1. Infants who screen positive for autism risk at 24 months already demonstrate significant neurodevelopmental impairments in motor, language, and social domains as early as 12 to 15 months of age.

2. Specific motor skills, particularly grasping and stationary stability, along with language development and adaptability, serve as key early indicators of later positive Modified Checklist for Autism in Toddlers (M-CHAT) screening results.

3. Factors such as a family history of autism, premature birth, and early childhood sleep disorders are strongly associated with a higher risk of neurodevelopmental delays.

4. Early objective assessment using standardized scales like Peabody Developmental Motor Scales-2 (PDMS-2) and Gesell Developmental Diagnosis Scale (GDDS) can help identify high-risk infants before traditional screening ages, facilitating timely and targeted interventions.

Autism spectrum disorder (ASD) is a broad neurodevelopmental condition characterized primarily by deficits in social communication, limited interests, and repetitive or stereotyped behaviors [1, 2]. Children diagnosed with ASD often have higher instances of co-occurring medical issues, including intellectual disability and attention deficit hyperactivity disorder [3]. In recent years, there has been a significant rise in the global prevalence of ASD. An epidemiological study conducted in the United States in 2021 indicated that the rate of ASD was 23 cases per 1000 among children who are 8 years old [2]. This surge in ASD prevalence has attracted considerable attention from healthcare and educational professionals, as well as parents and the wider public.

Early identification and prediction of ASD development is vital for improving the prognosis of these high-risk infants [1]. Several reports have developed models to predict the occurrence of ASD, with a focus on using the Modified Checklist for Autism in Toddlers (M-CHAT) as a screening tool for 24-month-old children. However, these techniques require comprehensive laboratory examinations or electroencephalography [4, 5]. Widespread screening of infants is difficult. Screening of ASD subjects is complicated due to their impaired areas of social and cognitive capabilities and levels of functioning. Notably, recent developments have suggested that a significant number of children with ASD display neurodevelopmental setbacks during their first or second years, including delays in language and social skills, along with motor development issues [6]. Most children with ASD who are under 5 years of age experience a combined global developmental delay, which may be evidenced by delays across two or more areas of development. What is more important is that for infants as young as 12–15 months, neurocognitive markers help predict the occurrence of ASD [7]. Therefore, for children with high-risk of ASD, early detection of these neurodevelopmental delays might help identify and predict the occurrence, and lead to early intervention. In addition, a number of other assessment tools were created to comprehensively represent the neurodevelopmental condition, including the Gesell Developmental Diagnosis Scale (GDDS), the Sensory Integration Schedule (SIS), and the Peabody Developmental Motor Scales (PDMS).

ASD is thought to arise from a mix of genetic and environmental influences. We selected family history (genetic), prematurity, and sleep disorders (environmental) based on: (a) strong epidemiological association (family history OR = 7.4 [8]; prematurity OR = 2.5 [9]; or sleep-disorder prevalence 50–80% in ASD [10]); (b) biological plausibility in neurodevelopment; (c) clinical feasibility for early screening. These factors represent core etiological pathways: genetic vulnerability (family history), perinatal compromise (prematurity), and postnatal neural dysfunction (sleep disorders). Consistent with this framework, current evidence has confirmed that various risk factors are strongly associated with heightened ASD risk. For example, an increasing body of evidence has suggested that prenatal experiences and the health of the mother during pregnancy can have enduring impacts on the neurodevelopment of children [11]. Parental rearing pattern [12] and nutritional status [13] have also been reported to be potential risk factors. However, the relationship between these risk factors and the neurodevelopmental delays is not quite clear, especially in children at high risk for ASD. Therefore, we hypothesized that the early neurological developmental patterns at 12-months to 15-months of age, in combination with key genetic or environmental factors, were significantly correlated with the onset risk of ASD. We retrospectively analyzed the key early neurodevelopmental patterns of infants (between 12-month- and 15-month-old) with high-risk for ASD at 24 months, matched with low-risk ASD infants born in the same month. The possible risk factors were also analyzed to reveal the possible contribution to these neurodevelopmental disorders and ASD progression. Our study may provide early clues of neurodevelopmental disorders for these children.

