Longitudinal progression, metrics, age-dependence, and modifiers of ataxia severity in SCA27B: a multicentre study of 219 patients Authors: Andreas Traschütz, Ralf-Dieter Hilgers, Friedrich Erdlenbruch, Christel Depienne, Thomas Wirth, Clarisse Delvallée, Astrid Nümann, Catherine Ashton, David Pellerin, Elisabetta Indelicato, Felix Heindl, Mathilde Renaud, Max Borsche, Marcus Grobe-Einsler, Jennifer Faber, Thomas Klockgether, Ludger Schöls, Bernard Brais, Mathieu Anheim, Dagmar Timmann, Matthis Synofzik Published: 2026 ABSTRACT SummaryBackgroundSpinocerebellar Ataxia 27B (SCA27B) is a novel, frequent and likely treatable late-onset autosomal-dominant ataxia caused by GAA repeat-expansions in FGF14. For understanding disease evolution and imminent trial planning, metrics of the most widely used clinical outcome assessment (Scale for the Assessment and Rating of Ataxia/SARA), longitudinal progression and modifiers thereof are warranted.MethodsMulticentre intercontinental observational study (2015–2024) of 661 assessments from 219 patients with SCA27B (age: 68 ± 10 years; SARA: 9 ± 6 points) with item-level distribution-based analyses to characterise SARA metrics relative to ageing-related impairment in 390 healthy controls; and linear mixed-effects modelling to determine longitudinal progression and demographic or genetic modifiers.FindingsAtaxia severity in SCA27B as assessed by SARA was primarily attributable to gait, stance, and lower-limb impairment; other ataxia domains scored ≤1 SARA point in 79–94% of patients. Discrimination of SCA27B motor performance from controls decreased with age due to ageing-related motor variability captured by SARA, thus limiting potential metric response windows for symptomatic treatments. Disease progression was faster in the presence of interfering ageing-related comorbidities in 14 (6%) patients. Overall longitudinal progression of SCA27B was 0.54 SARA points/year [95% CI: 0.37–0.71]. Expansions of (GAA)> 180 repeats were frequent also on the shorter allele (n = 18 (8%), range: 196–348 repeats), and associated with faster progression (+1.6 SARA points/year, [95% CI: 0.9–2.2]), including also otherwise less affected ataxia domains speech and sitting.InterpretationDisease progression in SCA27B is characterised by mild progression, ageing-related motor variabilities and comorbidities, and associated with repeat size on both alleles.FundingElse-Kröner-Fresenius-Stiftung, EU, DFG, BMBF, CIHR, NAF, Ataxia-UK, CSC.Keywords: Ataxia, SCA27B, Natural history, Progression, Modifiers, Trials Research in context Evidence before this study Spinocerebellar Ataxia 27B (SCA27B) is a recently identified, frequent and likely treatable cause of late-onset ataxia due to GAA repeat expansions in one of two alleles of the FGF14 gene. Small, mostly cross-sectional studies in the PubMed database until August 2025 have described the clinical phenotype and provided preliminary estimates of disease progression in SCA27B. However, imminent trial planning in SCA27B requires a better understanding of longitudinal disease progression; of outcome metrics to best measure progression, in particular with the Scale for the Assessment and Rating of Ataxia (SARA) as most widely used clinical outcome assessment; and of potential disease modifiers to better predict progression. Further, it remains unclear how ageing-related motor impairment and comorbidities affect the assessment of disease progression and outcome metrics in the elderly population suffering from SCA27B. Added value of this study This large multicentre study including 219 patients with SCA27B provides a robust estimate of longitudinal progression (0.54 SARA points per year), and identifies impairment of gait and stance as strongest contributors to disease severity and progression. The study further demonstrates that frequent ageing-related comorbidity interferes with the assessment of disease severity in populations with SCA27B, and that capture of ageing-related motor variability impairs SARA's discrimination of motor performance between patients with SCA27B and healthy controls, thus limiting its metric response window for future symptomatic trials. Importantly, the study identifies repeat length on the shorter FGF14 allele as a potential genetic modifier of progression, suggesting that repeat size of the shorter allele is associated with a faster disease course, even already above only >180 GAA repeats. Together, these findings define clinical, ageing-related, and genetic factors that are crucial for the design and interpretation of upcoming interventional studies in SCA27B. Implications of all the available evidence This study highlights that disease progression in SCA27B, while overall mild, is primarily attributable to impairment in gait and balance, and influenced by ageing-related motor variability and comorbidity. Future clinical trials should thus prioritise outcome measures with high sensitivity in the gait and balance domains, such as digital gait metrics or balance-focused scales, and apply careful