The timing and impact of psychiatric, cognitive and motor manifestations of Huntington’s disease

Objective To assess the prevalence, timing and functional impact of psychiatric, cognitive and motor manifestations of Huntington’s disease (HD), we analysed retrospective clinical data from individuals with manifest HD. Methods The clinical manifestations of HD were analysed for 6316 individuals in the European REGISTRY study from 161 sites across 17 countries. Data came from clinical history and the Clinical Characteristics Questionnaire that assessed eight symptoms: motor, cognitive, apathy, depression, perseverative/obsessive behavior, irritability, violent/aggressive behavior, and psychosis. Multiple logistic regression was used to analyse relationships between symptoms and functional outcomes. Results The initial manifestation of HD is increasingly likely to be motor, and less likely to be psychiatric, as age at presentation increases. The nature of the first manifestation is not associated with pathogenic CAG repeat length. Symptom prevalence data from the patient-completed Clinical Characteristics Questionnaire correlate specifically with validated clinical measures. Using these data, we show that psychiatric and cognitive symptoms are common in HD, with earlier onsets associated with longer CAG repeats. 42.4% of HD patients reported at least one psychiatric or cognitive symptom before motor onset, with depression most common. Apathy and cognitive impairment tend to come later in the disease course. Each psychiatric or cognitive manifestation of HD was associated with significantly reduced total functional capacity scores. Conclusions Psychiatric and cognitive symptoms occur before motor onset in many more HD patients than previously reported. They have a greater negative impact on daily life than involuntary movements and should be specifically targeted with clinical outcome measures and treatments.


Introduction
Huntington's disease (HD) is a central neurodegenerative disorder caused by an expanded CAG repeat (>35 CAGs) in exon 1 of the Huntingtin gene (HTT) 1 . The length of the expanded CAG tract is inversely correlated with the age at first clinical manifestation of HD 2,3 . Degeneration is most prominent in the medium spiny neurons of the striatum but occurs widely in the brain, leading to a progressive movement disorder, cognitive decline and ultimately death. The motor disorder usually includes chorea, and may also involve dystonia, coordination problems, oculomotor abnormalities and parkinsonism. In addition, individuals with HD develop a variable constellation of debilitating behavioural and psychiatric symptoms alongside their motor problems 4,5 : these 'non-motor' symptoms remain under-recognised and undertreated. Prospective studies of individuals carrying a pathogenic HTT CAG repeat but many years from predicted clinical onset have shown they have few significant motor, cognitive or psychiatric deficits when compared with age and sex-matched controls [6][7][8] . This implies there is a window for therapeutic intervention to preserve normal brain functions. It is therefore crucial to understand in detail the timing and impact of different manifestations of HD in order to target and assess potential disease-modifying therapies. Here we analyse in a large population the prevalence and timing of psychiatric, cognitive and motor manifestations of HD, their relationships with CAG repeat length, and their impact on day-to-day life.
The age at onset of the first clinical manifestation of HD, often difficult for patients, care partners and clinicians to define precisely, has been used as a specific milestone in the natural history of HD in individuals. The age at onset of chorea and/or other motor features ('motor onset') has proven particularly useful in genetic modifier studies of HD 9,10 , although it is only a crude measure of a progressive neuropathological process. The timing of HD clinical onset is typically recorded by a rating physician in observational studies such as the European REGISTRY, based on clinical information and symptom history from patients and care partners 11-13 . This retrospective assessment cannot pick up the subtle earliest motor and nonmotor signs, which are unnoticed by patients but detectable by HD specialists in prospective studies 7 . The rater also records the initial major presenting symptom out of a choice of six: motor, cognitive, psychiatric, oculomotor, other or mixed. Given the complexity of the HD phenotype across multiple domains, later versions of REGISTRY incorporated the HD Clinical Characteristics Questionnaire (HD-CCQ) 11 , asking individuals whether they had ever experienced a range of motor and psychiatric symptoms, and at what age those symptoms started 14 . The eight symptom groups recorded were motor (any type), depression, irritability, violent or aggressive behaviour, apathy, perseverative/obsessive behaviour, psychosis and cognitive impairment sufficient to impact on work or daily living. Importantly, individuals with HD and their care partners, present in clinic for over 93%, reported the presence or absence of these symptoms independent of whether they considered them directly related to HD or not. These data on symptom onsets were collected alongside multiple quantitative assessments of HD phenotype, including the Unified Huntington's Disease Rating Scale (UHDRS) 15 , Hospital Anxiety/Depression Rating Scale (HADS) and the Snaith Irritability Scale (SIS). The UHDRS consists of a series of validated questionnaires, tools and examinations related to motor, cognitive, behavioural and functional impairments seen in HD.