ASD is a neurodevelopmental disorder characterized primarily by deficits in social communication and the presence of restricted, repetitive behaviors, affecting approximately 2.3% of children in the United States. A rising incidence has also been observed in China [14]. With the increasing prevalence of ASD, early screening and intervention have become critically important. The clinical significance of this study lies in the early identification of neurodevelopmental impairments, particularly in language, motor skills, and behavior, in high-risk ASD infants. These early indicators could provide a foundation for timely intervention strategies.

The American Academy of Pediatrics advises that every child undergo screening at 18 and 24 months, with M-CHAT being among the most commonly used assessment tools. In the present study, we screened children with positive M-CHAT results and retrospectively analyzed the neurodevelopmental scores between 12 and 15 months of age. It is important to note that M-CHAT is a screening tool; children who tested positive were referred for diagnostic follow-up. Our results showed that infants with a positive M-CHAT screen often have higher ASD history and a higher incidence of child sleep disorders. Moreover, these ASD high-risk infants showed lower PDMS-2, GDDS, and SIS scores than did low-risk ones. Spearman correlation analysis showed that the key subscales were correlated with demographic, pregnant, perinatal, or lifestyle factors, as well as with M-CHAT screening status. It is interesting that the raw scores of three subscales from PDMS-2, three subscales from GDDS, and one subscale from SIS cooperatively correlated with M-CHAT positivity at 24 months. Collectively, these results offered early neurodevelopmental indicators for ASD high-risk infants; this may serve as ultra-early symptoms for timely screening and intervention. By evaluating neurodevelopmental patterns between 12 and 15 months, we would be able to identify potential ASD risks earlier, guiding individualized intervention strategies. This is critical for minimizing long-term social-adaptation challenges and improving the quality of life for children with ASD.

Previous research has suggested that both genetic and environmental influences play a role in the emergence of ASD. In a population-based study involving 22,156 individuals diagnosed with ASD, genetic factors were found to account for 81% of the variability in ASD traits, whereas environmental influences were linked to 14% to 22% of the ASD risk. Furthermore, genetic risk elements for ASD are seen to overlap with various other developmental and psychiatric conditions. Nevertheless, no genetic or environmental factors are completely exclusive to the development of ASD. Examining additional risk elements, a meta-analysis highlighted several maternal contributions associated with elevated ASD rates in offspring, including gestational hypertension [OR = 1.4, 95% CI (1.2–1.5)], pre-pregnancy or ante-natal obesity [RR = 1.3, 95% CI (1.2–1.4)], preeclampsia [RR = 1.3, 95% CI (1.2–1.5)], and maternal age of 35 years or older [RR = 1.3, 95% CI (1.2–1.5)] [23, 24, 25]. Other research has indicated that the use of certain medications during pregnancy may elevate ASD risk for children [26].

For maternal alcohol exposure—a factor correlated with impaired grasping skills in our PDMS-2 results (Table 3)—experimental studies have provided mechanistic insights: animal models have demonstrated that prenatal ethanol exposure disrupts cerebellar Purkinje cell development and synaptic plasticity, leading to persistent motor-coordination deficits [27]. Human cohort studies further confirmed a dose-dependent association between prenatal alcohol exposure and reduced fine motor proficiency in infants, potentially mediated by altered GABAergic signaling [28]. Our observed correlation between maternal alcohol consumption and PDMS-2 grasping scores aligned with those experimental findings, strengthening the biological plausibility of this environmental-risk pathway. Recent studies have used multi-center machine-learning approaches in order to predict ASD from maternal risk factors. Although these methods show promise, their clinical application remains limited. Based on the results of the present study, we recommend integrating traditional neurodevelopmental assessment tools, such as PDMS, GDDS, and SIS, into clinical practice to improve the accuracy of ASD screening [29].