patient selection and stratification. Repeat expansions on the shorter FGF14 allele as a modifier of progression suggest a bi-allelic contribution to disease mechanisms that warrants further investigation in larger genetic and clinical studies. Introduction Intronic GAA repeat expansions in FGF14 have been recently identified as genetic cause of Spinocerebellar Ataxia 27B (SCA27B),1,2 by now already widely acknowledged as the most frequent cause of late-onset autosomal dominant cerebellar ataxia in many populations.3, 4, 5, 6, 7 Moreover, SCA27B may be amenable to already available drug treatment, with accumulating observational evidence indicating a treatment effect of 4-aminopyrdine (4-AP) in SCA27B.1,8, 9, 10, 11 Taken together, the high prevalence plus probable treatability of SCA27B underline the urgent need for a timely understanding of optimal outcome metrics, longitudinal progression, and potential modifiers thereof–not only to better understand this disease, but also to guide imminent trial planning. The Scale for the Assessment and Rating of Ataxia (SARA) presents the most common clinical outcome assessment (COA) in genetic ataxias,12, 13, 14 but available data on its metrics or progression in SCA27B is limited to cross-sectional data and/or small cohorts.4,8,15 Moreover, existing data suggests that SCA27B presents a rather mild ataxia that could be difficult to capture by the SARA in clinical trials.4,7,8 Implementation of the SARA as COA could be further complicated by the predominant manifestation of SCA27B in late-adult and elderly patients, in whom ageing-related motor impairment or comorbidities become more common.8,16,17 Such SCA27B-unrelated factors may interfere with the clinical assessment of ataxia severity per se, and/or limit the window of responsiveness of the SARA to treatment effects (given that a SARA score increased due to general ageing or other comorbidities might limit the change in SARA score achievable by any treatment in SCA27B). Compiling a large, intercontinental, multicentre, cross-sectional and longitudinal cohort of 219 patients with SCA27B assessed with the SARA, we here (i) analyse metrics of the SARA by distribution-based analysis at the item-level; (ii) compare its responsiveness to a large cohort of age-matched healthy controls up to the 8th decade of age; (iii) estimate longitudinal disease progression in SCA27B after controlling for comorbidity that may confound SARA-based ataxia severity and progression estimates; and (iv) identify modifiers of disease progression, including the size of the GAA repeats on the shorter allele. These results will be important to better understand this recently identified disease; and in particular to inform data-driven design of imminently upcoming SCA27B trials. Methods Study cohort Patients with SCA27B—prospectively or retrospectively genetically confirmed (≥250 uninterrupted GAA repeats) since the discovery of the genetic cause of the disease1—were collected from nine centres in Europe and North America (n = 219; for details across sites, see Table S1). To characterise disease evolution, all consecutive patients with SCA27B until January 2024 with at least one historical or prospective assessment with the SARA as key COA for ataxia,12 documented at the item-level, were included (observation period across centres: 2015–2024). The SARA yields total scores from 0 (no ataxia) to 40 (most severe ataxia), based on clinical examination across eight items (gait: 0–8; stance: 0–6 and sitting: 0–4; speech: 0–6; finger chase, nose-finger test, fast alternating hand movements, and heel-shin slide: 0–4 points). Patients and their medical records, respectively, were systematically reviewed by the local physician for all previous SARA assessments to avoid sampling bias; for the presence of non-ataxia features; and for further clinical information that might influence disease evolution of SCA27B (=demographics, age of onset, visit dates), or otherwise interfere with the assessment of motor impairment, e.g. diseases unrelated to SCA27B (=comorbidities), or effects due to treatment with 4-AP or acetazolamide. Sex was not self-reported for the purposes of this study. It was extracted by the local investigators from the administrative patient records of each participating centre, where it is registered as legal/administrative sex on the basis of health-insurance registration data; gender was not assessed. Data was collected and linked using a common structured data sheet, including prespecified queries about genetics, demographics, comorbidity, treatment, and item-level SARA scores. Genotyping Genetic testing for GAA repeat expansions in FGF14 was performed by long-read sequencing, or sequential testing by long-range PCR, repeat-primed PCR, and Sanger sequencing, as previously described.1,18,19 GAA repeat number was determined for both the longer and the shorter allele in all patients, using a pathogenic cutoff of ≥250 uninterrupted GAA repeats on the longer allele.1,2,15 Ethics This study adhered to the 2024 