Social cognitive skills training utilize cognitive behavioral strategies such as problem identification, affective education, performance feedback, and weekly homework activities to target impairments in nonverbal synchrony, emotional expression, and interpretation of intent. The NETT curriculum is manualized and anchored in CBI strategies, such as problem identification, affective education, performance feedback, and weekly homework activities. Parent education sessions run concurrently with child groups to help facilitate generalization.

The facilitated play therapy group is a manualized treatment designed to tailor play to the interests and abilities of group members. Therapists use general therapeutics strategies such as reflective functioning statements to foster communication with therapists as well as between peers. Standard educational practices for children with ASD such as visual supports, schedules, and short-directed statements are also used. The concurrent parent group is supportive in nature.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

Global Functioning will be assessed using the Clinical Global Impressions-Improvement (CGI-I) Scale. A study physician followed standard CGI protocols to evaluate global improvement at each time point based on all available sources of information (e.g., caregiver interviews, behavior rating forms). The CGI-I employs a 7-point scale with the lowest score, 1, indicating the best outcome and the highest score, 7, indicating the worst outcome. Participants receiving scores of 1 (very much improved) or 2 (much improved) were considered responders. Participants receiving scores of 5 (minimally worse), 6 (much worse), or 7 (very much worse) were considered non-responders; note, no participant received a score worse than 5 which was used as the cut-off for non-responders. Participants receiving scores of 3 (minimally improved) or 4 (no change) were considered to show no significant change and omitted from the analysis.

Global Functioning will be assessed using the Clinical Global Impressions-Improvement (CGI-I) Scale. A study physician followed standard CGI protocols to evaluate global improvement at each time point based on all available sources of information (e.g., caregiver interviews, behavior rating forms). The CGI-I employs a 7-point scale with the lowest score, 1, indicating the best outcome and the highest score, 7, indicating the worst outcome. Participants receiving scores of 1 (very much improved) or 2 (much improved) were considered responders. Participants receiving scores of 5 (minimally worse), 6 (much worse), or 7 (very much worse) were considered non-responders; note, no participant received a score worse than 5 which was used as the cut-off for non-responders . Participants receiving scores of 3 (minimally improved) or 4 (no change) were considered to show no significant change and omitted from the analysis.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

The active comparison condition is a facilitated play therapy group.

Facilitated Play Therapy: The facilitated play therapy group is a manualized treatment designed to tailor play to the interests and abilities of group members. Therapists use general therapeutics strategies such as reflective functioning statements to foster communication with therapists as well as between peers. Standard educational practices for children with ASD such as visual supports, schedules, and short-directed statements are also used. The concurrent parent group is supportive in nature.

Total of all reporting groups

The Children's Communication Checklist-2 (CCC-2) Nonverbal subscale contains 7 items scored on a 4-point scale (0/less than once a week to 3/every day) for a total range of 0-21; higher total scores reflect greater impairment. Items estimate children's nonverbal communication skills (e.g., eye contact, social distance recognizing other's emotions) and are dispersed in a 70-item parent questionnaire intended for children 4-16-years-old.

The Children's Communication Checklist-2 (CCC-2) Social Relations subscale contains 7 items scored on a 4-point scale (0/less than once a week to 3/every day) for a total range of 0-21; higher total scores reflect greater impairment. Items estimate children's social competence (e.g., comfort around others, sensitivity to and interest in others) and are dispersed in a 70-item parent questionnaire intended for children 4-16-years-old.

The Griffith Empathy Measure (GEM) is a brief parent report survey adapted from an adult empathy measure intended to capture cognitive (i.e., identifying another's situation and taking their perspective) and affective (i.e., responding according to the other's perspective rather than one's own) empathy skills in pre- and through adolescence. Total scores range from -92 to 92, the sum of 23 items scored on a 9-point scale (-4/strongly disagree to 4/strongly agree), with lower total scores reflecting greater impairment.

