Results from an Early Behavioral Intervention Randomized Controlled Clinical Trial for Social Communication in Toddlers with Tuberous Sclerosis Complex
Article information
Abstract
Purpose
This study aimed to evaluate the effectiveness of a caregiver-mediated intervention in improving social communication skills among toddlers with tuberous sclerosis complex (TSC) and to assess the feasibility of behavioral interventions in medically complex populations.
Methods
In a randomized controlled trial with a waitlist-control design, 59 toddlers aged 12–36 months with TSC were randomly assigned to receive a 12-week caregiver-mediated intervention or to a waitlist group. The primary outcome was joint engagement, defined as the proportion of time the caregiver and child were mutually engaged during structured play. The trial was initially conducted in person and later transitioned to remote delivery to enhance accessibility and comply with pandemic restrictions.
Results
Both groups demonstrated significant improvements in joint engagement; however, there were no significant between-group differences at the end of treatment or follow-up after adjusting for baseline adaptive behavior. Toddlers with higher baseline adaptive behavior showed greater gains in joint engagement, regardless of group assignment. Hours of community-based interventions also increased across the study period in both groups, with a larger increase in the treatment group.
Conclusion
The findings suggest that early developmental surveillance for toddlers with genetic neurodevelopmental disorders may itself have therapeutic benefits. Gains were greatest among children with higher baseline adaptive behavior, highlighting the importance of developmental readiness in caregiver-mediated interventions. The study also highlights the need to account for concurrent community-based services when interpreting outcomes and demonstrates both the feasibility and challenges of implementing behavioral intervention trials in rare and medically complex populations.
Introduction
Early detection and intervention for autism and related neurodevelopmental disorders (NDDs) are known to improve developmental outcomes [1,2]. Much of the current understanding of early development in autism originates from prospective studies of infants with a family history of autism [3,4]. These studies have identified early alterations in brain network connectivity and nonverbal communication impairments during the first year of life, creating opportunities for pre-symptomatic risk stratification, ongoing surveillance, and intervention before a clinical diagnosis is made.
This approach has been extended to toddlers with tuberous sclerosis complex (TSC), a genetic neurocutaneous disorder caused by mutations in the TSC1 or TSC2 genes that lead to constitutive overactivation of the mammalian target of rapamycin (mTOR) pathway, resulting in benign hamartomas throughout the body and brain and disrupted neural connectivity. TSC represents a promising population for early autism detection and intervention for three key reasons: (1) it is often diagnosed prenatally or in infancy due to the presence of cardiac or brain hamartomas or early-onset epilepsy [5-7]; (2) up to 60% of children meet diagnostic criteria for autism, with approximately 50% exhibiting cognitive impairment, reflecting the high prevalence of neurobehavioral and neuropsychiatric sequelae collectively described under the umbrella of TSC-associated neuropsychiatric disorders (TAND) [8-10]; and (3) targeted disease-modifying therapies, such as mTOR inhibitors, have been tested in older children and are currently being studied in infants for epilepsy prevention [11]. Prospective studies have also identified early nonverbal cognitive and social communication delays in infants with TSC who later develop autism [12]. Our group previously demonstrated that by 9–12 months of age, social communication impairments—including slower disengagement of attention and limited eye contact, shared affect, social interest, and social referencing—can distinguish those later diagnosed with autism [13].
These findings, indicating both a high prevalence of autism and the presence of early behavioral differences preceding diagnosis, motivated the design of a clinical trial targeting early social communication skills. Despite the significant promise of behavioral interventions as complements to pharmacologic treatments, no studies had rigorously evaluated their potential effectiveness in TSC. Joint attention, symbolic play, engagement, and regulation (JASPER) is a caregiver-mediated naturalistic developmental behavioral intervention shown to improve social communication in toddlers with autism [14,15]. Prior to this trial, we conducted a small pilot study demonstrating the feasibility of JASPER in toddlers with TSC [16], which informed the design and implementation of the JASPER Early Intervention for TSC (JETS) trial for children aged 12 to 36 months (ClinicalTrials.gov NCT03422367) [17].
JETS represents the first randomized controlled trial of a behavioral intervention in TSC. The trial was initially designed as an in-person, weekly, parent-mediated program. However, enrollment proved challenging due to the burden of travel, particularly given the medical complexities of infants with TSC. Caregiver feedback emphasized time and cost barriers, especially for families managing epilepsy. In response, the trial was adapted to a hybrid delivery model, substantially improving accessibility, and the adaptation process was published to guide others developing behavioral interventions for rare disorders [17]. Shortly thereafter, the coronavirus disease 2019 (COVID-19) pandemic necessitated further modifications to accommodate virtual assessments and remote intervention delivery due to restrictions on in-person interactions.