Here we find that motor presentation of HD is increasingly common with later clinical onset but there is no relationship between CAG length and type of presentation. We validate the use of HD-CCQ data in a large HD population of over 6000 participants by showing strong and specific associations with validated scores of depression, irritability and cognition. We then use HD-CCQ data to show the high prevalence of psychiatric and cognitive symptoms in HD, often in advance of motor onset, and how they impact negatively on the daily lives of patients.

Standard protocol approvals, registrations and patient consents
Participants were in the multicentre, multinational, observational REGISTRY study of and 1289 were derived by local diagnostic laboratories. Two estimates of the age at clinical onset of HD were used in this study. First, the clinician-estimated age at first HD manifestation based upon all available clinical evidence at the first REGISTRY visit (coded as 'sxrater'). Having a 'sxrater' age at onset was required for inclusion in this study. Onset type was classified as motor, cognitive, psychiatric, oculomotor, other or mixed. As the clinician's estimate was given as a date, age estimates were calculated using the participant's anonymised birthday; where only a year was given, July 15 th was used for estimation (15/07/xxxx). Second, the ages at onset of different symptoms were determined from the HD Clinical Characteristics Questionnaire (HD-CCQ) which was completed by a healthcare professional, usually a HD-specialist nurse or similarly-qualified person, using responses from the individual with HD and their care partners (present in clinic in 93.1% of cases). The HD-CCQ comprises questions about eight symptoms commonly observed in HD, asking whether the participant has ever had the symptom (yes or no), and, if yes, the age at which the symptom was first experienced (Appendix 3). Information was available, at least in part, for 5609 individuals. The symptoms recorded (number of individuals with data) were: motor (5603), cognitive impairment sufficient to impact on work or daily living (5591), apathy (5584), depression (5595), perseverative/obsessive behaviour (5588), irritability (5586), violent or aggressive behaviour (5586) and psychosis (5589). For subsequent analyses, missing data were handled using pairwise deletion to maximise the number of individuals. Typically, the rater estimate of onset and initial HD-CCQ would be recorded at the first REGISTRY visit, sometimes by one clinician, and sometimes by a clinician and another qualified staff member such as HD-specialist nurse, depending on local clinic set-up. Subsequent visits updated the HD-CCQ: we used data from the most recent clinic visit.
The Hospital Anxiety/Depression Scale (HADS) and Snaith Irritability Scale (SIS) were completed by the participant at each clinic visit and provide measures of anxiety, depression and irritability at that specific time. We used lifetime highest total depression and total irritability scores from both the HADS and the SIS in analyses.
Similarly, the symbol-digit modalities test (SDMT) and Stroop tests of cognitive ability were administered as part of the UHDRS at each clinic visit. For the SDMT and Stroop, we used the total correct scores from the most recent clinic visit. Disease duration was estimated by taking the most recent visit and subtracting the clinician's estimate of disease onset. The product of Problem Behaviours Assessment (PBA-s) severity and frequency scores from the most recent clinic was used for modelling purposes.

Statistical analyses of clinical data
Total depression scores from HADS, total irritability scores from SIS, the number of correct answers in the SDMT, the number of correct answers in Stroop tests or composite PBA-s scores were regressed on HD clinical characteristics data, age, CAG length, sex and disease duration (table 1). To calculate coefficients of determination (R 2 values, table 2), HD-CCQ age at onset data were natural log transformed. Only individuals with a known sex and a symptom onset ≥3 years were considered, and a residual vs leverage plot identified one influential data point passing Cook's distance that was removed from all R 2 calculations. P values were calculated comparing male and female R 2 values using Fisher's transformation 16 . A chi-square test was used to test for differences in symptom frequency, derived from the yes/no component of the HD-CCQ, between males and females.