Declaration of Helsinki, data collection was approved by the institutional review board of the medical faculty of the University of Tübingen (598/2011BO1), and patients at all contributing sites provided written informed consent. Statistics To determine how disease severity and its progression in SCA27B is reflected by the SARA across different ataxia domains, its metrics were characterised by distribution-based analysis of its total score as well as its item subscores across all baseline assessments. This allowed to analyse the relative share of patients at each item severity level and potential floor/ceiling effects. To characterise how assessment by the SARA is affected by general, i.e. SCA27B-independent, ageing-related motor impairments -as might be expected in late-adult and elderly populations like SCA27B-, the distribution of the SARA and its items in all 219 patients with SCA27B at baseline were next compared to cross-sectional assessments in 390 healthy elderly controls between 18 and 79 years of age (mean: 47.5 ± 17.1 years) from an independent validation cohort, also assessed by experienced clinical raters as part of distinct SARA validation studies and in non-mutation carriers in prospective natural history studies of presymptomatic spinocerebellar ataxias.20 This comparison was stratified by the decade of each subject's age at SARA assessment, and the area under the curve (AUC) was calculated as effect size for the discrimination between patients with SCA27B and healthy controls. We thereby aimed to quantify the potential “window of responsiveness” of the SARA in SCA27B, i.e. the delta between the SARA of the disease state (SCA27B) and the SARA of the general, ageing-related healthy state in the corresponding age decade. Longitudinal disease progression in SARA for SCA27B was analysed by linear mixed-effect modelling (LMEM, restricted maximum likelihood, covariance structure: variance components, Kenward– Roger degrees of freedom) with random effects on intercept (score at baseline) and slope (points per year of follow-up).21 LMEM was performed both in the overall cohort, and in the clinically assumed most trial-relevant subcohort of ambulatory patients, defined by a SARA gait item ≤3, i.e. milder than levels requiring intermittent support or walking aids. To obtain valid progression estimates, clinical records of patients exceeding the 95th percentile of SARA total scores per decade of age, and/or the 95th percentile of population intercepts and slopes of a preliminary LMEM were manually reviewed for potential outliers before the final analysis. These patients were excluded from all final LMEM analyses if SCA27B-independent diseases, multi-morbidity with reduced functional status, and/or recent health events were identified that likely interfered with the SARA at the time of motor assessment, as evaluated independently by two assessors (A.T., M.Sy.). Two separate LMEMs were calculated for (i) the SARA total score and (ii) its item-level subscores to explore and compare disease evolution in respective ataxia effector domains. For the SARA total score, LMEM included ataxia severity (SARA) and disease duration at baseline as well as 4-AP treatment as covariates; and sex, age of onset, and GAA repeat number as exploratory independent factors, selected a priori based on expert knowledge and previous literature. For the item-level analyses, we applied a parsimonious modelling approach by restricting models to factors that had shown relevant associations at the total score level, in order to avoid overfitting and maintain interpretability. The item-level model included the normalised subscores in a joint LMEM for simultaneous assessment of most relevant item progression. The results were retransformed for presentation purposes. Descriptive and test statistics, linear regression analysis, Spearman correlations, and distribution-based analyses, including receiver operating characteristics, were calculated using Prism 10 (GraphPad Software, La Jolla, CA). LMEM was performed with SAS, version 9.4 (procedure MIXED, NLMIXED). Data are shown as mean ± standard deviation, interquartile range (IQR), 95% confidence intervals (95% CI), or standard error of the mean (SEM). No formal sample-size calculation was performed for this observational cohort. Role of funders The funders of this study had no role in study design, data collection, data analysis, data interpretation, the writing of the report, or the decision to submit the paper for publication. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Cohort characteristics This study included 661 SARA assessments from 219 patients (101 female) with genetically confirmed SCA27B across nine centres in Germany (Tübingen: n = 54; Essen: n = 34; Berlin: n = 21; Munich: n = 12; Lübeck: n = 7), Canada (Montreal: n = 44), France (Strasbourg: n = 26; Nancy: n = 11), and Austria (Innsbruck: n = 10), including 116 patients across all centres with longitudinal assessments over an average of 4 [IQR: 3–6] follow-up visits in 4.7 ± 3.4 years (range: 0.3–15.4; Table 1, see Table S1 for details across sites). The full study cohort spanned a wide range of patient age (68 ± 10 years, range: 24–86), age of onset (59 ± 12 years, range: 21–79), disease duration (9 ± 7 years, range: 0–32), and ataxia severity at baseline (SARA: 9 ± 6 points, range: 0–34), thus allowing for representative analyses of SARA metrics and progression in SCA27B. Mild to moderate impairment of the vibration sense at the ankle indicating sensory neuropathy (93/215 patients assessed, 43%) and an impaired head-impulse test indicating vestibular dysfunction (20/100, 20%) were frequent non-ataxia features known in SCA27B; less common non-ataxia features were mild cognitive impairment (19/216, 8%), signs of pyramidal tract involvement in the legs (hyperreflexia, spasticity or extensor plantar response: 16/216, 7%), mild postural or head tremor (12/216, 6%), and mild parkinsonism (11/219, 5%; see Table S1 for details). Genetically, the number of GAA repeats on the longer allele ranged from 251 to 937 repeats (median: 356, IQR: 303–407) and was negatively associated with the age of onset, decreasing the average age of onset by 4.5 years [95% CI: 2.3–6.6] per 100 repeats between 250 and 600 repeats (linear regression, p < 0.001; n = 218, i.e. excluding one outlier patient with 937 repeats, Fig. 1). Table 1Cohort characteristics at baseline visit.Total cohort (n = 219)Age (years)68 ± 10 (24–86)Sex101 female (46%)118 male (54%)0 other (0%)Age of onset (years)59 ± 12 (21–79)Disease duration (years)9 ± 7 (0–32)Ataxia severity (SARA score)9 ± 6 (0–34)Number of FGF14 GAA repeats Longer allele356 [303–407] Shorter allele16 [9–53]Drug treatment On 4-AP or 3,4-DAP40 (18%) On Acetazolamide14 (6%)Longitudinal follow-up Number of patients116 (53%) Number of follow-up visits4 [3–6] Follow-up time (years)4.7 ± 3.4 (0.3–15.4)Non-ataxia features Impaired vibration sense at ankle93/215 (43%) Impaired head impulse test20/100 (20%) Pyramidal signs in the legs16/216 (7%) Paresis3/218 (1%) Postural or head tremor12/216 (6%) Parkinsonism11/219 (5%) Cognitive impairment19/216 (8%)Data presented as mean ± standard deviation (range), median [interquartile range], or number of patients (%cohort).4-AP = 4-aminopyridine; 3,4-DAP = 3,4-diaminopyridine. Cohort characteristics at baseline visit. Data presented as mean ± standard deviation (range), median [interquartile range], or number of patients (%cohort). 4-AP = 4-aminopyridine; 3,4-DAP = 3,4-diaminopyridine. Association between age of onset and GAA repeat number on the longer allele in 218 patients with SCA27B, excluding one patient with an extreme value of 937 repeats. Solid line indicates linear regression, dashed lines reflect 95% confidence bands of the best-fit line. Item-level SARA metrics SARA total score and all item subscores at baseline correlated with disease duration, thus capturing disease severity over time, albeit with only small-to-moderate effect sizes, indicating that factors other than just disease duration influence the SARA levels (Spearman correlation, SARA: rho = 0.30 [95% CI: 0.17–0.42], p < 0.001; items: rho = 0.17–0.32 [95% CI: 0.03–0.44], all p ≤ 0.030; Fig. S2). Distribution of severity degrees per item, i.e. score levels per item, showed that ataxia severity in SCA27B was mainly captured by SARA items gait and stance in the balance domain, and the heel-shin item in the lower limb domain (Fig. 2). In items gait and stance, at least 50% of patients exceeded a subscore of 2; and 24% and 10% had subscores of 4 or higher, thus requiring at least intermittent support for gait and stance, respectively. In contrast, all SARA items apart from gait showed considerable floor effects, and patients with SCA27B predominantly scored ≤1 in other axial (sitting: 94%, speech: 81%) and upper limb items (finger–chase: 84%, nose-finger: 91%; diadochokinesia: 79%). Disease severity in SCA27B across ataxias domains, as measured by the distribution of scores across SARA items among 219 patients. Numbers in brackets indicate maximum scoring range per item. No statistical test was applied. SARA scores in SCA27B relative to healthy controls across age groups We next explored how motor impairments in general ageing, i.e. unrelated to SCA27B (or any other specific ataxia or other neurologic disease), might influence the SARA levels; and determined the relative delta in SARA levels in SCA27B vs. general ageing. For this, we compared the cross-sectional distribution of the SARA and its items in the SCA27B cohort to a large cohort of 390 healthy controls in the respective same age decade (Fig. 3). SARA total scores showed a consistent difference between patients with SCA27B and controls across decades, indicated by the fact that the progression of ataxia severity in SCA27B exceeded the average ageing-related increase of motor impairment (e.g. median [IQR] at 40–50 years: 2.5 [1.5–11] vs. 0 [0–1] points; 70–80 years: 8.5 [5–12.5] vs. 1.25 [0–3] points). However, the inter-individual variability of motor impairment—as assessed