The trial’s primary preregistered outcome was caregiver-child joint engagement, a core social communication skill fundamental to language development [18]. Joint engagement, defined as mutual involvement in a shared activity, serves as the basis for more complex social behaviors and language acquisition. Prior studies have shown that joint engagement is limited in toddlers who go on to develop autism [19,20]. In this study, joint engagement was assessed using structured parent-child play observations, during which the proportion of time the dyad spent in supported and coordinated joint engagement was recorded. Coders blinded to group assignment applied established systems developed by Kasari et al. [21,22], which have shown strong sensitivity to behavioral change. We hypothesized that toddlers in the JASPER group would demonstrate greater improvements in joint engagement than the waitlist group, with baseline clinical features (e.g., seizure severity, adaptive skills, developmental level) moderating outcomes. Overall, the JETS trial represents the first attempt to evaluate the impact of a behavioral intervention on developmental outcomes in toddlers with TSC, offering critical insights for future studies and informing the design and evaluation of interventions for this high-risk population.
Materials and Methods
1. Study design
1) Participant recruitment, eligibility, and enrollment
Children aged 12 to 39 months with a clinical diagnosis of TSC and their primary caregivers were recruited at one of two sites: the University of California, Los Angeles (UCLA) and Boston Children’s Hospital (BCH). Recruitment was conducted through referrals from TSC specialty clinics, the TSC Alliance, online social media postings, and institution-specific medical record queries. Exclusion criteria included a planned epilepsy surgery during the trial period or a nonverbal developmental level below 6 months of age as measured by the Mullen Scales of Early Learning [23]. The exclusion of children scheduled for epilepsy surgery was necessary to avoid confounding effects on developmental outcomes, as such surgery can lead to substantial and rapid changes in seizure burden, cognition, and behavior independent of any behavioral intervention. The surgery would also have prevented consistent participation during the immediate pre- and postoperative periods. The lower developmental threshold of 6 months was established to ensure participants possessed foundational skills necessary for meaningful engagement in a caregiver-mediated social communication intervention and to allow reliable assessment of early social communication outcomes. During recruitment, a brief screening questionnaire was completed by phone, and eligibility was confirmed through developmental testing at the baseline assessment visit. The first participant was enrolled in April 2017, and the last in May 2023, with all long-term follow-up assessments completed by fall 2024.
2) Standard protocol approvals, registrations, and patient consent
The study protocol was approved by the Institutional Review Boards at UCLA and BCH and was registered at ClinicalTrials.gov (NCT03422367). Before participation, caregivers completed the informed consent process in person with a trained research coordinator, during which they were given the opportunity to review the consent form and ask questions.
3) Power calculations
Target enrollment was set at 60 toddlers based on a detailed power analysis to ensure adequate detection of treatment effects. Anticipated dropout rates were 10% at the end of treatment and 20% at follow-up, with an additional 10% data loss expected at each time point. The study was powered at 80%, with a significance level of α=0.05, to detect a treatment effect size of d=0.83 standard deviations (SDs) between groups, assuming linear growth in effects over time. To achieve this, the design required complete data from at least 23 participants per group to detect a 0.9 SD difference at the end of follow-up visit. These assumptions were informed by prior findings reporting effect sizes of d=0.87 for joint engagement [14].
4) Clinical trial design
Participation in the waitlist-control randomized controlled study lasted 15 to 21 months (15 months for the immediate treatment group and 21 months for the waitlist group). A waitlist-control design was selected to ensure that all families ultimately received JASPER, which was considered ethically appropriate given the rarity of TSC and the limited availability of evidence-based interventions. This design also follows established precedent in behavioral intervention research, including prior JASPER trials and methodological guidance for psychosocial intervention studies [15,21,24]. The study took place across two research sites (n=34 at UCLA; n=27 at BCH). Assessment visits were conducted at baseline, end of treatment (3 months), and end of follow-up (6 months) (Fig. 1). The original in-person visit design included direct assessments, parent interviews, structured parent-child interactions (PCX), and electroencephalogram (EEG) recordings. EEG results will be reported separately, as the present analysis focuses on the primary preregistered endpoint.
Study design and assessments with remote adaptations. EEG, electroencephalogram; TAND, TSC-associated neuropsychiatric disorders.
