scores for each subtest were converted into standard scores and recorded as dependent measures. The test-retest reliability of the sensorimotor domain ranges from 0.75 to 0.98 (Korkman et al., 2007).
The DCDQ is a 15-item parent questionnaire to evaluate the gross-and fine-motor coordination of children aged 5-15 years during everyday functional/play skills in their natural environment. The DCDQ yields three subscales' scores (control, fine motor/handwriting, and general coordination), which were recorded as dependent measures. High internal consistency (0.89) has been reported for the questionnaire (Wilson et al., 2009).
Lastly, three-dimensional (3D) kinematic data of a reach-to-drop movement were collected by a motion capture system (the SMART D from BTS Bioengineering ® -Garbagnate Milanese, Italy). The reach-to-drop task consists in reaching and grasping a rubber ball placed over support and dropping it in a small basket ( Figure 1; Crippa et al., 2015). The task was included here based on previous evidence from our group showing that the kinematic analysis of this simple upper-limb movement can reliably identify preschoolaged, low-functioning autistic children (Crippa et al., 2015;Forti et al., 2011).
Kinematic data were preprocessed as previously done (Crippa et al., 2015). The topography of this experimental task has been previously standardized (Crippa et al., 2015(Crippa et al., , 2021Forti et al., 2011). Specifically, the reach-to-drop movement was divided into two submovements: Submovement 1-movement necessary to reach and grasp the ball; Submovement 2-a movement to drag the ball to the target basket and drop it. For each submovement, total movement duration, the number of movement units (a movement unit is defined as an acceleration phase followed by a deceleration phase higher than 10 mm/s, starting from the moment at which the increase or decrease in cumulative velocity is over 20 mm/s; Von Hofsten, 1991), peak velocity, time of peak velocity from submovement onset, peak acceleration, time of peak acceleration, peak deceleration, and time of peak deceleration were recorded as dependent measures. Wrist inclination at ball drop was further measured as the dependent variable.

Statistical analysis
Statistical analyses were performed using SPSS Statistics Software (Version 21) and open software JAMOVI 1.1.9.0 (retrieved from https://www.jamovi.org). The alpha level was set to 0.05 for all data analyses. After checking for linear model assumptions, between-group differences for demographic variables (age, IQ, and SES) were analyzed under a univariate analysis of variance (ANOVA). Potential between-group differences in SRS were assessed by an analysis of covariance (ANCOVA, with age and IQ as covariates). Differences in SCQ between ADHD and TD children were analyzed using a one-way ANOVA.
Second, three separate principal components analyses (PCAs) with direct Oblimin rotation and Kaiser normalization were conducted on the three DCDQ subscales score, on the seven motor features obtained from MABC2 and NEPSY-II and the 17 kinematic features to reduce the number of dependent variables while retaining the variation present in the data set. The Oblimin rotation was selected because it is reasonable to assume that caregiver reports and motor features would be correlated, whereas kinematic features have been previously shown to be correlated (Crippa et al., 2015(Crippa et al., , 2021. The number of components was estimated based on the eigenvalue criterion from Kaiser (1960) and the screen test (Cattell, 1966). Data suitability for PCAs was assessed with the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett's test of sphericity (Tabachnick & Fidell, 2001). Measures with loadings on components at or above 0.40 were retained (Tabachnick & Fidell, 2001).
Third, a multivariate analysis of covariance (MANCOVA, with age and IQ as covariates) was performed to assess potential between-group differences in caregiver reports, motor, and kinematic components obtained from the PCAs. Afterward, to understand whether motor and kinematic variables could predict group membership, three logistic regression analyses (autistic vs. TD, ADHD vs. TD, and autistic vs. ADHD) were conducted. In those regression models, only the measures showing significant between-group differences at MANCOVA were entered as predictors of group status. This study is considered exploratory, so no correction was applied for the family-wise error rate; however, 95% confidence intervals (CIs) for p values were calculated using a bootstrapping methodology (based on 1000 bootstraps resamples), and the outcome effect was significant if intervals between the lower and upper limits of a 95% CI of p values did not contain zero.
The reach-to-drop task consisted in reaching and grasping a ball (2) placed over a support (1, a), and dropping it in a hole (3). The hole (1, b) was located inside a transparent box (21 cm high, 20 cm wide) and was large enough so as not to require fine movements. Four passive markers are placed on the basket under the goal area, 2 on the ball and 3 on the hand. Reprint with permission from Crippa et al. (2015).

