Multi-observational Study for the Early Identification of Autism Spectrum Disorder (ASD) and Other Neurodevelopmental Disorders (NDD) in At-risk Populations

Background: The global prevalence of ASD has shown a marked increase in the last decades, with a median prevalence of 65/10,000. In Italy, a recent study of the National ASD Observatory estimated a prevalence rate of 1 every 77 children in the age range 7-9 (2019). Baby siblings of ASD children, Small for Gestational Age (SGA, weight below the 3rd percentile) and premature infants (born between 26 and 31 gestational weeks) are three populations sharing higher risk of ASD and NDDs than the general population. The NIDA Network, the largest multi-center and multi-disciplinary Network in Italy, aimed to guide surveillance programs for the early diagnosis and intervention of ASD and to improve the quality of life of children with ASD and their families. Moreover, NIDA research activities focused on the identification of behavioral/biological markers for early detection of ASD. Delays or unusual patterns in several developmental domains (motor, language, social, visual processing) are early warnings of NDDs in the first three years of life. Promising biomarkers (i.e. genetic, epigenetic, immune) have been developed and their role in the diagnosis of NDDs need to be validated. This proposal aims to empower experimental and clinical protocols using sophisticated and non-invasive technologies that can be applied to at-risk populations.

Description of activities The BABY@NET project requires a multidisciplinary approach satisfied by the partners backgrounds, scientific experiences and expertise. The sponsor, ISS, will take care of the Ethics Committee approval and coordinate the entire project making sure everyone adheres to the established protocol. In particular, the following activities will be performed by the five operating units: Recruitment of at-risk populations; Clinical surveillance of at-risk populations (at five time-points from 6 to 36 months) through the national NIDA protocol and newly store-and-forward telehealth measures; Experimental tasks in the first 12 months of age using cutting-edge techniques to detect early risk markers of ASD and other NDDs (motor assessment, infant crying, sensory processing and social communication development using EEG); Telehealth support activities for health professionals of NICUs and CPUs; Biological samples collection for piloting genetic/epigenetic, metabolomic and secretome analyses in NICUs; Improvement of web-based NIDA platform for data-entry of clinical, telehealth, and experimental data; Statistical Analyses; Dissemination of the results and scientific publications.

Aims: The first aim is to establish a standardized surveillance network for developing age-specific clinical/experimental protocols able to early identify infant/children at risk for ASD and other NDDs. The NIDA Network has provided a well-established clinical protocol for the surveillance of 6 to 36 months infants at-risk for ASD/NDDs in all Italian regions. The purpose of the BABY@NET project is: 1. to expand the NIDA protocol to the whole national territory; 2. to extend the research field by collecting behavioral, neurophysiological and biological material to further investigate and get new insights into the pathogenetic features and early clinical presentation of ASD and other NDDs in order to identify early specific biomarkers to be included in clinical surveillance programs. In detail, the study intends to investigate at specific time points in the first 12 months of at-risk infants' life, vocal, motor and social performances, EEG activity, genetic/epigenetic, metabolomic and secretome profiles, that have been found altered in ASD/NDDs children. These purposes are based on promising experimental data collected since 2010 in previous pilot showing that early motor, language, and social attentional trajectories of high-risk infants for ASD, who later received a diagnosis of NDDs, appeared different compared to typically developing infants. By applying novel automatic and semiautomatic technological tools, the NIDA research activities identified potential early behavioral precursors of altered development linked to the NDD diagnosis. Emerging evidence suggests that epigenetic processes may help promote inflammatory processes that influence the risk of chronic disease, including NDDs. Moreover, metabolic imbalances are strictly linked to inflammation. Therefore, elucidation of the specific epigenetic and metabolic pathways involved in the modulation of inflammation is of great interest in seeking a clear mechanistic understanding of the pathogenesis of ASD. A specific pilot study will be performed on preterms and SGA infants, by cord blood and meconium sampling at Fondazione IRCCS Policlinico Ca' Granda and Azienda Ospedaliera Universitaria Federico II, with the aim of identifying epigenetic, metabolomic, and secretome (inflammatory-related) markers linking early life exposures to neurodevelopmental outcome.