Randomization was conducted after confirming eligibility through baseline developmental assessments; however, the randomization process itself was conducted independently of baseline results to prevent bias. Participants were randomly assigned to either the treatment or waitlist group using an online randomization program (Semel Institute Biostatistics Core [Sistat]), with stratification by age (≤24 months vs. >24 months). Allocation was concealed within the database system, and group assignments were revealed only after enrollment was completed, ensuring that study staff could not anticipate assignments. Age stratification was implemented to achieve a balanced distribution of developmental stages across groups, as age is a critical prognostic factor influencing social communication development and response to early behavioral interventions in this population [25,26]. Stratified randomization is recommended in smaller clinical trials to minimize baseline imbalances on key variables [27], and age-based stratification has precedent in early autism intervention trials, including studies that blocked randomization by younger versus older toddlers or used stratified minimization procedures with defined age bands in preschoolers [28,29]. Because the intervention was caregiver-mediated, caregivers and study coordinators were not blinded to treatment condition. However, interventionists and assessors were masked to group assignment. In the treatment condition, participants received the intervention immediately following the baseline assessment, whereas in the waitlist condition, participants received the intervention after completing the 6-month follow-up assessment. All caregivers in both groups were provided a summary of baseline assessment results and offered the opportunity to speak with a clinical psychologist and/or pediatric neurologist from the study team to review clinically relevant findings and receive guidance for recommended follow-up. Letters were provided to service providers upon request to facilitate access to community-based services for developmental delays. Referrals to specialists were also provided as needed by the two physicians at each treatment site (SJ and EAT).
5) Study design adaptations
Due to the COVID-19 pandemic, the originally planned in-person assessment visits were adapted to remote formats to ensure participant safety and compliance with public health guidelines. Importantly, the core trial design, randomization procedures, and primary outcome measures remained unchanged. Over the course of the pandemic, as safety requirements evolved and to accommodate families’ needs, some visits were conducted in person with modifications, while others were completed entirely remotely. A comparison of the full assessment batteries for in-person and remote assessments is shown in Fig. 1. Remote adaptations utilized secure, a secure video conferencing platform video conferencing platforms, and all protocols were adjusted to parallel in-person assessments as closely as possible (e.g., standardized toy kits were mailed to families, and identical coding procedures were applied). To maximize the analyzable sample size and maintain comparability across participants and timepoints, this report focuses on measures collected in both in-person and remote formats. Assessments available only in person (e.g., laboratory-based measures) were excluded because they could not be obtained remotely and therefore resulted in substantial differential missingness. When completion of the full battery was not feasible, partial data were collected, with priority given to the primary outcome measure of joint engagement.
6) Visit completion
All visits were scheduled within a four-week window of the target date (baseline, then 3 or 6 months post-baseline). Families missed assessments primarily due to medical reasons—most often seizure activity—or because of scheduling conflicts. When appropriate, these events were reported as adverse events unrelated to the intervention or study protocol (e.g., illness, hospitalization, surgery, seizures), though they contributed to missed or delayed data collection. Assessment visits were rescheduled as soon as possible. If a visit could not be completed within the designated window, it was marked as missing, and subsequent assessments proceeded as scheduled. As part of the study’s adaptive procedures, efforts were made to capture partial data whenever possible to prevent loss of critical information. The study team also provided financial assistance to support travel for in-person visits. Participants who requested support received reimbursement for flights and one night of hotel lodging. These costs were incorporated into the study budget and supplemented by private foundation support, including BeCureful. All participants received monetary compensation, either cash or gift cards, for completion of each assessment visit.
7) Behavioral intervention
The 12-week JASPER intervention program targeted children’s social communication, play, and engagement skills through evidence-based strategies delivered within a caregiver-mediated model [14,30]. A primary caregiver was identified to participate in the intervention, though other family members were permitted to join sessions if they wished. During each weekly session, a site-trained interventionist introduced and actively coached caregivers in implementing core JASPER strategies. These strategies included promoting joint attention by following the child’s lead, commenting, and imitating; facilitating symbolic play by modeling and expanding play schemes; enhancing engagement through responsive interaction; and supporting regulation by structuring the environment and responding appropriately to child cues. The intervention was tailored to the developmental profiles of toddlers with TSC by emphasizing foundational social communication skills, as participants represented a wide range of neurodevelopmental levels.
Originally designed for in-person delivery, the intervention was adapted prior to the COVID-19 pandemic to accommodate families who lived far from the study sites or faced medical complexities, increasing flexibility in the requirement for in-person sessions [17]. These protocol modifications were developed collaboratively with the study’s funding agency, approved by the institutional review board, and published subsequently to ensure transparency. One participant completed the entire 12-week intervention fully in person before remote adaptations were introduced. For the remaining 29 participants in the treatment group and all participants in the waitlist group, a hybrid delivery model combining in-person and remote sessions was implemented. All 30 participants in the treatment group completed at least one in-person session. Across all participants, the average number of sessions completed per participant was 9.9 (range, 1 to 12 total sessions). On average, participants completed 6.0 remote sessions (range, 0 to 11), with the remaining sessions conducted in person or missed. This flexible, hybrid approach was essential to improve accessibility and accommodate the medical complexities and geographic dispersion of the participant population, particularly during the pandemic period.