RESULTS
Sample sociodemographic and clinical characteristics are depicted in Table 1. The three groups were perfectly matched by sex. No statistically significant differences were found in age and SES (both p > 0.05). In contrast, there was a significant group difference in IQ, with TD children scoring higher than autistic peers (p = 0.013). Autistic children also had significantly higher scores on SRS compared with TD (p < 0.001) and ADHD peers ( p = 0.003), which in turn presented higher scores than TD ones ( p < 0.001). Finally, ADHD children also showed higher scores on SCQ than TD participants ( p = 0.001).
Concerning the motor measures, DCDQ subscales scores were unavailable for two participants (one autistic and one TD), while NEPSY-II Manual Motor Sequences scores were missing for one autistic child. DCDQ subscale scores were suitable for PCA (KMO = 0.686; Bartlett's test, p < 0.001). Results from PCA revealed one PC, labeled as Caregiver Reports, accounting for the 81.87% of the total variance. Balance PC were loaded by MABC2, aiming and catching, and static and dynamic balance scores. In Table 3, the results of the PCA and saturation of PCs show how individual motor variables contributed to each PC.
Kinematic data from the present sample were also suitable for PCA (KMO = 0.795; Bartlett's test, p < 0.001). This second PCA identified four PCs, explaining 79.74% of the total variance. Table 4 describes the results of the PCA on kinematic variables. In line with Crippa et al. (2021), PC1 was loaded by the parameters of velocity and acceleration of both submovements of the experimental task, while PC2 and PC3 involved the temporal parameters of Submovement 2 and Submovement 1, respectively. Finally, PC4 accounted for the wrist inclination at the end of Submovement 2. Standardization of the measures normalized the variance, and kurtosis and skewness values were lower than 1. Thus, scores on PCs from the two PCAs (on motor and kinematic measures, respectively) represented z scores.
A MANCOVA model was conducted to assess between-group differences PCs identified by the three separate PCAs on, caregiver reports, motor, and kinematic variables, controlling for age and IQ. The results of MANCOVA are presented in Table 5.  The results showed a significant group difference for Caregiver Report PC, with TD participants having higher scores compared with both autistic and ADHD peers (both p < 0.001). Significant between-group differences were also found for Motor Imitation PC, with TD children having higher scores than autistic ones ( p = 0.010), for Fine-Motor Skills PC, where TD children reached higher scores than ADHD peers ( p = 0.001), and for Object Interception and Balance PC with the autistic group presenting lower scores than TD and ADHD groups ( p < 0.001 and p = 0.021, respectively). Finally, statistically significant group differences were detected for kinematic PC3, where autistic children reported higher scores than ADHD children (p = 0.005).
Logistic regression was then run on three models (autistic vs. TD, ADHD vs. TD, and autistic vs. ADHD) to explore the ability of motor measures and kinematics to predict group membership correctly. The model comparing autistic and TD children explained between 60.9% (Cox & Snell R 2 ) and 81.2% (Nagelkerke R 2 ) of variance in-group status and correctly distinguished autistic from T A B L E 3 Motor variables' contributions to principal components.

Measures
Motor imitation Fine-motor skills Object interception and balance TD peers with 87.2% accuracy. Only two independent variables significantly contributed to the model-Caregiver Report PC and Motor Imitation PC-with odds ratios of 7.4 and 13.3, respectively. The model comparing ADHD and TD explained between 38.8% (Cox & Snell R 2 ) and 51.8% (Nagelkerke R 2 ) of variance and correctly discriminated the two groups of participants with an accuracy of 73.5%. Caregiver Report PC and Fine-Motor Skills PC were the variables significantly contributing, with the same odds ratio of 3.7. Lastly, the model contrasting autistic and ADHD groups explained between 29.2% (Cox & Snell R 2 ) and 39.0% (Nagelkerke R 2 ) of variance in-group status and correctly differentiated autistic participants from their ADHD peers with 73.5% accuracy. Object Interception and Balance PC predicted autistic group membership with an odds ratio of 2.9. An additional logistic regression was performed including the two MABC2 subscales (aiming/catching and static and dynamic balance, respectively) which loaded the PC to more finely parse this factor and identify the contribution of each subtest. This additional model comparing autistic and ADHD group accounted for between 21.9% (Cox & Snell R 2 ) and 29.2% (Nagelkerke R 2 ) of variance, with an in-group status accuracy of 70%. Only aiming and catching subscale significantly contributed to this model with an odds ratio of 2.8.

