Symptom rates and profile clustering in tuberous sclerosis complex-associated neuropsychiatric disorders (TAND)

Tuberous-sclerosis complex (TSC) is an autosomal-dominant genetic disorder caused by pathogenic variants in TSC1 and TSC2, affecting about 1 in 6000 live births and about 1 million people worldwide [1]. TSC affects all major organ systems with high variability in disease severity. TSC is also associated with a range of behavioral, intellectual and academic, psychiatric, neuropsychological, and psychosocial difficulties impacting approximately 90% of individuals with TSC and representing a significant burden [2, 3]. It has been shown that mental health is a significant part of the disease and that cost is a major barrier to improvement [4].

To improve identification of neuropsychiatric difficulties experienced by individuals with TSC, the Neuropsychiatric Panel at the 2012 TSC Consensus Conference developed the term TAND (TSC-associated Neuropsychiatric Disorders) [5]. TAND encompasses six “levels” of investigation: (1) behavioral level, (2) psychiatric level, (3) intellectual level, (4) academic level, (5) neuropsychological level, and (6) psychosocial level [3]. TAND symptoms are highly prevalent in patients with TSC overall, yet only 20% of patients receive appropriate assessment and treatment [3]. The TuberOus Sclerosis registry to increase disease Awareness study (TOSCA), a large international registry capturing the natural history of TSC, noted high rates of missing and unreported TAND data presumably from lack of assessment and treatment, sometimes greater than 60% for behavioral symptoms and intellectual levels [6].

TSC Consensus Conference recommendations include that individuals with TSC be assessed for TAND every year with the goal of lessening the identification gap between symptoms and recognition and identification by clinicians [3, 5]. With this consensus in mind, the TAND Checklist-L was developed as a screening tool to guide clinicians in discussing the different levels of TAND [3]. The checklist includes a total of 11 sections, with mostly yes/no questions and covers each level of investigation and was validated by Leclezio and colleagues in 2014 [7].

Leclezio and colleagues also attempted to identify natural grouping of symptoms or “clusters” with the goal of identifying groups of specific characteristics across levels of TAND and subsequently enabling healthcare providers to implement appropriate treatments focused on individual clusters/symptom groups. Initially, a feasibility study was completed (with 56 participants from South Africa and Australia) [8], followed by a larger replication study (with 85 participants from the TOSCA registry) which showed a six-cluster solution [9]. These were namely (1) scholastic, (2) neuropsychological, (3) mood/anxiety, (4) ASD-like, (5) challenging behaviors, and (6) hyperactive/impulsive [8, 9]. In a larger study with bootstrapping and internal consistency, a final seven-cluster solution was selected, including the addition of one for eating/sleeping that previously had been divided into the ASD-like cluster (eating) and mood/anxiety cluster (sleeping) rather than its own stand-alone cluster [10].

A total of 1545 checklists were completed for 668 participants (250 (37%) adults and 418 (63%) children) within the 3-year time period (an additional 99 records were removed for > 50% missing data). Demographics for the cohort can be seen in Table 1. Approximately 37% of the participants were only seen one time. Of those seen for multiple visits, the average interval between appointments was 324 days, with 70.8% seen within one year of previous appointment (73% in children and 67% in adults). Of note, 65% of the participants did not self-report intellectual ability (as reported by themselves, parent, or caregiver of previous formal evaluation). Of those who did self-report, about 38.5% of the participants had normal intellectual ability, with another 40% with mild-moderate intellectual disability. Parental/guardian assessment of intellectual ability and official IQ testing was generally consistent (weighted-kappa κ = 0.72, 95% CI 0.68–0.77).