During weekly remote sessions, caregivers and interventionists met for 30 to 60 minutes via videoconferencing to introduce new strategies and discuss their implementation. These meetings were conducted using Zoom Video Communications, a secure and a secure video conferencing platform online platform approved in consultation with the institution’s Security Compliance Office. Each week, caregivers were asked to record and upload a video demonstrating their practice of the intervention skills with their child at home. Interventionists reviewed these videos and provided individualized feedback during the subsequent session to support caregiver fidelity. Although formal quantification of fidelity was not included in the current analyses, it will be examined in future work. Caregivers were also encouraged to practice JASPER strategies for up to 30 minutes daily and could report practice time via text message. Importantly, the core JASPER content and strategies were identical across in-person and remote formats; adaptations were limited to the delivery mode and the inclusion of video uploads for feedback.
2. Participant characteristics
1) Demographics and intervention history
A demographic questionnaire was administered at the baseline visit, including items on TSC diagnosis, prenatal history, developmental milestones, seizure history, and family demographics. Caregivers also completed a structured survey reporting the number of hours their child received in community-based interventions, including speech and language therapy, occupational therapy, physical therapy, intensive behavioral therapy, biomedical therapy, and other services. The total number of intervention hours at baseline was calculated and recorded. At each subsequent assessment visit, caregivers provided updates regarding community interventions and any changes in seizure activity.
2) Parent-child interaction assessment
The primary outcome measure was coordinated time spent in joint engagement, as assessed by the PCX. This measure was chosen as the primary outcome due to its strong theoretical foundation, established measurability in early development, and direct relevance to the targets of the JASPER intervention [20,31]. In contrast, outcomes such as language or adaptive behavior, while clinically important, were considered more distal and expected to show less change over the relatively brief intervention period. Joint engagement was therefore selected as the most proximal and sensitive indicator of JASPER’s anticipated effects in toddlers with TSC. In this population, early social communication impairments—such as limited eye contact, reduced shared affect, diminished social interest, and slower disengagement of attention—often precede an autism diagnosis, making joint engagement a particularly relevant and timely intervention target [13]. Other prespecified outcomes from the trial protocol (e.g., expressive social communication skills, symbolic play, cognition and language, and parent social communication) were collected and will be reported in subsequent manuscripts. The PCX was conducted with the primary caregiver who participated in the intervention, and this caregiver remained consistent across all assessments. The PCX was administered either in person or remotely, using a standardized set of toys provided to all participants. For remote assessments, the standardized toys were mailed to participants’ homes to ensure uniformity. Each session consisted of a 10-minute free play period that was video recorded while the parent and child interacted with the toy set. These sessions were designed to capture naturalistic interaction within a controlled framework, allowing for a detailed evaluation of play behaviors and interaction quality [14]. Trained coders subsequently analyzed the video recordings using a detailed coding system that assessed both parent and child joint engagement. Graduate student coders, uninvolved in other aspects of the study and blind to treatment group and timepoint, coded all videos following established protocols used in prior studies [21,22,32]. Inter-rater reliability for key observational variables—including total time jointly engaged (intraclass correlation coefficient [ICC]=0.95) and initiating joint attention skills (ICC=0.97)—was excellent, based on in-lab reliability testing and direct coding of 20% of the sessions. The primary outcome measure reported in this manuscript is the total percentage of time the dyad spent in parent- or child-initiated joint engagement.
3) Vineland adaptive behavior scales
The vineland adaptive behavior scales, third edition (VABS-3), was administered as a parent interview to assess functional skills at each assessment visit. The vineland adaptive behavior scales adaptive behavior composite score (VABS-ABC) score, which includes the domains of communication, daily living skills, and socialization, was used as a key covariate in subsequent analyses. Given the substantial heterogeneity in developmental profiles among children with TSC, the VABS-ABC provided essential descriptive data on participants’ adaptive functioning at baseline and across the study period.
4) Autism concern
All children in the study received an autism assessment tailored to their developmental level, age, and in-person availability, as detailed in Table 1. The assessment tools included the Brief Observation of Symptoms of Autism (BOSA), the Autism Observation Scale for Infants (AOSI), and the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2). Before the COVID-19 pandemic, ADOS-2 or AOSI assessments were administered in person. During the pandemic, the BOSA was administered either in person or remotely via Zoom, with standardized materials mailed to families. The AOSI is an examiner-administered instrument designed for early identification of autism-related behaviors in infants and toddlers [33]. The ADOS-2, a gold-standard examiner-led tool for diagnosing autism, involves structured observation of social communication and play [34]. The BOSA, developed for remote assessment during the pandemic and incorporating elements of the ADOS-2 administered by an unmasked caregiver, has demonstrated diagnostic convergence with the ADOS [35]. However, BOSA scores cannot be directly compared with ADOS scores. Results were categorized as either autism concern or no autism concern based on validated cutoffs. Four children had nonverbal developmental levels between 6 and 12 months, warranting cautious interpretation of autism-related symptoms in this subgroup.
