Lessons from Kids-HD

In 2009, the National Institute of Health (NIH) -funded Kids-HD study (ClinicalTrials.gov Identifier: NCT01860339) was launched at the University of Iowa, investigating the impact of mHTT on brain development in children at risk for HD [7]. For research purposes only, children aged 6–18 years-old at risk for HD (defined as having a parent/grandparent with HD) were genotyped and categorized as gene-expanded (i.e., GE, CAG >= 36) or gene-non-expanded (i.e., GNE, CAG < 36). The GNE children served as an excellent control group, given that being raised in an HD family can produce unique psychosocial strain not captured by children of non-HD families. A separate group of age-matched participants with no familial link to HD was also recruited as a comparison group. The Kids-HD study modeled the progression of HD several decades before the predicted age of motor onset in the GE participants. In total, we collected data from 406 individuals (13.73 ± 4.33 year; 177 Males, 227 Females; 2 unspecified) across 663 visits. Over the years, the Kids-HD study produced several key findings (Table 1). Some highlights include:

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Summary of Kids-HD papers/findings.

https://doi.org/10.1371/journal.pone.0336088.t001

• HTT impacts brain structure and function along the entire spectrum of repeats: Higher repeats are associated with better cognitive skill and greater brain volume in GE and GNE individuals [13,14].
• Methodological challenges of modeling age and CAG separately, versus years-to-onset model (which incorporates both): The Kids-HD study employed an accelerated longitudinal design (ALD), the gold standard for evaluating brain development in children, which combines cross-sectional and longitudinal data to model age-related changes across a broad developmental window. This model showed not only interesting findings of accelerated striatal development, where GE children reached peak striatal volume years before the GNE group, but that these trajectories were profoundly affected by CAG repeat length [15]. We also found that every CAG repeat conferred a greater advantage in general cognitive ability [14]. This effect peaked at repeats of 43 and individuals with longer CAG repeats showed trajectories that were declining in function after age 15, likely representing changes due to the degenerative phase of the disease. The age-trajectory modeling results highlighted the profound consequences HTT and CAG repeat-length has on neurodevelopment. These findings recapitulate the well-established negative relationship between CAG repeat length and age of motor onset and require an updated model to take both factors, and their interactions, into account. Given the influence CAG repeats and age exert on longitudinal brain development, we began modeling the data in a Years-To-Onset (YTO) fashion to evaluate the time course of the disease from development to degeneration. The YTO model predicts the age of motor onset based on the participant’s current age and CAG repeat length. Importantly, the YTO model can reveal information about disease trajectory that is obscured when plotting data by age. For example, Reasoner et al. (2022) found that cortical volumes are similar across age groups [16], while Neema et al. (2024) found evidence for higher and lower cortical volumes depending on the YTO of the individual [17]. The critical difference between approaches is that while higher age and lower YTO both predict decline in cognitive and motoric function, age-only models ignore variability in disease trajectory caused by CAG repeat length. An example of the difference between an age-only trajectory and a YTO model is as follows: An 8-year-old with a CAG repeat length of 48 and a 22-year-old with a CAG length of 43 both have 26 YTO, despite being on opposite ends of the age trajectory model. Thus, potential differences in cortical volumes could have been missed by Reasoner et al. (2022) because participants at similar ages were at different points in their disease trajectory. Additionally, when evaluating a biomarker of degeneration in Kids-HD (neurofilament light; “NfL”), we found that NfL levels are normal as far back as 40 YTO but rise around 20 YTO [18]. These findings support the notion that the YTO model can accurately characterize the developmental phase of the disease process, which occurs 20 years before the onset of motor symptoms.
• The antagonistic pleiotropy theory of HTT and its implications for neurodevelopment: In our most recent publication, we revealed that GE individuals from 50−20 YTO have larger cerebral volumes, more cortical surface area, and higher IQs compared to GNE subjects, but the opposite is true after 20 YTO [17]. Striatal and pallidal volumes remain comparable in volume to GNE until 20 YTO after which they dramatically decline in volume, consistent with early vulnerability of the basal nuclei. These findings challenge the traditional concept of HD as a uniquely neurodegenerative disorder and highlight the neurodevelopmental role of mHTT. Additionally, Kids-HD research uncovered potential early advantages of mHTT in GE children, such as larger brain development that supports enhanced cognitive development and reduced depression/anxiety. However, these children will indeed develop HD, highlighting the yoking of advantage with disadvantage. Our findings support the notion of antagonistic pleiotropy in HD, where the same neurodevelopmental differences that confer early advantage eventually lead to vulnerability and subsequent degeneration [19]. Collectively, they reshape our understanding of the evolutionary implications of mHTT and its potential role in brain development