Reading the Mind in the Eyes Test (RMET) % correct out of items administered (maximum 28), participant task

Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2) % correct out of items administered (maximum 96), participant task

Wechsler Abbreviated Scale of Intelligence, 2nd ed (WASI-II) and Wechsler Intelligence Scale for Children, 4th ed (WISC-IV) are clinician administered intelligence tests. Both tests provide a full scale, non-verbal and verbal intelligence quotient (IQ). Verbal IQ is estimated using the Verbal Comprehension Index (VCI). The WASI-II Vocabulary and Similarities subtests form the VCI. The WISC-IV Vocabulary, Similarity, and Comprehension subtests form the VCI. VCI provides an estimated verbal IQ ranging from 40-160, mean of 100 (standard deviation of 15); higher scores indicate better performance.

Data was missing for one participant's VIQ

The Attention Deficit Hyperactivity Disorder (ADHD) combined subscale from the Child & Adolescent Symptom Inventory-5 (CASI-5) measures the frequency of ADHD symptoms in children ages 5-18-years-old. Parents rate the frequency of 18 behavioral items on a 173-item form from 0/never to 3/very often. Items scores are totaled, and raw total subscale scores are converted into ADHD symptom severity T-scores for youth (5-12 yo) by sex. T-scores have a normative mean of 50 (SD=10) with severity categorized as 50-59/low, 60-69/moderate, 70 and above/high indicating clinical impairment.

The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores convert to T-scores using age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment.

The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SBI is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SBI is a composite score based on Soorya et al. (2015) which identified measures that comprise this metric. Scores from the Children's Communication Checklist-2 (CCC-2) Social Relations and Nonverbal Communication subscales and the Griffith Empathy Measure (GEM) were standardized as z-scores using the sample means and standard deviations at baseline. CCC-2 subscale scores and reversed GEM total scores were used so higher z-scores reflect more impairment across measures. The SBI outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SBI z-score equals 0; higher individual scores indicate relatively more impairment in social behavior skills. Larger, negative slopes represent improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in SC is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. SC is a composite score, including the Reading the Mind in the Eyes Test (RMET) and the Diagnostic Analysis of Nonverbal Accuracy-2 (DANVA2), based on Soorya et al. (2015). RMET has 28 items rated correct/incorrect (total 0-28). DANVA2 contains 4 sets of 24 items rated correct/incorrect (total 0-96). Percent correct was calculated for each measure given the difference in denominators and to allow for administrative omissions. Percentages were standardized as z-scores using the sample means and standard deviations at baseline. The SC outcome was subsequently derived by averaging the z-scores. At baseline, the sample average SC z-score equals 0; higher individual scores reflect stronger skills on social cognitive tasks compared to lower scores. Larger, positive slopes indicate skill improvement.

All participants contributed to the intent to treat (ITT) model.

Global Functioning will be assessed using the Clinical Global Impressions-Improvement (CGI-I) Scale. A study physician followed standard CGI protocols to evaluate global improvement at each time point based on all available sources of information (e.g., caregiver interviews, behavior rating forms). The CGI-I employs a 7-point scale with the lowest score, 1, indicating the best outcome and the highest score, 7, indicating the worst outcome. Participants receiving scores of 1 (very much improved) or 2 (much improved) were considered responders. Participants receiving scores of 5 (minimally worse), 6 (much worse), or 7 (very much worse) were considered non-responders; note, no participant received a score worse than 5 which was used as the cut-off for non-responders. Participants receiving scores of 3 (minimally improved) or 4 (no change) were considered to show no significant change and omitted from the analysis.

If participants were missing CGI-I ratings at week 12 (Facilitated Play: n=2), the rating from the last available visit (i.e., week 6) was used. Participants with scores of 3 or 4, indicating no significant change, were not included in the responder analyses (ION-ASD: n=12; Facilitated Play: n=12). One participant (Facilitated Play) had no data to carry forward. Additionally, two participants dropped out of the study (ION-ASD: n=1; Facilitated Play: n=1) and were excluded from the analyses.