Summary of autism assessment administration criteria, cutoffs for autism concern, and frequency of use in the study cohort
5) Statistical methods
All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). Baseline characteristics between the treatment and waitlist groups were compared using two-sided t-tests for continuous variables (e.g., age, percent time in joint engagement, VABS-ABC, hours of intervention) and chi-squared tests for categorical variables (e.g., sex, genotype, seizure history, autism concern, primary caregiver, distance from site, income, education, race, ethnicity). The treatment group demonstrated significantly lower baseline VABS-ABC scores than the waitlist group (mean±SD, 71.5±12.3 vs. 82.2±16.1, P=0.013). No other baseline characteristics differed significantly, although sex (P=0.091) and history of infantile spasms (P=0.074) approached significance (Table 2).
Categorical demographics for the primary caregivers participating in the intervention, randomized to treatment and waitlist groups
The prespecified primary analysis assessed changes in joint engagement from baseline to end of treatment, while a secondary analysis evaluated changes from baseline to end of follow-up. These outcomes were analyzed using analyses of covariance, with end of treatment or end of follow-up joint engagement as the dependent variable, group (treatment vs. waitlist) as the independent variable, and baseline joint engagement as a covariate. Because baseline VABS-ABC scores differed between groups, post hoc sensitivity analyses were conducted that included baseline VABS-ABC as an additional covariate to evaluate the robustness of the findings.
Joint engagement percentages were transformed using a logit function with a regularization constant (f(x)=log{(x+0.01)/(1–x)}) to reduce skewness. Covariates were centered on their baseline means so that intercepts could be interpreted as mean group differences at average baseline values.
Missing outcome data were addressed using multiple imputation under a multivariate normal model implemented via Monte Carlo Markov Chain methods (PROC MI in SAS). Thirty imputed datasets were generated, incorporating group, baseline VABS-ABC scores, baseline joint engagement, and the dependent variable. Results were combined using Rubin’s rules. Analyses were conducted on an intent-to-treat basis, including all randomized participants with available baseline data. Complete case analyses yielded results consistent with the imputed models. All P values <0.05 were considered statistically significant.
Results
1. Descriptive results
Sixty-one participants initially provided consent. Two were disqualified due to developmental level, resulting in 59 randomized participants (treatment, 30; waitlist, 29), as illustrated in the Consolidated Standards of Reporting Trials (CONSORT) diagram (Fig. 2).
Consolidated Standards of Reporting Trials (CONSORT) diagram for visit completion. 'Missing: n=' indicates participants for whom primary outcome (joint engagement) data was unavailable at that time point. This may reflect fully missed visits, partially completed visits where parent-child interactions data could not be obtained, or specific data collection failures for the primary outcome measure. Reasons for disqualification and dropouts are detailed in the 'Descriptive results' section. COVID-19, coronavirus disease 2019; Tx, treatment.
Across the study, seven participants withdrew from the randomized pool, representing 11.9% (7/59) of the total sample. As detailed in Fig. 2, dropouts occurred in both groups and at multiple time points. Reasons for withdrawal included scheduling difficulties and limited family availability (e.g., conflicting work or therapy schedules), medical complications (e.g., unexpected surgeries), and challenges maintaining engagement or contact with the research team (e.g., unresponsiveness or loss to follow-up). COVID-19–related circumstances were also cited as contributing factors. Specific numbers of dropouts by group are shown in Fig. 2.
Assessments completed outside the designated time window—defined as more than 4 weeks before or after the planned visit—were also documented. For the end of treatment visit, eight assessments fell outside the window (four in the treatment group and four in the waitlist group). For the end of follow-up visit, seven assessments were completed late (four in the treatment group and three in the waitlist group). Overall, 8.5% of all assessments were conducted outside the predefined window, most commonly due to scheduling conflicts, illness, or COVID-19–related disruptions.
Participant flow and availability of primary outcome data (joint engagement) are summarized in Fig. 2. Within the diagram, ‘missing’ indicates that primary outcome data were unavailable for a given visit. Across the study, 7.3% of all primary outcome data were missing, primarily due to incomplete assessments, behavioral challenges, or technical issues with video quality and data export.
Combining missing data and dropouts, overall data loss for key outcomes was 3.4% at the baseline visit, 18.6% at the end of treatment visit, and 18.6% at the end of follow-up visit. Based on a priori power calculations (which accounted for up to 10%, 20%, and 30% data loss, respectively), the available data were considered sufficient for detecting meaningful group differences.