Beyond Kids-HD: Embracing a new era with ChANGE-HD

While the Kids-HD study provided novel insights, it also raised several critical questions that required further investigation. Kids-HD’s modest single-site sample (202 GE children) produced valuable data but required a necessary expansion for greater reliability and broader application. This prompted the launch of the NIH-funded, multi-site ChANGE-HD study in 2019, which builds upon the Kids-HD findings and addresses the following key areas:

• Expansion and Replication: No single study can provide definitive conclusions. ChANGE-HD aims to reproduce the Kids-HD findings using a larger sample size across multiple sites to ensure more robust, generalizable data.
• CAG-specific effects: Kids-HD identified CAG-specific effects in the GE group, where different repeat lengths produce distinct neurodevelopmental trajectories. To fully understand these effects, ChANGE-HD will recruit a larger cohort to enable a thorough analysis across a fuller range of repeats.
• Wider age range amongst participants: Kids-HD originally recruited participants from age 6 through age 18, but maturational changes extend into late 20s and early 30s [26]. Although YTO models address the strict need for sampling participants across a broad spectrum (as demonstrated by the example above), it is still beneficial to examine older participants who are approaching their date of motor onset to better model trajectories at low YTOs.
• Evaluation of serum biomarkers of disease progression – NfL: Serum NfL is a marker of neuronal damage in premanifest HD (among other neuronal diseases and disorders), with levels rising approximately 15–20 YTO, and is strongly correlated with the age of motor onset. Although NfL is classically absent in the developmental phase and appears after the onset of subtle neurodegeneration, its use as a biomarker for HD remains unclear and requires further investigation. The CHANGE-HD study aims to clarify the trajectory of NfL in younger at-risk individuals and explore its relationship with clinical outcomes and MRI findings.
• The ChANGE-Juvenile Onset Huntington’s disease (JOHD) Study: Expanding our sample size across multiple investigation sites allows us to delve deeper into the 5–10% of HD cases where symptom onset occurs before the age of 21, or JOHD [27–29]. Therefore, along with studying adolescent and adult-onset HD, we have included a companion study of ChANGE-JOHD where we will investigate the neuropathology of this rarer and more aggressive form of HD. This study will leverage the infrastructure of the ChANGE-HD study and follow a similar protocol.

The study also seeks to expand on other HD-related studies, such as TRACK-, PREDICT-, IMAGE-HD, and HD-YAS. Beyond the fact that we are studying a population that is much further from the possibility of disease manifestation, ChANGE-HD has other distinctions. First, whereas TRACK-, PREDICT-, and IMAGE-HD examined the disease from the earliest point of neurodegeneration and just prior to cortical and subcortical volumetric atrophy, the ChANGE-HD study focuses on developmental alterations in neural connectivity and structure, with the hypothesis that mHTT confers a pre-manifest benefit to inheritors. Second, whereas other studies have sought to determine the earliest potential time for interventions we seek to determine whether there is a point where early intervention may be detrimental. Third, we expand on HD-YAS by determining whether increases or decreases in cellular stress markers, such as NfL can be observed during the neurodevelopmental phase of the disease. Fourth, our broad toolbox of instruments has been updated to include more sensitive measures that can capture performance variability in individuals < 10 years old, such as the Q-Motor test, which relies on force transducers, rather than observation, to quantify motor system function [30]. In addition to self-reports, ChANGE-HD also extends beyond these studies by recording participant data from parents’ reports as well. Lastly, our neuroimaging protocols have been updated to conform to other large-scale studies of children and adolescents (see Magnetic Resonance Imagine section below) [31]. We note that although ChANGE-HD is distinct in many ways, we maintain several measures used in previous studies, such as the core neuroimaging scans to measure neural volume (including T1 and T2-weighted scans), verbal fluency tests, CAG repeat length assessments, and NfL quantification.