Associations between binary responses in the HD-CCQ (1; experienced the symptom and 0; symptom not experienced) and clinical covariates were tested using logistic regression. The covariates used were sex, CAG length, alcohol consumption (units per week), tobacco use (cigarettes per day), education (years of education), total functional capacity (TFC) score and total motor score (TMS). An additional analysis regressed the type of HD onset defined by the clinician, coded as a binary variable, on the clinician's onset or CAG length (table e-2, doi:10.5061/dryad.pk0p2ngkz). This analysis was restricted to HD participants with CAGs 36-59, to be consistent with figure 1 subgroups, and also to adult-onset HD individuals (≥20 years). We also tested whether symptom presence was associated with the length of the wild-type (6-35 CAGs) and expanded CAGs  alleles in individuals of known sex, and for whom both CAG lengths were known (table e-3, doi:10.5061/dryad.pk0p2ngkz). 19 individuals with a coincident formal diagnosis of schizophrenia, schizotypal disorder or schizoaffective disorder (ICD-10 F20, F21 or F25) were excluded from all models. It was not possible to formally exclude these symptoms being part of the HD phenotype in these few individuals.

Data availability
Further information and requests for data should be directed to Thomas H. Massey MA, BM BCh, DPhil (MasseyT1@cardiff.ac.uk). Anonymised summary data is available on request by qualified investigators. Furthermore, anonymised patient data is available from the European Huntington's Disease Network (EHDN) upon request given their institutional assurance patient confidentiality will be upheld, and no attempt will be made to discover the identity of patients.

The initial manifestation of HD varies with age and CAG length
We analysed the initial manifestation of HD for 6316 participants in REGISTRY 11 , including 3083 males (48.8%) and 3233 females (51.2%). All participants had a confirmed genetic diagnosis of HD with a pathogenic CAG repeat length of  (figure e-1, doi:10.5061/dryad.pk0p2ngkz). The first symptom or sign of HD, determined by the rating physician, varied with the age at which HD became clinically manifest (figure 1A and table e-1, doi:10.5061/dryad.pk0p2ngkz).
Individuals with onset before the age of 20, defined as juvenile HD, were equally likely to present with motor (24.5%), cognitive (21.8%) or psychiatric features (28.2%). In contrast, the initial HD-related symptom or sign was more likely to be motor than psychiatric in adult-onset HD. As age at first manifestation increased (figure 1A and table e-2A, doi:10.5061/dryad.pk0p2ngkz) motor presentations became more likely (odds ratio (OR) = 1.06 per ten year increase in onset age, 95% confidence interval (CI) 1.04-1.07; P = 7.4 x 10 -22 ) but psychiatric presentations became less likely (OR = 0.96 per ten year increase in onset age, 95% CI 0.95-0.97; P = 9.4 x 10 -16 ). For people presenting over the age of 60, over two-thirds (68.6%) had motor manifestations at clinical onset with far fewer having psychiatric (11.5%) or cognitive (6.7%) presentations. Next, we tested whether there was any relationship between pathogenic CAG repeat length, known to be inversely correlated with age at clinical onset, and the presenting phenotype. Interestingly, there was no significant relationship between CAG length (36-59 inclusive) and the relative proportions of motor, cognitive and psychiatric onset cases (figure 1B and table e-2B, doi:10.5061/dryad.pk0p2ngkz). For the few cases with data and repeat lengths of more than 59 CAG we observed a more balanced distribution of motor, cognitive and psychiatric onsets, mirroring the trends seen for the juvenile HD cases.

Questionnaire correlate with scores from validated clinical tools
The HD Clinical Characteristics Questionnaire (HD-CCQ) is completed directly by individuals with HD and their care partners about lifetime history and age at onset of eight motor, psychiatric and cognitive symptoms. In REGISTRY, this information was updated at each annual clinic visit. Since prevalence data from HD-CCQ have not been used in large analyses before we first tested how well they correlated with data from validated measures of depression (HADS), irritability (SIS) and cognition (SDMT and Stroop). To mitigate against potential effects of medication at certain times, we used the lifetime highest total depression and total irritability scores for each individual. For cognitive tests we used scores at the last recorded clinic visit as these would be expected to worsen progressively and be little affected by medication. Total depression score from HADS was significantly increased in individuals with depression recorded in HD-CCQ (table 1; increase of 1.49 units, 95% CI 1.09-1.89; P = 5.7 x 10 -13 ). An increase in HADS score was also observed in individuals with HD-CCQ apathy, probably because apathy, common in HD, may be mistaken for depression by individuals and their care partners when completing the HD-CCQ. Total irritability score from SIS was significantly increased in individuals with HD-CCQ irritability (increase of 1.82 units, 95% CI 1.37-2.27; P = 2.0 x 10 -15 ), and also with violent/aggressive behaviour (increase of 1.57 units, 95% CI 1.08-2.06; P = 3.3 x 10 -10 ), as expected. Both SDMT and Stroop scores of cognitive ability were significantly decreased in individuals with cognitive impairment as recorded in HD-CCQ (reductions of 3.52 units, 95% CI 2.71-4.33; P = 2.3 x 10 -17 and 3.41 units, 95% CI 2-65-4.17; P = 1.4 x 10 -22 , respectively). Significant associations between cognitive scores and motor and apathy symptoms were also observed. In addition, we found robust and specific associations between neuropsychiatric symptoms recorded in HD-CCQ and their related symptoms scored using the validated shortform Problem Behaviors Assessment (PBA-s; supplemental table e-4, doi:10.5061/dryad.pk0p2ngkz). The specificity of the associations between HD-CCQ data and recognised clinical scales validated the use of HD-CCQ data in subsequent analyses.