Our investigation revealed that both demographic and pregnancy-related factors were present in both high-risk and low-risk infants diagnosed with ASD, indicating significant correlations with ASD risk of family history of ASD, rates of prematurity, child sleep disorders, and low serum 25(OH)D levels. A recent review indicated that predicting later autism in infants is feasible by concurrently monitoring early autism development across multiple analytical levels (genetic, brain, health, and behavior), with behavioral assessments emerging as the most practical and cost-effective approach [30]. ASD is typically recognized as a neurodevelopmental disorder that can be accurately diagnosed in children by 24 months. However, recent research has shown that developmental patterns observed between 12 and 15 months can provide early predictive markers for ASD. Our study demonstrated that neurodevelopmental assessments conducted during this critical window (12 to 15 months) are strongly associated with high-risk ASD infants, providing valuable insights for early screening and intervention.

Various public health systems have made efforts to detect ASD in very young children within the general population. Nevertheless, screening techniques have often lacked the necessary sensitivity, as they have failed to identify the majority of children with ASD in the general population whose parents had not previously recognized any delays [31]. Longitudinal studies following infants from 12–15 months, who are subsequently diagnosed with ASD, have revealed early developmental indicators of the disorder. Structural and functional changes in the brain have been noted in children with ASD, even prior to the emergence of autism-related behaviors, with several techniques being utilized, including MRI, electroencephalography, and near-infrared spectroscopy. Additionally, a variety of behavioral markers for autism can be detected during this critical window (12–15 months). According to Dawson et al. [30], these variations include attention, the development of prelinguistic communication skills, emotional expression, temperament, social integration, motor skills, play with toys, and restrictive and repetitive behaviors. Dong et al. [32] also reported that ASD children exhibited impaired developmental quotients. Moreover, previous studies reported that infant neurocognitive markers were associated with ASD in mid-childhood [7] or later ASD diagnosis [33], indicating that several neurodevelopmental disorders in infants may provide clues for the occurrence for ASD development. Another study also demonstrated that late preterm infants who had positive results on M-CHAT showed unbalanced abilities in GDDS scales [34]. Therefore, investigating the neurodevelopmental status for infants might provide clues for early ASD prediction. In the present study, we retrospectively collected the neurodevelopmental status for suspected M-CHAT positive infants with PDMS, GDDS, and SIS tools. We demonstrated that ASD high-risk infants exhibited impaired function in most of the subscales for PDMS, GDDS, and SIS, which showed a close correlation with M-CHAT results. We also revealed that some demographic, perinatal, or lifestyle-associated factors were correlated with the scores for these subscales. Therefore, impaired subscales for PDMS, GDDS, and SIS in the early period (12 months to 15 months of age) might provide clues for the prediction of ASD in the future.

Methodologically, we used age-standardized DQ for GDDS and T-scores for SIS to enable developmental comparisons; PDMS-2 raw scores were retained due to lack of validated norms for 12–15 month-olds in our cohort. Sleep disorders were rigorously defined using BISQ-R criteria, and lifestyle variables were operationally defined with precise cutoffs to minimize measurement bias.

To mitigate false-positive risks from multiple subscale comparisons, we implemented FDR correction which maintained significance for 10/12 subscales. The logistic regression model demonstrated excellent predictive performance (AUC = 0.92) for M-CHAT screening status, with robust internal validation (bootstrap AUC = 0.90). These rigorous statistical approaches strengthened confidence in our findings despite the multiplicity of outcomes.

Several methodological considerations merit discussion. (a) Although we matched controls by birth month and adjusted for exact assessment age, residual confounding may persist due to developmental trajectories within the 12–15-months window. (b) Seasonal variation could influence vitamin D levels [35], which were measured at 24 months but may not fully reflect status during earlier neurodevelopment. Although we adjusted for season in our analyses, unmeasured factors like sunlight-exposure duration could introduce residual confounding. (c) Although we controlled for prematurity and key demographic factors, other unmeasured covariates (e.g., exact nutritional intake, environmental toxins) might affect results.