The goal of ChANGE-HD is to model development (versus degeneration) by focusing on individuals whose brain changes likely reflect neurodevelopment rather than disease progression. The broader objectives include:

1. Using structural and functional MRIs to monitor brain development over time. A specific focus is given to striatal circuitry, due to its centrality to HD and its prominence in earlier data.
2. Evaluate the relationship between brain structure and cognitive, behavioral, and neurologic assessments. Here, we aim to elucidate the precise effects of mHTT on cognition.
3. Investigate the link between CAG repeat length and clinical (motor scores, cognitive evaluations etc.) and biological (neuroimaging, biomarkers) measures. This includes evaluation of NfL as an appropriate marker of neurodegeneration in HD.
4. Investigate the putative roles of genetic variants on development. CAG repeat length is the most significant driver of symptom progression in individuals with mHTT. However, there still exists variation in the age of motor onset in individuals with similar CAG repeat lengths. This fact has motivated genome-wide association studies to identify variants that delay (e.g., FAN1) or accelerate (e.g., MSH3) the age of motor onset in patients with HD [32–34]. Several of the identified variants exist on genes implicated in DNA repair [33,35], ultimately exacerbating or preventing somatic instability. While these genes are typically studied in the context of HD neurodegeneration, several play a role in disorders associated with altered neurodevelopment, such as schizophrenia and autism [36], and HD and muscular dystrophy [35]. Using blood or saliva samples obtained at each visit, we will explore whether these genetic variants accelerate or delay development.

The study objectives will be achieved by improving upon the Kids-HD infrastructure, including adding more study sites and adding more precise measures (e.g., Q-Motor). We expect the study to validate pre-manifest biomarkers of neurodegeneration and their relationship with the age of motor symptom onset. We also expect to reveal novel and potentially opposite effects of these biomarkers on neurodevelopment:

• Molecular/Genetic
  1. ◦ Neurofilament light [18]
  2. ◦ Genetic modifiers, including MSH3, FAN1, MLH1, PMS1, PMS2, MLH3, TCERG1, GPR161, RRM2B, and CCDC82 [33]
• Neuroimaging
  1. ◦ Striatal, cortical, and white matter volume [17,37,38]
  2. ◦ Cortical surface area, folding index, and curvature [17]
  3. ◦ White matter integrity in corpus callosum and corticostriatal pathways [39]
  4. ◦ Prefrontal-striatal, motor-striatal, and default mode network resting state functional connectivity [40]
  5. ◦ Hub organization [41]
  6. ◦ Network segregation [41]
• Cognitive
  1. ◦ General intelligence [38,42]
  2. ◦ Cognitive processing speed [42]
• Motor
  • Finger tapping variability (increased irregularity in rhythm) [37,43]
  • Grip force fluctuations [37,43]
  • Frequency of repetitive movements [37,43]

Ethics (protocol approvals, consents, blinding, and safety)

Study approval and centralized oversight were provided by the WIRB-Copernicus Group (WCG; formerly WIRB; study number: 1269202; IRB Tracking Number: 20192908). WCG ensures compliance with federal regulations, ethical guidelines, local laws, participant safety, and study integrity. The study was conducted according to the principles expressed in the Declaration of Helsinki. Consent (or assent) procedures vary based on the age range of our participants. Assent is a simplified version of the consent document presented in age-appropriate language:

• Ages 6–7: The assent form is verbally explained to minor participants, and they will choose whether to provide verbal assent. If verbal assent is provided, it is indicated on the physical consent document by the research team memeber who explained the assent form to the minor participant. A parent or legal guardian reviews and chooses whether to provide full written consent on the participant’s behalf.
• Ages 8–11: Literate participants have the assent form verbally explained to them, read, and decide whether to provide written assent. A parent or guardian then decides whether to provide written consent on their behalf.
• Ages 12–17: Both the participant and their parent or legal guardian review the full consent form with a research team member and decide whether to provide written consent.
• Ages 18–30: Adult participants review the full consent form with a research team member and decide whether to provide written consent on their own behalf.

Collecting predictive genetic data for HD in individuals below 18 years of age raises the risk of confidentiality and privacy breaches, the results of which can have highly negative psychological consequences on those individuals. Thus, a study of this kind warrants commensurate security measures. In accordance, all testing is performed in a double-blind manner and all research team members who interact with participants or their families remain blind to the participant’s genetic status. Those who have previously undergone genetic testing are asked not to disclose their results. Team members responsible for data analysis only have access to de-identified data and do not have any contact with participants or their families. At no point, including after the study’s completion, will any team member have access to both identifying information and genetic status. Genetic results are never shared with participants, their families, medical providers, or anyone with access to identifying information. Additionally, participants are free to withdraw from the study at any time. Data obtained prior to the time of withdrawal are retained, unless the participant or their caregiver specifically requests that all collected data be discarded. To verify that our protocols ascend to the appropriate level of security required to protect the records of all participants giving genetic samples in our study, the ChANGE‐HD study protocols were submitted to several HD patient advocacy groups, including Help4HD International, the Huntington’s Disease Society of America (HDSA), and the Huntington’s Disease Youth Organization (HDYO). All groups have reviewed and endorsed our security measures.