Baseline demographic and clinical characteristics for children and caregivers are presented in Table 2. Descriptive data on total hours of community-based interventions at each visit are provided in Table 3. Although formal statistical comparisons were not performed for these measures, descriptive trends indicated that both the treatment and waitlist groups showed increases in community intervention hours over time, with a greater increase observed in the treatment group.
Treatment hours at baseline, end of treatment, and end of follow-up
1) Primary analysis: baseline to end of treatment
Model 1.1 included the main effect of group while controlling for baseline joint engagement. By the end of treatment, both groups combined showed significant increases in joint engagement (t=2.814, P=0.0050). There was no significant difference between the waitlist and treatment groups (t=–0.126, P=0.9000) when controlling for baseline joint engagement. Baseline joint engagement had a significant positive effect on end of treatment joint engagement (t=6.222, P<0.0001), indicating that participants with higher baseline joint engagement tended to maintain higher engagement levels at the end of treatment in both groups (Fig. 3). Model 1.2 assessed group differences in joint engagement at the end of treatment while controlling for both baseline joint engagement and baseline vineland adaptive behavior scales adaptive behavior composite (VABS-ABC) scores. Adjusting for baseline VABS-ABC scores did not alter the results. Both groups combined continued to show significant improvement over time (t=2.017, P=0.0441). There remained no significant group effect (t=0.666, P=0.5058), while baseline joint engagement (t=5.231, P<0.0001) and baseline VABS-ABC scores (t=2.222, P=0.0265) had significant positive effects on joint engagement at the end of treatment. All results are presented in Table 4.
Log odds of joint engagement at baseline and end of treatment for (A) the treatment group, and (B) the waitlist group. The estimated group means from model 1.1 with multiple imputation at the baseline and end of treatment are plotted in black dashed lines.
Fitting results for models
2) Secondary analysis: baseline to end of follow-up
Model 2.1 included the main effect of group while controlling for baseline joint engagement. By the end of follow-up, both groups combined demonstrated significant increases in joint engagement (t=6.246, P<0.0001), along with a significant group effect. The treatment group showed lower average joint engagement than the waitlist group at follow-up (t=–2.339, P=0.0195), controlling for baseline joint engagement. Baseline joint engagement was a strong positive predictor of joint engagement at the end of follow-up (t=5.454, P<0.0001), suggesting that participants with higher initial engagement maintained higher levels across both groups (Fig. 4). Model 2.2 evaluated the same relationship while controlling for both baseline joint engagement and baseline VABS-ABC scores. After adjusting for baseline VABS-ABC, the group difference was no longer significant (t=–1.531, P=0.1261). Both baseline joint engagement (t=4.627, P<0.0001) and baseline VABS-ABC scores (t=1.968, P=0.0493) remained significant positive predictors of joint engagement at follow-up. Thus, participants with higher initial joint engagement and/or adaptive behavior showed greater engagement at the end of follow-up. Because the waitlist group had a higher mean VABS-ABC score at baseline, these baseline differences likely contributed to the relatively greater gains in joint engagement observed for the waitlist group at follow-up. All results are summarized in Table 4.
Log odds of joint engagement at baseline and end of follow-up for (A) the treatment group, and (B) the waitlist group. The estimated group means from model 2.1 with multiple imputation at the baseline and end of follow-up are plotted in black dashed lines.
Discussion
As with many genetic syndromes associated with epilepsy and aberrant brain development, toddlers with TSC are at high risk for NDDs, with autism rates exceeding 50%, necessitating early surveillance and intervention. These neurodevelopmental conditions fall under the broader diagnostic framework of TAND, and it is increasingly recognized that most—if not all—individuals with TSC experience some manifestation of TAND during their lifetime. Individuals with TSC typically present with multiple TAND manifestations that may vary in nature and severity over time. Screening and diagnosis of TAND have now become core recommendations for TSC clinics worldwide. Although several natural history studies have examined early development and detection of NDDs in TSC, there have been no prior studies translating early detection into early behavioral intervention [36]. Rigorous studies of early intervention can inform clinical monitoring practices and improve advocacy for families navigating developmental services. JETS represents the first randomized clinical trial of a behavioral intervention in toddlers with TSC, motivated by prior findings identifying nonverbal communication impairments during the first year of life. We selected an intervention specifically designed to enhance social communication skills in toddlers and school-aged children with non-syndromic autism. Children aged 12–36 months with TSC and a wide range of neurodevelopmental levels at baseline were included. Recruitment proved challenging, and these difficulties were compounded by COVID-19–related restrictions, necessitating creativity and flexibility in study design that may inform future clinical trials. Although the hypothesis that the treatment group would demonstrate greater gains in joint engagement was not confirmed, improvements were observed across both groups, with the greatest gains seen in toddlers who began the trial with higher baseline adaptive behavior. While between-group differences were not statistically significant, both groups improved over time. These gains may reflect the combined influence of developmental surveillance and feedback provided to caregivers. Accordingly, we interpret these findings not as evidence of a lack of treatment benefit but as support for the therapeutic value of surveillance itself and the need to consider developmental readiness when designing caregiver-mediated interventions for TSC.