Psychiatric symptoms are common in HD and are associated with CAG repeat length We next analysed the lifetime prevalence of the eight symptoms recorded in HD-CCQ in 5609 individuals with HD at their most recent clinic visit ( Although these symptoms are not recorded explicitly in HD-CCQ, contemporaneous data from UHDRS showed that 96.8% of our study population had chorea, alongside variable amounts of incoordination, dystonia and rigidity. The next most prevalent symptom was depression, occurring in 64.5% of HD individuals with significantly more females affected than males (70.4% vs 58.2%; OR 1.70, 95% CI 1.52-1.90; P = 2.6 x 10 -21 ). Cognitive impairment sufficient to impact upon work or activities of daily living, apathy and irritability were also each observed in over half of our HD population. Cognitive impairment and apathy were equally likely in males and females, but there was significantly more irritability observed in males (62.9% vs 56.9%; OR 0.78, 95% CI 0.70-0.87; P = 4.0 x 10 -6 ). An excess of violent or aggressive behaviour was also observed in the male group (34.9% vs 27.0%; OR 0.69, 95% CI 0.62-0.77; P = 2.0 x 10 -10 ). Psychosis was the least prevalent of the eight recorded symptoms, although this was still observed in over 11% of individuals with HD with no significant difference in prevalence between males and females.
There was a strong inverse correlation between pathogenic CAG repeat length (40-55 CAG inclusive) and mean age at symptom onset for all symptoms analysed (figure 2). We found no effect of wild-type CAG allele length on any symptom onset, nor any significant statistical interaction between expanded and wild-type repeat lengths (table e-3, doi:10.5061/dryad.pk0p2ngkz). Pathogenic CAG length explained 66.3% of the variance in age at onset of motor symptoms, in line with previous estimates 2,3,17-23 , but also between 37.5% and 61.9% of the variance in onset of each of the psychiatric symptoms analysed (table 2). Depression had the weakest association with CAG repeat length (R 2 = 37.5%), likely reflecting the high prevalence of the symptom in the general population independent of HD and the lack of use of universal diagnostic criteria. CAG length accounted for significantly more of the variance in age at onset of perseverative/obsessive behaviour in males (table 2; P = 3.7 x 10 -3 ) and irritability in females (P = 1.3 x 10 -3 ).

The timing of motor and psychiatric symptoms in HD varies with symptom type and CAG length
Given that motor onset is often used as a specific milestone in the natural history of HD, we investigated the timing of each of the seven psychiatric/cognitive symptoms relative to the age at which motor symptoms first occurred (figure 2). The difference in age between motor onset and each of the psychiatric symptom onsets was approximately normally distributed, with a wide range of at least +/-20 years in each case (figures 2 and e-2, doi:10.5061/dryad.pk0p2ngkz). In those patients reporting depression, onset occurred before motor symptoms in 39.2% (N= 1369/3495). For patients with irritability, onset occurred before motor symptoms in 30.8% (N= 996/3235). Perseverative/obsessive behaviour tended to occur later in the disease course, after motor onset, as did psychosis although numbers were smaller.

Discussion
In this large study of over 6000 patients we have shown that the initial clinical presentation of HD varies significantly with age: late onset (>60 years) is usually associated with motor manifestations, whereas early onset (<20 years; juvenile HD) is associated with a wider range of motor, cognitive and psychiatric presentations (figure 1A). These results extend prior studies which have shown that motor presentation of HD is common in late-onset disease (65.5% of an earlier REGISTRY cohort 24 ), with more variable presentations in juvenile HD 25,26 . Approximately 20% of HD patients present with rater-determined psychiatric features, in line with previous findings (table e-1, doi:10.5061/dryad.pk0p2ngkz) 11 . Cognitive onset of HD might be under-reported in older age-groups due to it being regarded as coincident 'agerelated' change. Importantly, our results show there is little relationship between pathogenic CAG repeat length and onset type in adult-onset HD (figure 1B), despite both being associated with age at clinical onset. These data fit a model in which age at clinical onset is driven primarily by CAG repeat length, but modified by environmental factors and variants at other genomic loci 10,23,27,28 . The age and physiology of the brain at clinical onset subsequently determines the types of symptoms that become manifest.