Design

The ChANGE-HD study employs an accelerated longitudinal design (ALD), which is widely regarded as the gold standard for assessing developmental changes in brain development in children [44]. Instead of a traditional longitudinal design with two necessary time points separated by a specified interval, all visits across all participants are combined to create the largest possible pool of assessments. This complementation across hundreds of participants allows for data to be obtained across a wide range of YTO and allows more flexibility for individuals to participate as they can have their follow-up visits anytime nine months after their previous visit, if desired. This design allows us to combine cross-sectional and longitudinal approaches by recruiting participants across the desired age range. The clinical and imaging protocols are modeled after the ‘Adolescent Brain Cognitive Development’ (ABCD) study, providing a well-established framework for cognitive, behavioral, and MRI assessments in a longitudinal multi-site setting [31]. While the protocol has been designed such that several of the measures are standardized for comparison across the range of 6–30 years-of-age (e.g., NIH Toolbox instruments), other measures, such as the Quantitative-Motor (Q-Motor) require age-correction. For every visit, informed consent is obtained, screening procedures are performed, and, if the participant is of childbearing potential, a pregnancy test is performed. The following assessments are completed at each study visit (Table 2; see S2 Table for more details):

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Procedures performed at each ChANGE-HD visit.

https://doi.org/10.1371/journal.pone.0336088.t002

Sites

The University of Iowa serves as the primary site for the ChANGE-HD study and received IRB approval from the WCG to begin recruitment on 11/18/2019. Four additional sites were added, including the Children’s Hospital of Philadelphia (CHOP), Columbia University, the University of Texas at Houston, and the University of California, Davis. Site were approved for recruitment by WCG on 11/16/2020, 11/16/2020, 5/12/2020, and 4/16/2020, respectively. Vanderbilt University was added to the study and received approval to recruit participants on 11/14/2024.

Study period and recruitment goals

ChANGE-HD was originally planned as a 5-year study, aiming to complete 1,500 assessments of 400 participants by the end of the fifth year. However, the COVID-19 pandemic delayed study progress by causing significant disruptions in recruitment and scheduling across the five original study sites and required us to update our timelines after Year 2 (Table 3). As a result, the assessment period, including pre-COVID data collection, was extended to approximately seven years and our recruitment targets were changed to 450 participants across 1,030 assessments. Note that the extension allowed for the inclusion of new baseline visits, as well as additional repeat visits for those already enrolled. Actual recruitment values by site and year are shown in Table 4. Despite the disruptions of COVID-19, our most recent power analysis shows that we are still on track to achieve our study aims (see Power analysis and sample size calculation).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Study timeline and recruitment targets. Numbers represent the targeted number of subjects and assessments, respectively.

https://doi.org/10.1371/journal.pone.0336088.t003

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Recruitment by Site and Year. * - As of 02/20/2026.

https://doi.org/10.1371/journal.pone.0336088.t004

Participants

At baseline, approximately 450 total participants will be enrolled across all sites. Although annual visits are not required for an ALD design, retention is best with close follow-up, and subjects are asked to return annually. We aim to recruit an even distribution of subjects across age groups, with 225 participants in the range of 6–18 years-old and the remaining participants in the range of 19–30 years-old.

Inclusion criteria:

• Between 6–30 years old
• Fluent in English
• Have a biological parent or grandparent who has been genetically or clinically diagnosed with HD, indicating that the participant is at risk for HD (determined from participant or family report)
• Demonstrate an age-appropriate understanding of HD and its risks (If the parents indicate that their child is unaware, we will not enroll the participant until an age-appropriate discussion has occurred)

Exclusion criteria:

• A history of medical diagnoses that could potentially confound the study results (e.g., major head trauma, seizures, tumors, or any other major medical illness requiring ongoing care). If an aspect of their medical history obscures their symptoms as being directly attributable to HD (e.g., a concomitant congenital disorder) they are excluded from participation.
• Have Juvenile-onset HD (JOHD)
• MRI contraindications. Participants of childbearing potential must undergo a urine pregnancy test before their MRI. If positive, they are excluded from the MRI but may still complete other assessments.
• Manifestation of overt motor features of HD. This is ascertained by reporting from parents (or participants themselves if they are over 18) during a screening visit. If the participant or the parent reports that they are currently manifesting any motor symptom of HD, the participant is referred to a child or adult neurologist for clinical evaluation. These individuals are also invited to participate in the JOHD study if they were diagnosed with JOHD before the age of 21 (see below).