DISCUSSION
The present work sought to investigate the evidence for motor challenges specificity in autism and ADHD using-for the first time, to our knowledge-a multimodal approach. We evaluated the motor profile of three groups of children (autistic, ADHD, and TD) aged 7-13 through the concurrent use of standardized motor tests, caregiver reports, and kinematic analysis of a reachto-drop task. Our findings indicate that specific motor profiles can be isolated in autism and ADHD through a multimodal investigation of motor skills. We found that caregiver reports (DCDQ) and motor imitation skills were the best predictors of autism when comparing autistic and TD children, with an accuracy of 87.2%. DCDQ findings align with previous literature and have been formerly replicated in different samples (Bhat, 2020(Bhat, , 2021(Bhat, , 2022Crippa et al., 2021). Results on motor imitation are also consistent with previous studies (e.g., Biscaldi et al., 2015;Dewey et al., 2007); interestingly, all three studies provided converging evidence of imitation peculiarities in autism by using different modalities of testing. This latter finding, therefore, substantiates earlier observations that autistic individuals may struggle with tasks that require an efficient translation of visually observed actions into internal action representations (for a recent review on this topic, see Lidstone & Mostofsky, 2021). Regarding differences between ADHD and TD children, caregiver reports and fine motor skills differentiated the two groups of participants with a level of accuracy of 73.5%. As for the comparison between autistic and nonautistic participants, findings from DCDQ are in line with earlier works suggesting decreased motor skills in ADHD based on caregivers' judgment (Lee et al., 2021;Montes-Montes et al., 2021). Fine motor PC was loaded by scores on Visuomotor Precision of NEPSY-II and MABC2 Manual Dexterity, which included tasks such as pegboard, lacing blocks, nuts and bolts, and tracing tasks. Our findings differ from Biscaldi et al. (2015), which showed no difference in pegboard test between ADHD children and TD peers, but are in line with earlier studies consistently reporting fine motor difficulties in school-aged ADHD children, especially in tracing/ drawing tests (e.g., Goulardins et al., 2017;Kleeren et al., 2023). With this regard, previous research has indicated that difficulties in fine motor control could be related to core ADHD characteristics, such as difficulty in inhibiting prepotent responses (Kaiser et al., 2015). However, there is still no consensus about whether such an association exists on a behavioral level or in terms of a potential anatomical basis (Goulardins et al., 2017).  Last, our analysis showed that Object Interception and Balance PC was the only predictor of group status when directly comparing autistic and ADHD participants, with a classification accuracy of 73.5%. This PC was loaded by two MABC2 subscales-aiming and catching, and static and dynamic balance, namely-. A closer inspection of the contribution of each subscale revealed that between-group difference was entirely driven by MABC2 aiming and catching, which predicted the autistic group status with an accuracy of 70%. For balance, these findings are generally in line with earlier studies based on standardized developmental tests that reported inconsistent results. Ament et al. (2015) found that decreased static balance performance was explicitly associated with autistic status. In contrast, Biscaldi et al. (2015) reported autistic participants with more significant difficulties in dynamic balance when compared to ADHD ones. While our analytical approach hampers direct comparison with previous studies, we encourage future studies to include more comprehensive motor tasks to determine whether the peculiarities in balance control are specific to autism. In contrast, alterations in ball throwing/catching performance observed in the present study closely replicated those from prior works (Ament et al., 2015;Whyatt & Craig, 2013), offering additional evidence that difficulties in acting on moving objects could be the best motor feature to discriminate autism and ADHD. As for difficulties in imitation described above, throwing/catching issues could result from an overall alteration of visual-motor integration, as recently proposed by Lidstone and Mostofsky (2021). In addition to the comparison with ADHD, a recent meta-analysis (Wang et al., 2022) showed that object control abilities, such as throwing and catching, are most effective in distinguishing autistic individuals from non-autistic individuals. With this respect, it is important to note that, while we found a significant difference ( p < 0.001) between autistic and TD children in Object Interception and Balance PC, this difference fell marginally short of significance ( p = 0.070) in predicting autistic group status, given the stronger effect of caregiver reports (DCDQ) and motor imitation skills. Taken together, the present findings further suggest that perception-action coupling could represent a specifically impaired mechanism within the motor domain, even associated with autistic core features (Lidstone & Mostofsky, 2021;Wang et al., 2022). Alterations in object interception and imitation have been associated with peculiarities in different but partly overlapping neural networks, including the visual, rostroparietal, and premotor cortex, and the cerebellar regions (e.g., Crippa et al., 2016;Mahajan et al., 2016; for a thorough discussion of these networks, see also Lidstone & Mostofsky, 2021).