Previous analysis of a smaller cohort showed that the expanded CAG repeat is fully dominant for motor onset of HD, with no effect of the normal HTT CAG allele 17 . We confirmed this result and showed that the same is true for all psychiatric and cognitive onsets (table e-3, doi:10.5061/dryad.pk0p2ngkz). The age at onset of each symptom recorded by HD-CCQ was inversely correlated with CAG length (figure 3), with motor symptoms best correlated and depression least correlated (table 3).
These data are consistent with previous reports showing that CAG length accounts for 47-72% of the variance in age at motor onset of HD 29 , but do not fit with previous studies that reported no correlation between CAG repeat length and psychiatric symptoms [30][31][32][33] . However, these studies were small and often examined incident psychiatric symptoms, which can fluctuate over time, rather than lifetime history as here. In addition, psychiatric manifestations might have been less easily defined as HD-related than motor features. Increasingly accurate CAG tract sizing will improve the accuracy of correlations between repeat length and symptoms 10,34,35 .
Most modifier studies of HD have used age at motor onset as a quantitative trait.
Motor signs (usually chorea) in the context of an expanded CAG repeat in HTT have high specificity for HD and most individuals with manifest HD report motor symptoms. However, psychiatric and cognitive symptoms are often more debilitating for patients and can occur early in the disease course 36 . These symptoms are common in HD (table 2), and much more prevalent than in the non-HD population 5,14,37 . Neuropsychiatric symptoms are also likely underestimated in HD due to pathological unawareness of these traits by patients 38 . Recent genetic evidence has shown that polygenic risk scores for psychiatric disorders, particularly depression and schizophrenia, are associated with increased risk of corresponding psychiatric symptoms in HD 39 . This suggests that the expanded HTT CAG repeat might lower the genetic threshold for manifestation of typical psychiatric symptoms 39 .
In agreement, we found the expected relationships between female sex and depression and male sex and irritability in our cohort (table 4). Our finding that psychosis in HD had a nominally significant negative association with educational level (table 4) also corroborates work showing that higher levels of education are associated with decreased schizophrenia risk 40 .
The HD-CCQ captures quantitative information not available elsewhere on symptom prevalence and timing in the HD population. It provides particular insight into neuropsychiatric symptoms but is not designed to capture subtle early motor signs.
The specificity of recorded HD-CCQ scores is shown by their correlation with validated scores of depression, irritability and cognition (table 1) showing that these symptoms track together in the disease trajectory 39,41 . There was also a significant association between cognitive impairment and psychosis, which fits the cognitive deficits observed in schizophrenia 42 . Conversely, validated depression and irritability scores correlated well with their respective prevalence data from HD-CCQ but were not associated with motor or cognitive impairment (table 1).
Despite considerable variation in the timing of psychiatric symptoms relative to motor onset, there are some conserved patterns (figures 2 & 3). Depression, and less often irritability, can precede motor onset by many years. Conversely, apathy and cognitive impairment tend to occur after motor onset. Overall, the HD-CCQ data show that 64.8% of our HD population (N = 3266/5042) reported at least one psychiatric or cognitive symptom by the time of motor onset. This is a much higher figure than previously reported based on clinician estimates of first HD symptom (figure 1) 11 . The discrepancy probably arises because psychiatric symptoms in 'premanifest' individuals carrying the HD mutation are often attributed by clinicians to coincident diagnoses rather than to HD itself. This is particularly true of depression which is common in young adults with or without HD 43 . Clinically, it is impossible to distinguish between symptoms arising as a result of the HD mutation and those from primary psychiatric disorders, particularly in younger pre-manifest patients who are potentially years from motor onset. Furthermore, environmental effects on mental health, such as living in a family with HD, should not be overlooked.