Participant recruitment is conducted across the United States. Potential participants along with their parents or grandparents are identified through local hospital or clinic records and HD registries. Recruitment platforms such as Research Match, HD Trial Finder, and ClinicalTrials.gov are used to share study information with the community. Additionally, study-dedicated social media channels and websites are maintained to disseminate updates. The study is also advertised through events, newsletters, social media, and websites of national HD organizations, including the HDSA, its state chapters, Help4HD International, National Youth Alliance (NYA) of HDSA, HDYO, and the Huntington Study Group (HSG).

Power analysis and sample size calculation

Based on our findings from the Kids-HD sample, our central hypothesis is that GE participants will demonstrate altered trajectories in general brain size and cognition across YTO. Specifically, we predicted that GE individuals will demonstrate significantly larger brain volume (primary outcome measure) and stronger cognition scores (secondary outcome measures) prior to 20 YTO and significantly smaller brain volume and weaker cognition between 20−0 YTO. Using our updated recruitment target of 1,030 assessments across 450 participants, we estimated at least 90% power to reproduce our previous findings at p = .01 (except for cortical thickness; Table 5). Note that Table 5, which shows the results of our power analysis, includes functional and structural measures, but that our power analysis is primarily based on the structural findings from Kids-HD. Assuming 40% of participants are GE (as in the previous study), we can assume at least 400 visits from 180 unique genetically affected participants. The power analysis used the Kids-HD sample from Neema et al. (2024), and was based on 136 visits from 79 gene-expanded participants. Because the mean number of visits (and hence, statistical information) per participant will be higher in ChANGE-HD, a conservative estimate of the overall increase in statistical information (i.e., log-likelihood statistics) is obtained from the ratio of unique participants, which is 2.278. If the findings from Kids-HD hold, then the expected values of the analogous F statistics (Table 5) will increase by at least 2.278-fold. Standard power calculations using these projected F statistics yield results that exceed the above-claimed minimum statistical power.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Tests used for hypotheses and power analysis (from Neema et al., 2024).

https://doi.org/10.1371/journal.pone.0336088.t005

Data collection and assessments

Cognitive, behavioral, environmental, and motor measures.

As mentioned previously, the clinical assessment protocol is modeled after the ABCD Study [31], which tracked healthy children (9–10 years old) through adolescence. Although we are studying individuals who are both younger (6–9 years old) and older (25–30 years old), the ABCD study provides the gold standard model for large-scale studies of children and adolescents, which make up a large segment of our population of interest. Table 6 details the full assessment battery, including measures of cognition, behavioral health, environment & health, and motor function.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. ChANGE-HD assessments. MTTC = Maximum time to complete in minutes.

https://doi.org/10.1371/journal.pone.0336088.t006

In the Kids-HD study, cognitive evaluation relied heavily on the Wechsler scales. However, in alignment with new NIH guidelines seeking to improve the inter-rater reliability of multi-site cognitive outcomes, a protocol change was implemented to use the NIH Toolbox (Version 2) [45]. The NIH Toolbox is an iPad-based assessment tool designed to obtain comprehensive metrics from a single visit in the following domains: language, verbal fluency, executive function, attention, working memory, episodic memory, processing speed, and visuospatial intelligence. It offers several advantages, including being digital, standardized, and comprehensive, reducing variability in test results, improving the precision of cognitive measurements, and enabling comparisons across diverse ages and sites. The NIH Toolbox is also validated for use in individuals across the lifespan (ages 3–85; [45,48,59]) and is widely used to study neurological disorders and diseases, such as stroke [60] and Huntington’s disease [61,62], and other clinical populations [63]. Critically, the NIH Toolbox provides a domain-specific score based on all cognitive tests. These standardized cognitive domain scores will serve as our primary metric to assess cognitive function, although we will also explore the effects of gene expansion on each individual test.

Motor functions are assessed using the Unified Huntington’s Disease Rating Scale (UHDRS) motor battery, the NIH Toolbox Nine Hole Peg Test, and the Quantitative Motor (Q-Motor; primary outcome measure for motor performance) assessment. The Q-Motor battery assesses tapping speed and regularity of movements produced by index fingers, hands, and feet. It also quantifies grasping, lifting, and involuntary choreiform movements. The Q-Motor battery is also likely to be more sensitive to pre-manifest changes in motor performance than the UHDRS-TMS [64]. Altogether, it provides a comprehensive and precise evaluation of motor function that has been validated in children [30].