Interestingly, kinematic analysis of an upper-limb action task did not differentiate school-aged autistic children from ADHD and TD peers. This result is in line with recent findings from our group (Crippa et al., 2021) but diverges from previous works using the same task to distinguish autistic from non-autistic toddlers (Crippa et al., 2015;Forti et al., 2011). Although the recent metaanalysis of Wang et al. (2022) reported more significant effects for standardized clinical assessment than kinetic/ kinematic measures, this result may be due in part to the reach-to-drop task itself and to other methodological differences, rather than to the use of kinematic measurement methods per se. As we usually include participants in a broader age range (2-11 years), our reach-to-drop task do not require a specific movement end-point for a successful drop, because the ball could be released into the hole from any location over the box. Moreover, participants in the current work and in Crippa et al. (2021) have higher scores on standardized measures of IQ and older ages compared to previous ones. Thus, the discrepancy between our results could likely reflect a ceiling effect of the performance in this specific movement for older autistic children with lower support needs. Furthermore, the reach-to-drop task differed from other motor measures included here being a well-rehearsed, highly stereotyped movement, whereas the skills included in the Object Interception and Balance PC have a high degree of difficulty because of their dynamic nature. Previous research (Gepner & Mestre, 2002;Lidstone et al., 2020;Mosconi et al., 2015) has indeed shown that autistic individuals may particularly struggle to manage dynamic tasks that present increased visual-motor integration demands than static ones, suggesting a peculiar internal representations of dynamic targets for predictive motor control in autism, especially in children.
We feel that the present results could provide valuable contributions to the ongoing debate about the possibility of including motor characteristics to the autism definition (Belmonte, 2022;Bhat, 2022;Bishop et al., 2022;Crippa, 2022;Ketcheson et al., 2022;Licari et al., 2022). First, we offered new empirical evidence that specific motor profiles in autism can be isolated through a multimodal investigation of motor skills. Second, we critically discussed how the modality of motor assessment (standardized tasks, caregiver reports, or kinematics) and, more importantly, the nature of the movement target/ goal (static versus dynamic) could moderate the magnitude of the motor difficulties at different ages. Lastly, although preliminary, our findings also add important information for developing support programs specific to each condition. Previous works have formerly indicated that the motor skills of school-aged autistic children could benefit from the reduction of the visual information speed and the concurrent increase of proprioceptive feedback (e.g., see Lidstone & Mostofsky, 2021). Concerning ADHD, our findings related to fine-motor control difficulties highlight the potential of including training programs focused on executive functions, (such as inhibition and self-monitoring) into standard motorbased interventions (e.g., see Kleeren et al., 2023), given the potential role of these functions in tasks such as drawing and tracing. A strength of this study is the use of a multimodal approach to determine the specificity of motor difficulties in autism and ADHD. A further novelty is that we directly contrast all three groups of participants (autistic vs. TD, ADHD vs. TD, and autistic vs. ADHD) using a supervised machine-learning method (i.e., logistic regression). This work presented some limitations, too. One main limitation is related to the small sample sizes of participant groups. Second, we did not adjust the statistical testing for multiple comparisons because many dependent variables were intercorrelated. Another potential limitation is that both autistic and ADHD participants were clinically referred; this could restrict the generalizability of present findings to the heterogeneous presentation of these conditions. The generalizability of the motor profiles could also be limited by the lack of combined ADHD/ASD group in this study, as these are often comorbid. Our findings were additionally limited to drug-naïve, school-aged samples of children. Last, we did not include participants with high support needs.

CONCLUSIONS
These findings provide further evidence for specific motor profiles in autistic and ADHD children. Using a multimodal assessment, we found that difficulties in object control and balance were distinctly associated with autism, while challenges in fine-motor skills characterized ADHD. The present work can suggest developing specific support programs for children with neurodevelopmental conditions. We therefore substantiate the need to consider the specific strengths and challenges of individuals to better tailor the demands posed by the environment, also in accordance with the International Classification of Functioning, Disability and Health (ICF) framework (Bölte et al., 2021;Busti Ceccarelli et al., 2020).

ACKNOWLEDGMENTS
The authors are especially grateful to all the families of the children who took part in this study. The authors are also grateful to Dr. Valentina Riva for her useful comments on the first draft of the manuscript.

FUNDING INFORMATION
This work was supported by grants from the Italian Ministry of Health to A.C. (Ricerca Finalizzata GR-2011-02348929; Ricerca Corrente 2023, "Progetto Mosaico").

CONFLICT OF INTEREST STATEMENT
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.