We acknowledge several potential biases in these data: they are retrospective, subject to recall bias, and cross-sectional. HD-CCQ data depend on patient and care partner interpretation of questions. The analyses are based on data from the most recent clinic visit which is at different points of the disease course in different individuals. We controlled for this by using disease duration, the time between first onset and last clinic visit, as a covariate in analyses. The use of psychoactive medications is found in up to 60% of HD patients and might confound motor and neuropsychiatric phenotypes 11,44 . Of drugs prescribed for chorea, tetrabenazine can induce depression and antipsychotics can reduce irritability. They also suppress motor manifestations which might affect the total motor scores used here as a covariate (table 4). It is hard to control for these effects. Drugs prescribed to treat symptoms once they are present will not influence symptom onset data. We used worst-ever depression and irritability scores when validating the use of HD-CCQ to mitigate against the effects of medication prescribed at certain times.
It is not feasible to conduct detailed prospective studies of symptom onsets over many years in large HD populations and so data from HD-CCQ are invaluable. Since the HD-CCQ is part of ongoing longitudinal observational studies such as a ENROLL-HD, future analyses of larger populations will be possible and of benefit.
The presence of psychiatric and cognitive symptoms in HD patients is associated with significantly reduced functional capacity, emphasising the importance of managing these symptoms that have deleterious effects on quality of life 45,46 . It is notable that recent models of HD staging and progression do not directly include these psychiatric and cognitive symptoms 47-49 . Work is underway to include detailed psychiatric and cognitive ratings in ongoing observational studies and clinical trials to improve the accuracy of clinical outcome measures. Given the prevalence, timing and impact of psychiatric and cognitive manifestations of HD demonstrated here, their treatment must form a central part of assessing the efficacy of novel treatments that are potentially disease-modifying, such as Huntingtin lowering 50 .

Acknowledgments
We thank all the patients who contributed data to this research. B.         Table e-2. Age at onset, but not CAG length, is associated with initial manifestation type in adult-onset HD. (A) Logistic regression of initial manifestation as recorded by the clinician, coded as binary variables (1: initial manifestation of that type, 0: initial manifestation of any other type) on age at onset. (B) Logistic regression of initial manifestation (defined as a binary variable, as above) on CAG length. Significant associations after Bonferroni correction for 12 models are shown in bold (P < 4.17 x 10 -3 ) and nominally significant associations in italics (P < 0.05). The odds ratio (OR) indicates the effect on the outcome probability associated with an increase of ten years in age. Only individuals with a confirmed onset ≥20 years, CAG 36-59 and no schizophrenia co-morbidity (ICD-10 F20, F21 or F25) were included. N=6051 for both sets of models. CI: confidence interval.  The CAG repeat length of the pathogenic HTT allele, but not the wild-type HTT allele, is associated with the presence of motor and psychiatric symptoms in HD. Logistic regression of binary symptom data (lifetime prevalence; 0 = no symptom, 1 = reported symptom) on expanded (36-93 CAGs) and wild-type (6-35 CAGs) repeat lengths, with an interaction term. Significant associations are in bold (P < 2.08 x 10 -3 , correcting for number of covariates and phenotypes), and nominal associations italicised (P < 0.05). The odds ratio (OR) indicates the effect on the outcome probability associated with an increase of one unit in the covariate. In addition to having a confirmed onset and pathogenic CAG length, to be included individuals must have a known sex and no co-morbid diagnosis of schizophrenia, schizotypy or schizoaffective disorder. CI: confidence interval; POB: perseverative/obsessive behaviour; VAB: violent or aggressive behaviour. Table e-4. Scores for neuropsychiatric symptoms from the short-form Problem Behaviors Assessment questionnaire (PBA-s) correlate with specific responses from the HD clinical characteristics questionnaire (HD-CCQ). Generalised linear models regressed the product of the frequency and severity scores from PBA-s questions against binary lifetime prevalence data on 8 symptoms from the HD Clinical Characteristics Questionnaire (italics, column 1) and other covariates. For binary covariates (CCQ symptoms and sex) "effect" is the increase/decrease in the clinical score associated with presence of that covariate. For quantitative covariates (age, CAG, duration), "effect" is the change in clinical score associated with an increase of one unit in the covariate. In addition to having a confirmed onset and pathogenic CAG length (36-93), individuals must have no co-morbid diagnosis of schizophrenia, schizotypy or schizoaffective disorder. Significant associations after Bonferroni correction for 11 phenotypes and 12 covariates are shown in bold (P < 3.79 x 10 -4 ) and nominally significant P values are italicised (P < 0.05). CI: confidence interval; POB: perseverative/ obsessive behaviour; VAB: violent or aggressive behaviour; Sex (F): female.