We measure behavioral metrics including participants’ adaptive and maladaptive functions, impulsivity, inhibition/reward seeking, and psychological resiliency using the Achenbach System of Empirically Based Assessment, the Urgency, Premeditation, Perseverance, Sensation Seeking, and Positive Urgency (UPPS-P) impulsive behavior scale, the Behavioral Inhibition/Approach System scale (BIS/BAS), and the Adult/Child-Youth Resilience measure, respectively. Finally, we also record participants’ physical, mental, and social well-being, sleep, the presence of sleep disorders, physical activities, history of traumatic brain injury, and substance use and abuse using Neuro-QoL (Quality of Life) [65], the Sleep Disturbance Scale for Children [55], the Sports and Activities Involvement Questionnaire [56], the Ohio State Traumatic Brain Injury Screen (short version) [57], and a substance abuse and use form (SASUF, made in-house) [58], respectively. All data collected under this protocol will be de-identified and uploaded to the NDA following data collection and analysis.

Magnetic resonance imaging.

Our main goal in designing the ChANGE-HD MRI protocol was to improve on the Kids-HD protocol, giving us the best chance of reproducing previous findings [17] and revealing new ones. To this end, the MRI protocol was designed to collect high resolution structural and functional MRI data within the time span of a typical research MR exam (~1 hour). To harmonize data across scanning sites, the scanning protocol was adapted from the ABCD study [31]. We used 3 Tesla research scanners to collect T1-, T2-, and diffusion-weighted, and resting-state functional MR brain images (see Table 7 for sequence parameter details). These sequences have been adapted and optimized for all six sites across Philips, General Electric, and Siemens MR scanner platforms.

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. ChANGE-HD harmonized imaging scanning parameters for 3T scanners. FOV = field of view, TR = repetition time, TE = echo time, TI = inversion time, MP = MPRAGE, GR = spoiled gradient echo. VUMC = Vanderbilt University Medical Center.

https://doi.org/10.1371/journal.pone.0336088.t007

As in the ABCD study, all images across sites were acquired in the axial plane because GE and Philips scanners may cause ghosting artifacts when acquired at oblique angles. Several improvements have been made to our MRI protocol since the Kids-HD study. While Kids-HD involved a single Siemens scanner at the University of Iowa, ChANGE-HD involves a mix of three vendors (Iowa: General Electric, CHOP: Siemens, Columbia: General Electric, UT Houston: Philips, UC Davis, Siemens, Vanderbilt: Philips). T1- and T2-weighted scan resolution has been increased by decreasing voxel sizes from 1.1 mm (T1) and 1.0x1.4x1.0 mm (T2) to 1.0 mm isotropic. To improve signal-to-noise ratio (SNR), we also increased the head coil channels from 12 in Kids-HD to between 32 and 48 (depending on scanner model) in ChANGE-HD.

Several improvements to resting-state functional scans were made. First, instead of acquiring one five-minute scan, we collected two scans (10 minutes total), each with opposite phase encoding directions. These improvements were made to reduce the effects of spurious noise and improve SNR [66], while reducing susceptibility effects [67]. Voxel and temporal resolution were also improved by decreasing voxel size from 3.4x3.4x4.0 mm to at least 2.4 mm isotropic and decreasing the TR from 2500 ms to 800 ms, respectively.

Finally, diffusion scanning was improved by decreasing voxel sizes from 2.0 to 1.7 mm isotropic and reducing distortion potential by lowering the echo time from 81 ms in Kids-HD to 69 ms in ChANGE-HD. Thus, the improvements in hardware and software represent a significant technological improvement between studies. Although scanning times differed by study site and participant visit, the approximate active scanning time was ~ 1 hour, including ~16 minutes for structural T1/T2 scans, ~ 18 minutes for diffusion scans, ~ 22 minutes for resting state scans, and ~2 minutes for BOLD field maps. Participants were given opportunities to take breaks and leave the bore, if needed.

During data collection periods, several quality control measures were followed to optimize scan quality. First, prior to study initiation, and once per year, each site was responsible for human phantom calibration. These calibrations were performed by designated credentialed imaging experts at each site (see S1 Table for each imaging expert). Second, each site’s imaging expert and team were required to regularly check scans for signs of poor scan quality, such as low tSNR or high resting state scan motion. In some cases, scans were repeated to ensure high image quality (e.g., the appearance of participant motion during resting state scanning). Third, deviations in protocol or changes to the study MRI system were noted by each site’s MR physicist. Post-data collection metrics of quality control will be reviewed to ensure that our final scans include only those with SNR of above 20, 45, 25, and 10 for anatomical, resting-state, b0, and b1000 diffusion scans, respectively. Resting state functional scans will be rejected if average framewise displacement exceeds 0.3 mm and frames will be censored if motion exceeds 0.3 mm.