This trial also provides insights to guide future studies and to better understand the reasons for the null treatment effects. First, the JASPER intervention was modified to accommodate remote delivery. Although all participants in the treatment group completed at least one in-person session, the trial primarily employed a hybrid delivery model combining in-person and remote sessions. This flexibility was crucial to enhancing accessibility and feasibility for a medically complex and geographically dispersed population, especially as the study progressed through the COVID-19 pandemic. The benefits and limitations of these remote adaptations are discussed in a dedicated section below.
Second, to ensure representativeness, this study intentionally included a heterogeneous cohort. Historically, autism intervention trials have often excluded individuals with developmental delays, intellectual disabilities [37,38], or comorbidities such as epilepsy [39]. Prior JASPER trials have focused on toddlers or children with clear signs of autism [30] or confirmed autism diagnoses [40]. While this approach optimizes the likelihood of detecting treatment effects, it limits generalizability. In contrast, the toddlers with TSC in this study exhibited a broad spectrum of developmental profiles and abilities. For instance, four children had developmental age equivalents below 1 year (6 to 12 months), whereas others demonstrated near age-expected developmental levels. Notably, the treatment group began the trial with significantly lower baseline adaptive behavior than the waitlist group. This difference, combined with wide developmental variability, may have influenced responsiveness, as children with higher baseline adaptive skills showed greater gains in joint engagement regardless of group assignment. Although the sample size was too small to test moderators statistically, qualitative examination revealed substantial variability: some children demonstrated marked improvements, while others declined. Future studies with larger samples will be critical for identifying responder profiles and refining intervention tailoring. Some participants had not yet shown clear developmental delays at enrollment and may have exhibited emerging symptoms during the trial period, whereas others demonstrated more stable developmental trajectories. This variability likely contributed to differential responsiveness and underscores the importance of defining inclusion criteria and assessing developmental readiness in future early intervention trials for TSC.
Third, the observed improvements in joint engagement across both the treatment and waitlist groups highlight the potential therapeutic value of developmental surveillance. This phenomenon, where consistent monitoring and feedback alone produce measurable benefits, has been documented in both prospective studies and randomized controlled trials involving children with autism [41,42]. In JETS, participants in both groups received detailed clinical assessment reports and expert guidance, which likely increased caregiver awareness of their child’s developmental needs and influenced their daily interactions. These reports may have enhanced the waitlist group’s ‘control’ condition beyond typical ‘community services as usual,’ facilitating access to additional interventions and modifying caregiver-child engagement patterns. Descriptive findings support this interpretation: both groups increased their community intervention hours over the study period, with a greater rise observed in the treatment group by the end of follow-up. It is also possible that, beyond the quantity of services, the quality of interventions improved as caregivers became more informed about their child’s specific developmental needs through participation in study assessments.
While this pattern reflects the clinical realities faced by families, it also illustrates a methodological challenge inherent to waitlist-control designs in behavioral research. Surveillance and increased service access can attenuate measurable treatment effects, reducing between-group differences. At the same time, this underscores a broader methodological tension in early intervention trials. Children are often excluded from autism studies if their outside services change during participation to minimize variability that could obscure treatment effects. However, enforcing such standardization in real-world settings is difficult and may even hinder developmental progress and family well-being—particularly in early childhood, when needs evolve rapidly. Researchers must therefore develop strategies to measure and account for the influence of concurrent community interventions, or potentially leverage them to guide clinicians in making more targeted service recommendations during routine care. Ethically, the waitlist-control design ensured that all families ultimately received the intervention in this rare-disease population while also providing a unique opportunity to examine the effects of surveillance alone. Exit interviews confirmed that families valued this aspect of participation, reinforcing the idea that developmental monitoring and feedback may serve as an active therapeutic component—one that helps shorten the diagnostic odyssey for families of children with rare NDDs [43].
1. Lessons from remote delivery
The JETS trial underwent two major phases of protocol adaptation. Initial modifications were implemented early in the study to address recruitment barriers, particularly given the rarity and medical complexity of the TSC population [17]. These changes increased feasibility and allowed the study to proceed despite initial enrollment challenges. Subsequently, the COVID-19 pandemic required additional protocol revisions, most notably a significant transition to remote delivery. JETS therefore represents an important early effort to systematically evaluate a hybrid intervention model integrating both in-person and remote components. This approach yielded valuable insights into how behavioral interventions can be flexibly delivered, raising key questions about optimizing accessibility and family autonomy while maintaining intervention fidelity and ensuring adequate ‘dose’ of in-person interaction [44-47].