A secondary aim is to provide telehealth support to professionals working in NICUs and CPUs to enhance monitoring of at-risk infants and speed up diagnostic and intervention processes with the ultimate goal of improving the quality of life of children with ASD/NDDs and their families. The shortage of personnel and expertise for the collection and analysis of experimental data and specific tests for monitoring at-risk infants in NICUs follow-up and CPUs in the Italian territory leads to the urgent establishment of a multidisciplinary telehealth team. It will promote and support health professionals in the neurodevelopmental and behavioral assessments of at-risk populations by audio and video analyses and test scoring. The telehealth support will be provided through the IT platform specifically implemented in the BABY@NET project on the National ASD Observatory website (https://osservatorionazionaleautismo.iss.it/rete-di-coordinamento-per-diagnosi-ed-intervento-precoce).

The third aim is to detect ASD/NDDs risk factors for planning prediction and prevention strategies. Early vocal, motor, sensory processing and social communication parameters collected and analyzed in the first 12 months of infants' life, along with clinical/behavioral/biological variables collected by CPUs and NICUs from 6 to 36 months of age will be entered into the IT platform specifically implemented in the present BABY@NET project. ISS Co-PI, leading the Epidemiology, Mathematical Models and Biostatistics unit of the ISS, will implement high-performance computing and artificial intelligence to the multidimensional clinical and biological datasets collected in the digital platform with the aim of identifying predictive risk factors associated with the ASD/NDDs diagnoses. The identification of risk factors will be used for planning and promotion of prediction and prevention strategies during the Well-Child Care Visits, as well as to further specialize the training of professionals in graduate schools of child neuropsychiatry, neonatology, and pediatrics.

Methods Infant behavioral, neurophysiological and clinical indexes will be monitored using a longitudinal multi-observational protocol specifically designed to be implemented as part of routine assessment visits at CPUs/NICUs.

Considering differences in the prevalence of ASD/NDDs in those three at-risk populations, an estimated sample of n=62 siblings of ASD children, 86 SGA, 86 preterm and a comparable sample of low-risk (LR) infants (at term and appropriate for gestational age) will be recruited between birth and 12 months of age through the CPUs/NICUs of the Operative Units (UOs)involved in the project and the NIDA Network. Up to now, the NIDA Network has already recruited and monitored 664 siblings of ASD children, 459 preterm, 153 SGA, and 158 LR infants from birth to 36 months through a clinical/experimental protocol. Given the limited time available in the present project for the enrollment and evaluation, the investigators will collect specific behavioral/experimental data in the first 12 months of age that have been found promising ASD/NDD risk predictors. These data will be pooled together with data previously collected through the NIDA Network to enhance the positive predictive value of behavioral/clinical variables previously found associated with ASD/NDD risk. Infants will be subjected to a multi-observational protocol up to 36 months (mo) of age. Preterm and/or SGA infants will be tested at the corrected postmenstrual age. The clinical and experimental protocol includes:

1. Study-specific and structured parent interview collected at the time of recruitment regarding gestational and family history, childbirth, socio-demographic variables, and environmental factors.
2. Experimental data collection at birth, 3, 6, and 12 mo: a. auxological and head circumference measurements; b. developmental trajectories of spontaneous and intentional motor movements, vocal/language/communication performances; c. early audio-visual sensory processing and social skills measured by an innovative integrated electrophysiological + eye tracking approach at 12 mo.
3. Clinical/diagnostic assessment for investigating longitudinal cognitive, motor, language, and adaptive profiles and determining the presence/absence of ASD/NDD diagnosis using standardized tools/tests and structured parental interviews at 6, 12, 18, 24 and 36 mo.

Infants identified at risk or with a provisional/stable diagnosis will be immediately included in early individualized intervention programs.

A specific pilot study will be performed on a subsample of 30 LR and 30 preterm infants. Cord blood and meconium will be collected at birth to identify genetic/environmental/inflammatory-related markers linking early life exposures to the neurodevelopmental outcome.

During the first three months of the BABY@NET project, an integrated online environment (platform) dedicated to the health professionals working at the CPUs and NICUs will be developed. The platform will provide a wide set of services (audio and video-uploading, analysis of motor, vocal and social features, data-entry and automated test scoring, file sharing, one to one live chats) supporting professionals during the longitudinal behavioral and clinical assessment in an efficient, effective, and trustworthy way, at low cost and with high clinical value.

The sponsor will establish and coordinate the BABY@NET multidisciplinary telehealth team, involving child neuropsychiatrists, psychologists, speech- and motor-therapists, neonatologists and researchers working within the UOs involved in the project.