Six hundred two individuals (90%) reported at least one TAND symptom at the first visit. The average number of symptoms at the first visit was approximately twelve. Rate of each TAND symptom at the initial visit can be seen in Table 2. The most common symptom was neuropsychologic difficulty with attention (59.4%). Other commonly reported symptoms were scholastic, neuropsychological (namely difficulties with multitasking and other executive skills), and behavioral (namely difficulties with attention, mood swings, and anxiety), all of which were present in more than half the participants at their first visit. When divided by age, the most common symptoms for adults were anxiety, depression, mood swings, sleep, and memory problems (all statistically more common than in children). Most common symptoms for children were temper tantrums, overactivity, impulsivity, and scholastic difficulties (often statistically more common than in adults). There was not a significant increase in rate of symptoms over the subsequent visits, with biggest increase noted for sleep difficulties (+1.6%).

Symptom rate by IQ was also analyzed for the 234 participants for whom a formal IQ level was known (Table 3). Most symptoms at the behavioral, intellectual, academic, and neuropsychological levels were significantly less common in those without intellectual disability (IQ ≥ 70) compared to those with intellectual disability (IQ < 70). The exception to this relationship was the psychiatric level, where only the diagnosis of autism spectrum disorder (ASD) and intellectual disability (IQ < 70) remained significantly associated with one another.

Using the TAND Checklist, we report on the first study to identify distinct clinical profiles within the TSC population via cluster and factor analysis followed by K-means cluster analysis. The majority of participants were administered the TAND checklist at yearly intervals, which is in line with current recommendations [3, 5]. The rate of TAND symptoms reported over the three-year period of the study remained overall stable. There were important differences in rates of symptoms when individuals were divided by age (children vs adult). Children tended to have more external behavioral symptoms, while adults tended have more internal mood symptoms. This pattern was consistent with the large, multinational TOSCA registry involving 2093 participants [6] and is consistent with age-based expectations in psychopathology [15], as well as other neurodevelopmental disorders like ADHD [16].

Of note, our results overall found similar rates of TAND symptoms as those of previous studies, with approximately 90% of individuals reporting at least one TAND symptom [3, 17]. Some individual TAND characteristics in our cohort were nearly identical compared to TOSCA with reported rates for intellectual disability (54%), ASD (21%), and ADHD (20%) [18]. However, we found that rates for most mood and behavioral symptoms in our cohort were consistently higher (22–57% vs. 7–21% in TOSCA). These differences could not be explained by age or sex, even when splitting the cohorts between children and adults. We suspect the differences are better explained by the methodology used for data collection. Whereas our study relied solely on the TAND Checklist for data collection, TOSCA employed a variety of methods, including patient self-report on the patient experience and quality of life, review of hospital discharge and clinic visit files and electronic medical records, patient questionnaires, and ad hoc clinical databases [19]. Furthermore, the TAND Checklist was administered to every patient seen in our TSC clinic over the reporting period while TAND-related data was reported missing for 25–40% of participants of TOSCA [18] despite the wide net for data collection used. To investigate these differences further, TAND Checklist responses from our cohort could be directly compared with the subset of TOSCA participants prospectively assessed using the TAND Checklist (n=85), results of which were published in 2020 but did not include individual checklist item frequency rates [9]. Perhaps a greater influence on observed differences of TAND symptoms in our cohort compared to TOSCA is the widespread discrepancies in TAND assessment and treatment generally [3, 20], which may be minimized at our center by the universal adoption of the TAND Checklist into our clinical practice as recommended by the 2012 International Consensus Group Recommendations for TSC [5] that creates heightened awareness of TAND in patients/parents/caregivers seen at CCHMC and our team of TSC clinicians. Furthermore, inclusion of a developmental pediatrician, neuropsychologists, and psychiatrists into our multidisciplinary team of TSC specialists removes barriers to formal assessment for TAND and consultation with TAND specialists.

Intellectual disability is known to be highly prevalent in TSC, with a reported rate of 45% in toddlers at 24 months in the TSC Autism Center of Excellence Research Network (TACERN) [21] to 55% in patients of all ages in the TOSCA study across all ages in TOSCA. In our study, the rate of parent/caregiver/self-reported intellectual disability was 62%. Although formal evaluations of IQ were only available for 35% of our sample, participants with reported intellectual disability (estimated IQ < 70) were concordant with formal evaluations. Intellectual disability correlated with higher rates in most TAND symptoms reported via the TAND Checklist including psychiatric mood disorders and ASD, consistent with previous studies [22, 23]. It is therefore even more important for providers to be especially careful when assessing TAND symptoms in those with intellectual disability. We also found an association between intellectual ability and ADHD diagnosis, though this was not found in other studies [23].