Global Functioning will be assessed using the Clinical Global Impressions-Improvement (CGI-I) Scale. A study physician followed standard CGI protocols to evaluate global improvement at each time point based on all available sources of information (e.g., caregiver interviews, behavior rating forms). The CGI-I employs a 7-point scale with the lowest score, 1, indicating the best outcome and the highest score, 7, indicating the worst outcome. Participants receiving scores of 1 (very much improved) or 2 (much improved) were considered responders. Participants receiving scores of 5 (minimally worse), 6 (much worse), or 7 (very much worse) were considered non-responders; note, no participant received a score worse than 5 which was used as the cut-off for non-responders. Participants receiving scores of 3 (minimally improved) or 4 (no change) were considered to show no significant change and omitted from the analysis.

If participants were missing CGI-I ratings at week 16 (ION-ASD: n=1; Facilitated Play: n=5), the rating from the last available visit (i.e., 5 from week 12 and 1 from week 6) was used. Participants with scores of 3 or 4, indicating no significant change, were not included in the responder analyses (ION-ASD: n=10; Facilitated Play: n=12). Additionally, from baseline to week 16, five participants dropped out of the study (ION-ASD: n=3; Facilitated Play: n=2) and were excluded from the analyses.

Global Functioning will be assessed using the Clinical Global Impressions-Improvement (CGI-I) Scale. A study physician followed standard CGI protocols to evaluate global improvement at each time point based on all available sources of information (e.g., caregiver interviews, behavior rating forms). The CGI-I employs a 7-point scale with the lowest score, 1, indicating the best outcome and the highest score, 7, indicating the worst outcome. Participants receiving scores of 1 (very much improved) or 2 (much improved) were considered responders. Participants receiving scores of 5 (minimally worse), 6 (much worse), or 7 (very much worse) were considered non-responders; note, no participant received a score worse than 5 which was used as the cut-off for non-responders . Participants receiving scores of 3 (minimally improved) or 4 (no change) were considered to show no significant change and omitted from the analysis.

If participants were missing CGI-I ratings at week 24 (ION-ASD: n=2), the rating from the last available visit (i.e., week 16) was used. Participants with scores of 3 or 4, indicating no significant change, were not included in the responder analyses (ION-ASD: n=6; Facilitated Play: n=11). Additionally, from baseline to week 24, seven participants dropped out of the study (ION-ASD: n=4; Facilitated Play: n=3) and were therefore excluded from the analyses.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Social Responsiveness Scale (SRS-2) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The SRS-2 is a caregiver reported measure of social behavior in the general population (3-18 yo) used to assess social functioning. The SRS-2 contains 65 items rated on a 4-point scale (1/not true to 4/almost always true). Scores are recoded 0-3 and totaled for a raw score. Raw scores are converted to T-scores based on age and sex. T-scores have a population mean of 50 with a standard deviation of 10. T-scores 60-65 indicate mild social impairment, 66-75 moderate deficits, and above 76 severe deficits strongly associated with autism. Higher SRS-2 T-scores indicate greater severity of social impairment. Larger, negative slopes reflect a greater reduction in severity.

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

All participants contributed to the intent to treat (ITT) model.

The change from baseline in the Caregiver Strain Questionnaire (CGSQ) is represented by the slope of each group. The primary outcome is the difference between groups in this rate of change or the group*time interaction. The CGSQ was used to assess quality of life. The CGSQ is a parent-rated questionnaire designed for parents of children and adolescents with emotional and behavioral disorders. It includes 21 items rated on a 5-point problem scale (1 = not at all to 5 = very much) rating subjective internal, subjective external, and objective strain. A global measure of strain can be calculated by averaging the scores together. Higher CGSQ scores indicate greater strain. Larger, negative slopes reflect a greater reduction in strain (i.e., improvement in score).

All participants contributed to the intent to treat (ITT) model.

The active comparison condition is a facilitated play therapy group.

Facilitated Play Therapy: The facilitated play therapy group is a manualized treatment designed to tailor play to the interests and abilities of group members. Therapists use general therapeutics strategies such as reflective functioning statements to foster communication with therapists as well as between peers. Standard educational practices for children with ASD such as visual supports, schedules, and short-directed statements are also used. The concurrent parent group is supportive in nature.