Throughout the entire duration of the BABY@NET surveillance, each CPU/NICU will have a specific web space on the IT platform. Within each CPU/NICU, the digital system will assign a personal alphanumeric code to each infant, and it will be possible to upload all audio-video recordings data (vocal, socio-communicative, motor), structured interviews/gold standard tests, cognitive assessment and raw data of the neurophysiological/biological variables included in the experimental protocol. The multidisciplinary telehealth team will provide spectrographic analysis of the infant's crying, analysis of spontaneous and intentional motor skills, consonant inventory, canonical syllables, nonvocal vocalizations, and perform test scoring.

Infants' data, collected through the standardized surveillance network protocol and online expert support provided by the telehealth team, will be merged with data collected since 2010 by the NIDA Network (including 158 LR infants, 664 ASD siblings, 459 preterm and 153 SGA infants) and used to track vocal, motor, and social developmental trajectories in each population. Thanks to the BABY@NET project, it will be possible to further outline the infants' developmental profile, including behavioral, neurophysiological and biological parameters.

In addition, a predictive model will be implemented to forecast the ASD/NDD risk based on the clinical, experimental (vocal, motor, sensory and social measures) and genetic/epigenetic, metabolomic and secretome parameters collected. Both classic statistical approaches and machine learning algorithms will be applied. The BABY@NET project, through the testing and validation of this predictive model, will allow the automatic and early identification of infants with scores indicative of ASD/NDD risk. In this way, infants will have the unique opportunity to access early, individualized, and intensive interventions.

Statistical analysis A statistician will perform analyses of variance (ANOVAs) on clinical infants' outcomes obtained at each timepoint with group (4 levels: siblings of children with ASD, preterm, SGA, and LR infants) as the main between-factor variable. Differences among groups (ASD siblings/preterm/SGA infants with or without a diagnosis of ASD/NDDs) will be analyzed using t-test or Mann-Whitney (Kruskall-Wallis) test - depending on data distribution - for continuous variables and chi square or Fisher's exact probability test for categorical variables. Additionally, models to track early developmental trajectories will be estimated in Mplus by latent class growth analysis with inter-individual variations in time of assessment and mixed-effects linear models with repeated measures to assess group differences in rates of longitudinal clinical changes. Cluster analysis will be used to detect infant subgroups with similar developmental patterns. Longitudinal trajectories of primary outcome variables will be modeled using generalized linear mixed models (GLMMs) with main effects of ASD/NDD outcomes at 36 months and time points along with their 2-way interactions. All available observations from each participant will be used in modeling via the GLMM. The GLMM will account for correlations between repeated measures within individuals, allowing for fixed and time-varying covariates and automatically handling missing data, thereby producing unbiased estimates if observations are missing at random. A model will be implemented to predict ASD/NDD outcomes based on both clinical and experimental measures (AIM 3). Both classic statistical approaches (e.g., logistic regressions and generalized linear models) and machine learning algorithms (e.g., Support Vector Machine, Random Forest, and Extreme Gradient Boosting) will be applied. Machine learning models will be trained on a subset of the sample and then tested in the remaining subset. Predictive performance of the models will be mainly measured using the receiving operating characteristic (ROC) curve analysis. The predictive performance of the different models will be compared with the DeLong method for comparison of area under curve. Motor, vocal and social features, EEG/ERP parameters and molecular variables collected in the first year of infants' life will be associated with ASD/NDD symptoms and related traits at 18, 24 and 36 months. Putative prognostic indexes of ASD/NDDs and their predictive power will be identified by multivariate logistic regression and ROC curve analyses. Sensitivity and specificity (for the detection of ASD/NDDs) and predictive accuracy (for the prediction of ASD/NDD outcome) of the identified prognostic indexes will be evaluated, comparing: (i) HR-ASD vs LR, (ii) HR-NDD vs LR, (iii) HR-no diagnosis vs LR; (iv) HR-ASD vs HR-NDD. Multivariable predictive model to predict ASD/NDDs will be obtained by carrying out logistic regression modeling. Odds ratios (ORs) and 95% confidence intervals (CI) will be obtained. A probabilistic score will be obtained by considering significant variables and its prognostic accuracy will be evaluated by ROC curve analysis. Sensitivity, specificity, positive predictive value, and negative predictive value will be calculated.