We identified seven TAND symptoms profiles occurring naturally in TSC, spanning the full spectrum of low to high symptom burden. The cluster and factor analysis used to identify these TAND symptom profiles were highly concordant in our study and similar to many aspects of the previous cluster analysis performed by Leclezio et al. in a pilot study of South African and Australian TSC patients (n = 56) [8] and a more recent analysis by de Vries et al. of TOSCA registry participants (n=85) [23]. All three studies found Ward’s method for hierarchical clustering best suited for identifying naturally occurring TAND symptom clusters in TSC. However, key differences are present. Our clustering solution identified 7 symptom groups, whereas the clustering solution of the two prior studies identified just 6 symptom groups. All three studies also identified scholastic, neuropsychological, and behavioral dysregulation symptom clusters, but individual TAND checklist items defining each symptom cluster frequently differed from study to study. This led to some unique findings in our analysis, such as the single neuropsychological cluster in the previous two studies being divided into separate memory/communication and executive function/attention clusters, and the single symptom cluster for mood/anxiety being divided into separate clusters for mood swings/anxiety and depressed mood/shyness. Perhaps the most surprising difference was the lack of specific ASD-like and hyperactive/impulsive clusters identified in both previous studies. Instead, these symptoms sorted in cohort within other clusters for the memory/communication, behavioral dysregulation, and vegetative symptoms. We cannot explain these differences other than the larger size of our cohort providing granularity not appreciated in smaller cohorts and the potential impact of differences that may be unique to our center to influence TAND symptom identification and reporting by patients, as discussed above. Valuable insight into these possibilities will be possible via analysis of TAND Checklist responses that are being collected in large, multicenter studies including the Developmental Synaptopathies Consortium of the Rare Diseases Clinical Research Network (clinicaltrials.​gov NCT02461459) and the TANDem Consortium (tandemconsortium​.​org).

Poor sleep, difficulties with eating, and difficulties with peers did not load well on any individual factor driving individual cluster symptomology. In the studies by Leclezio et al .[8] and de Vries et al .[23], poor sleep loaded onto the factor best correlated with the mood/anxiety cluster and difficulties with eating and peers loaded onto the factor best correlated with the ASD-like cluster. Intuitively, sleep, eating, and interaction with peers may be closely connected with a variety of other symptoms investigated via the TAND checklist and therefore not closely loading with any one factor but rather many of them. Sleep disturbances in particular may be difficult to load on individual factors contributing to TAND symptomology as it may also be driven by factors fully independent of TAND. For example, circadian rhythms are known to be perturbed directly via the mTOR pathway [24], as well as influenced by epilepsy and anti-seizure medications [25].

We extended our study beyond the hierarchical cluster analysis and factor loading of individual symptoms to identify naturally occurring TAND patient profiles defined by the constellation of symptom clusters occurring together in individual patients. We found seven unique TAND patient profiles using this approach, ranging from high to low symptom burden. Approximately 20% of participants exhibited the highest symptom burden with involvement of all factors and another 19% exhibited high symptom burden involving five or more factors. These patients highlight the need for TSC specialists to be familiar with the full spectrum of symptoms contributing to each level of TAND (behavioral, intellectual and academic, psychiatric, neuropsychological and psychosocial), including their diagnosis and management. While 31% (n=198/642) of patients were classified as having the lowest symptom profile in which no single common factor was identified as responsible for their individual TAND profile, only one third of individuals in this group (n=66) had zero TAND symptoms reported at the initial visit. Therefore, individuals in this group still periodically warrant further evaluation as per current guidelines recommended [3, 5].

TAND represents a common and often burdensome aspect of TSC. Our findings validate previous research on natural clustering of TAND symptoms and identify distinct individual TAND symptom profiles defined by constellations of individual symptom clusters occurring together within similar individuals.