Remote adaptations offered both advantages and challenges. Families reported that eliminating travel reduced participation burden and, for some, enhanced engagement. Remote delivery expanded accessibility for geographically distant families and aligned with the realities of caring for toddlers with complex medical and developmental needs. However, caregivers also noted that home-based sessions introduced distractions such as siblings and household noise, which sometimes reduced child engagement compared to private laboratory settings. Practical challenges—such as finding a quiet space, maintaining stable internet connections, and resolving technical difficulties—occasionally affected session quality or caused delays. Engagement, attendance, and adherence were similarly mixed: some families found remote delivery more convenient and motivating, while others reported that it made skipping sessions easier.
For assessments, the shift to remote administration introduced heterogeneity across participants, raising questions about the comparability of in-person and remote measures and interventions. Although this trial was not designed to test these differences, future research should evaluate their equivalence directly. While these modifications added new considerations to the statistical analysis, they also highlight the importance of designing flexible and accessible trials for these populations. Adaptations required ongoing IRB and compliance review, which were implemented successfully but added procedural complexity. This experience underscores a persistent challenge in rare disorder research: determining which aspects of interventions and assessments must remain consistent and which can be adapted. Balancing flexibility, feasibility, and methodological rigor is essential given the continued challenges of recruitment, enrollment, and attrition in rare-disease trials [48]. An ongoing fully remote JASPER intervention study in TSC (NCT04698538) may offer further insight into the effects of a consistently delivered remote design. Future studies should build on these design lessons to develop shorter-duration trials with broader applicability.
2. Future research
Precision health extends beyond pharmacology and gene editing. Behavioral intervention studies can also pursue personalization by identifying predictors of response and tailoring intervention strategies to each child’s developmental and behavioral profile. Motivated by the observed individual variability in joint engagement trajectories, future analyses will focus on identifying subgroups of children within the JASPER treatment group who benefited most, refining recommendations for clinicians advocating for developmental services for toddlers with TSC. Understanding the characteristics of participants who responded well versus those who did not will facilitate more individualized interventions. Kasari et al. [49] have applied a precision-health framework through sequential multiple assignment randomized trials, in which treatment dosage and type are modified based on interval assessments of response. Following this approach, future analyses will examine the outcomes of the waitlist group after participation in the JASPER intervention, allowing exploration of whether a period of developmental surveillance followed by treatment produces greater benefit than immediate intervention alone. Detailed characterization of the JETS cohort—particularly regarding autism symptomatology—may further clarify sources of differential responsiveness to both intervention and surveillance. Alternative methodological approaches, such as modular designs or preference-based randomized trials that incorporate family characteristics and preferences into randomization, could enhance personalization. Future analyses may also examine the various types of engagement and parenting behaviors recorded in this study to better understand which changes resulted from the intervention, as well as the influence of caregiver factors—such as quality of life, available supports, and stress—on engagement in a parent-mediated model. Finally, understanding the broader impact of developmental surveillance and monitoring on caregiver-child interactions will be critical. Simple, feasible enhancements to clinical visits, implemented by informed providers, may help strengthen developmental skills and reduce the TAND burden in TSC.
3. Conclusions
The JETS trial represents a systematic effort to evaluate the effects of behavioral intervention on autism-related outcomes in TSC. It also demonstrated the potential therapeutic value of developmental surveillance and the adaptability required to conduct effective behavioral intervention trials in medically complex developmental populations. The findings reinforce the importance of personalized, developmentally informed approaches and underscore the need to balance accessibility with methodological standardization in early interventions for toddlers at risk for NDDs.
Notes
Conflicts of interest
No potential conflict of interest relevant to this article was reported.
Author contribution
Conceptualization: SJ, DS, EAT, CAN, and CK. Data curation: SJ, CHT, NM, MP, and CK. Formal analysis: SJ, CHT, DS, WS, and MD. Funding acquisition: SJ, DS, EAT, CAN, and CK. Methodology: SJ, DS, EAT, CAN, and CK. Project administration: SJ, CHT, NM, WS, MP, JP, CAN, and CK. Visualization: DS and MD. Writing - original draft: SJ, CHT, DS, and MD. Writing - review & editing: SJ, CHT, DS, NM, WS, MD, MP, JP, EAT, CAN, and CK.
Acknowledgments
This study was funded by the National Institute of Child Health and Human Development (1R01HD09013801A1). The study also received a donation from the TSC Alliance to support mailing materials for at-home assessments during the COVID-19 pandemic.
All data for this study are publicly available in the National Database for Autism Research (NDAR).