All raw (uncensored, uncorrected, etc.) neuroimaging data collected under this protocol will be de-identified and uploaded to the NDA following data collection and analysis. Structural MRIs will be defaced to prevent participant identification.

Data analysis

Statistical analysis using YTO Models.

We will employ a YTO model, which provides a more detailed representation of disease-related changes (between 50 and 0 YTO) than an age-based analysis [15,16,18,23]. Our statistical analyses are derived from Langbehn et al. [80] and our own work [17] investigating neural and behavioral changes across YTO. This method is based primarily on data from individuals with CAG repeats between 41 and 56 [80] but extrapolation has generalized usefully to CAG repeats extending beyond this range (40, 57–70, etc.) [17]. Because individuals with between 36–39 CAG repeats represent an incomplete penetrance group, they will be excluded from the YTO analysis and will be examined separately in exploratory analyses. Individuals with >70 CAG repeats will also be excluded as they are likely to have already begun showing motor symptoms of HD and are likely classified as patients with JOHD. All hypotheses relate to either systematic group difference between GE and GNE individuals or to differences within the GE group driven by CAG expansion length and age (or years to onset as in the NfL analysis). Our analysis approach will include the following steps: 1) age and sex correction based on the data from healthy controls, 2), estimation of YTO, and 3) YTO modeling with testing for potential additional sex differences and sex by YTO interactions.

For each dependent variable, we apply statistical corrections to separate developmental effects of typical aging from those experienced by GE individuals as part of the disease trajectory. We perform this correction by modeling age and sex effects in GNE controls using a regression that includes within-subject and within-family random effects. The models incorporate both linear and potential non-linear age relationships. Nonlinearity is modeled flexibly via restricted cubic splines (“natural splines”). The degree of non-linearity (up to five degrees of freedom) is determined by choosing the model with the lowest Akaike Information Criterion (AIC), with tied AIC values (≤ 1.0) favoring the simpler model. We also test the model for sex and sex-by-age interactions, which are excluded from the final model if found to be nonsignificant. The fitted values for this model are then subtracted from the dependent value to complete the correction.

Estimation of age of motor symptom onset is calculated using the Langbehn model [80] shown below:

YTO is then calculated by simply subtracting participants’ ages (in years) from age of motor symptom onset. (The abbreviation YTO, now established by precedent, stands for a slight misnomer. The value is the number of years from the mean age of onset for those with the given CAG length. It is not the individual’s number of years to onset—which is unknowable—and is slightly different from the individual’s expected years to onset, given that they have reached their current age without already experiencing onset.)

A limitation of this YTO approach is that it can only approximate the relationship between true years to HD onset and the outcome measures. As noted earlier, the YTO calculations are slightly biased estimates of participants’ expected years to onset. Furthermore, a person’s true years to onset has a probability distribution around the CAG-specific average. For concave relationships, such as those we typically expect in the neurodevelopment of brain volumes, substitution of an average value for a probability distribution causes at least slight underestimation of the predicted value that would be obtained from a model based on the true (but unknowable) number of years from onset (Jensen’s inequality).

We assess the relationship between YTO and age- and sex-adjusted outcome measures in GE individuals using an additional mixed effects regression, again including random effects for repeated measures of the same participant and for similarities between participants from the same family. Similar to the GNE model described above, the potentially nonlinear influence of YTO is modeled using restricted cubic splines with degrees-of-freedom chosen by AIC. Sex and sex interactions with YTO are tested and included in the model if significant. The mixed effect models are fitted by maximum likelihood and F-test degrees of freedom are estimated using the Satterthwaite approximation. The summary of the resulting model indicates a p-value for the effect of YTO, which, if significant, indicates a statistically meaningful difference of GE from GNE along the YTO trajectory. Finally, we plot these relationships and their 95% confidence intervals to determine where these differences occur along the trajectory (50−0 YTO)

To assess the relationship between age- and sex-adjusted neuroimaging variables (e.g., cortical grey and white matter volume, striatal volume, corticostriatal structural and functional connectivity, etc.) and clinical outcomes in GE individuals, we apply modeling techniques as described above. The models include random effects of subject and family, but exclude age and sex, as the values will have already been corrected using the data from GNE participants. Imaging measures are generally predictor variables, and clinical measures are the outcomes, reflecting the potential causal relationships between the two measurement classes. As in our previous work [17], we also split our GE participants into two groups with a cutoff of 20 YTO and test for differences in the imaging versus clinical relationships. This is based empirically on our work [17] showing divergence in neural and behavioral measures at this point in the YTO trajectory and because of work showing an upward trend in NfL at 20 YTO [18]. As described above, potential non-linear relationships between biomarkers and clinical measures are interrogated using restricted cubic splines (up to five degrees of freedom). For all analyses, a p value of < 0.05 is considered statistically significant.