Additionally, to strengthen the clinical applicability of our findings and provide a broader perspective on assessment strategies, future research could incorporate complementary tools designed to evaluate executive function (EF) in preschool populations. While our study utilized well-validated measures of general neurodevelopment (GDDS), motor skills (PDMS-2), and sensory integration (SIS), these may not fully capture the specific EF deficits—such as in working memory, cognitive flexibility, and inhibitory control—that are increasingly recognized as core features in ASD [36]. The Behavior Rating Inventory of Executive Function for Preschoolers (BRIEF-P) is one such parent-report questionnaire that has demonstrated utility in identifying nuanced EF differences in young children with neurodevelopmental disorders, including ASD [36]. However, its application must consider potential measurement limitations, such as floor and ceiling effects, which may obscure the true extent of deficits or strengths in very young or severely affected populations, as highlighted in a recent systematic review [36]. Integrating performance-based measures (like those used in our study) with parent-reported behavioral ratings (like the BRIEF-P) could offer a more comprehensive profile of a child’s strengths and weaknesses, ultimately guiding more targeted early intervention strategies.

The limitations of this study include: First, although our sample (n = 98) exceeded the minimum required for detecting large effects (n = 40/group), it remained underpowered for detecting small-to-medium effect sizes (d 0.5) and subgroup analyses. Second, the retrospective design posed risks of selection and information bias: (a) Recall bias: Parent-reported variables may have been exaggerated due to heightened concern for high-risk infants. (b) Misclassification bias: Inaccurate reporting of exposure variables may have led to bias toward null or distorted associations unpredictably. Despite employing validated instruments and structured diaries, retrospective reporting remained a limitation. Third, the children enrolled were screened with M-CHAT without definite ASD diagnosis. Fourth, our exclusion of neurological comorbidities (e.g., epilepsy) may have limited generalizability to the broader ASD population. Although this was methodologically necessary to isolate ASD-specific neurodevelopmental patterns, it warrants caution when applying findings to children with complex neurodevelopmental profiles. Fifth, this study lacked long-term follow-up data beyond 24 months (e.g., at 5 years of age) to confirm the stability of early neurodevelopmental predictions for ASD. To address these limitations, we are initiating a prospective cohort that will: (a) enroll 300 infants at 18 months with serial neurodevelopmental assessments; (b) incorporate gold-standard ADOS-2 diagnoses at 36 months; (c) collect longitudinal environmental-exposure data through validated diaries.

These results suggest that infants who screen positive on the M-CHAT at 24 months may show early neurodevelopmental impairments at 12–15 months, including motor, language, behavior, and sensory issues. These neurodevelopmental impairments were associated with several factors, including family history of ASD, prematurity, and sleep disorders. Notably, preterm birth and sleep disorders were more prevalent in the M-CHAT positive group. Early detection of these neurodevelopmental impairments can help identify the likelihood of a positive M-CHAT screen at 24 months, which is a crucial step in the screening pathway for ASD. It is important to note that a positive M-CHAT is a screening indicator, not a clinical diagnosis of ASD; children with positive screens require comprehensive diagnostic evaluation. The clinical value lies in potentially flagging children earlier in the screening process for closer monitoring and, if later diagnosed, earlier intervention. However, it is crucial to emphasize the preliminary nature of these findings, which are derived from a retrospective design. Therefore, these results must be validated in prospective longitudinal cohorts that include gold-standard diagnostic confirmation before any definitive clinical recommendations can be made.

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

HC designed the study, conducted the data analysis, and drafted the manuscript. YL handled data collection and performed the initial bioinformatics analysis. JL was responsible for acquiring and preprocessing the dataset and contributed to the statistical analysis. QL was involved in the design of the work and the substantial interpretation of data, provided critical revisions to the manuscript, and provided overall supervision for the study. JX conceptualized the research framework, guided the overall methodology, and finalized the manuscript revisions. XW handled data collection, guided the overall methodology, and finalized the manuscript revisions. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

Approval for this study was obtained from the Ethics Committee at Hebei Children’s Hospital affiliated to Hebei Medical University (No. 202136). All participants under the age of 16 were informed in detail about this research and gave written consent from their legal guardians. The study was carried out in accordance with the guidelines of the Declaration of Helsinki.

Not applicable.

The project was supported by the Hebei Provincial Health Commission [grant number 20220776].

The authors declare no conflict of interest.