In addition to our hypothesis-driven analyses, we also plan to perform detailed exploratory analyses to uncover novel neuroimaging biomarkers that both distinguish GE and GNE individuals and individuals with variations in HD-related single nucleotide polymorphisms (SNPs). For example, machine learning approaches, such as deep learning, have been applied to neuroimaging data from patients with neurodegenerative disorders to identify how those SNPs affect brain anatomy and function [81]. As mentioned above, SNPs in several genes explain variation in the eventual onset of motor symptoms in HD [32–34]. Using deep learning and other machine learning approaches as exploratory tools may reveal new mechanistic information about how these SNPs shape neurodevelopment and -degeneration.

The ChANGE-Juvenile Onset HD (ChANGE-JOHD) Study

As part of the same study and funding mechanism, we are also conducting the “ChANGE-Juvenile Onset HD (ChANGE-JOHD) Study,” which will leverage the same infrastructure and multi-site design as the ChANGE-HD study. Although most HD patients typically manifest obvious motor signs between the ages of 40 and 50 (known as Adult-Onset HD, AOHD), a much smaller percentage of patients are diagnosed with motor features before the age of 21 (classified as JOHD) [27]. Patients with JOHD experience the same triad of cognitive, behavioral, and motor symptoms as those with AOHD, albeit with distinct phenotypes [82]. Specifically, JOHD patients tend to exhibit less hyperkinesia and more hypokinesia compared to AOHD. However, patients with JOHD experience CAG repeat expansion and symptom progression at a faster rate compared to patients with AOHD [83]. The progression of JOHD is theorized to involve several neuropathologic mechanisms, including reduced gestational neurogenesis and cell differentiation, neuronal migration, and synaptogenesis, as well as hyper and hypo-synaptic pruning during puberty [84]. However, due to its rarity, many aspects of the neurobiology and progression of JOHD are unclear.

While large-scale observational studies of AOHD have significantly advanced our understanding of HD, JOHD patients have been largely excluded from these efforts [85]. As such, large-scale longitudinal studies in JOHD are needed to deepen our understanding of the pathophysiology, as well as therapeutic strategies. When the Kids-HD study began at the University of Iowa in 2009, children at risk for HD who were already manifesting motor signs were followed as well. This was called the Kids-JOHD study (which ran in parallel to the Kids-HD study), whose primary objective was to deepen our understanding of the course and progression of JOHD. Cross-sectional analyses showed that patients with JOHD have significant subcortical degeneration shortly after diagnosis, along with reduced general measures of growth (such as weight and BMI) compared to GNE controls [21,86,87]. Furthermore, the longitudinal trajectory of disease severity is markedly hastened and amplified in JOHD compared to AOHD [88]. Despite these initial assessments from the Kids-JOHD study, larger-scale longitudinal studies in JOHD are still lacking. To this end, we have developed the first ever multi-site, prospective and comprehensive study of JOHD, leveraging the established infrastructure of the titular ChANGE-HD study to record clinical and neuroimaging metrics in this population. This study, which is considered separate from the ChANGE-HD but also follows an ALD, aims to investigate whether comprehensive longitudinal quantitative assessments of motor skills, cognition, and neuroimaging in JOHD patients can yield reliable biomarkers of disease progression. A total of 40 patients with JOHD will be recruited. In the first year, 20 patients with JOHD (3–4 per site) will complete baseline assessments, with the remaining 20 patients undergoing baseline assessments in Year 2. Year 2 will also include annual follow-up visits for the first 20 patients. In Year 3, follow-up visits will be conducted for the second group of 20 patients. As of the time of this writing, we have collected 49 assessments from 34 participants. For the study, the comparator group will include GNE participants recruited in the ChANGE-HD protocol. Data and biospecimens collected from the ChANGE-JOHD study will be shared separately from the ChANGE-HD study after data collection and analysis of primary outcome measures are complete, which we estimate will occur in September of 2027. To participate, JOHD participants must:

• Be diagnosed with JOHD by a trained neurologist prior to the age of 21, based on motor symptomology (the participant’s genetic status is confirmed after consent is obtained).
• Be between 6–30 years old at baseline. Thus, even though the diagnosis is made prior to 21, assessment for the study is not restricted to ages under 21.
• Not be bed-ridden, unable to communicate, or beyond travel conditions that are too strenuous for the individual.