by the SARA—increased in healthy aged controls, leading to a decreasing difference between item levels of patients with SCA27B and controls with increasing age (e.g. median [IQR] of patients with SCA27B vs. 90th percentile in controls; 50–60 years: 6 [3.5–12] vs. 1 point; 60–70 years: 6 [4–10] vs. 2.5 points; 70–80 years: 8.5 [5–12.5] vs. 5 points). Accordingly, the effect size of the discrimination between patients with SCA27B and controls by the SARA worsened with age (receiver operating characteristics, AUC: 0.97 (50–60), 0.95 (60–70), and 0.91 (70–80), respectively, though all p < 0.001). This decreasing difference, which reflects the potential treatment-related improvement delta SARA in SCA27B relative to merely ageing-related SARA levels, suggests that the metric response window of the SARA decreases with higher age in patients with SCA27B. Comparison of SCA27B severity to healthy ageing-related motor impairment. SARA total score and item scores of 219 patients with SCA27B at the age of baseline assessment vs. normative reference values in 390 healthy aged controls. Solid lines and dashed lines reflect medians and 90th percentile calculated per decade of age, respectively. The effect size of the discrimination between patients with SCA27B and controls by the SARA worsened with age, with AUCs of 0.97 (50–60 years), 0.95 (60–70 years), and 0.91 (70–80 years), respectively, for the total SARA score (receiver operating characteristics, all p < 0.001). This difference between patients with SCA27B and age-stratified controls corresponds to the possible response window for a change in SARA in a hypothetical trial of a symptomatic treatment. At the item level, only SARA items gait and stance showed up to a 2-point difference between the median of patients with SCA27B vs. controls, while items finger–chase and heel-shin had a maximum 1-point difference (Fig. 3), indicating a potentially higher response window for gait and stance than for the latter appendicular items. Such larger differences for gait and stance between SCA27B and controls were observed across all age groups, thus suggesting even an age-independent, general aspect of SCA27B disease. This was further corroborated by the observation that ageing-related motor impairments increased in particular the variability of the appendicular items of the SARA. For example, a 90% percentile of up to 1.5 SARA points was observed in healthy controls for the items diadochokinesia and heel-shin above age 70 years. In contrast to gait and stance, this ageing-related variability in the appendicular items exceeded even the median item level observed in the patients with SCA27B. Progression modelling and interference by frequent comorbidity in late-adult populations Visual inspection of SARA scores across age groups and preliminary progression modelling highlighted several patients with SCA27B with unusually high ataxia severity and/or fast progression (Fig. 4). An in-depth clinical review of 29 patients fulfilling objective outlier criteria (see Methods) revealed significant comorbidities that interfered with the SARA assessment in 14 of these patients (6% of the total SCA27B cohort). These comorbidities included acquired and genetic second diseases (e.g. amyotrophic lateral sclerosis or RFC1-associated ataxia), multi-morbidity (e.g. dementia or cancer with reduced functional status), and sequelae of recent events (e.g. falls or surgery; see Table S3 for list of comorbidities). After removing these outliers, the estimated annual progression in SCA27B was 0.54 SARA points/year [95% CI: 0.37–0.71] in the overall cohort (LMEM, p < 0.001; Fig. 5A; see Table S4 for details of the model). In the most trial-relevant subcohort of patients still able to walk independently (SARA gait item ≤3), the estimated annual progression was 0.40 SARA points/year [95% CI: 0.26–0.55; p < 0.001]. At the item level, two thirds of the overall SARA progression was captured by its items gait (0.20 points/year [95% CI: 0.16–0.24], p < 0.001) and stance (0.16 points/year [95% CI: 0.13–0.19], p < 0.001). The relative progression in these items—normalised to their scoring range—exceeded all other items and the total score of the SARA (Fig. 5B; see Table S5 for item-level statistical results), highlighting gait and balance as key ataxia domains affected by longitudinal progression in SCA27B. Sex, age of onset, disease duration, and ataxia severity (SARA score) at baseline were not associated with progression of the total SARA or any of its items (LMEM, all p > 0.05; Tables S4 and S5). However, age of onset (i.e. patient age) was associated with higher subscores in items stance (0.08 points per decade, [95% CI: 0.01–0.15], p = 0.034) and finger–chase (0.06 points per decade, [95% CI: 0.02–0.11], p = 0.009), thus indicating that stance assessment in SCA27B is susceptible to ageing-related motor impairment independent of its large difference between patients with SCA27B and controls. Outlier detection of excess ataxia severity and/or progression in patients with SCA27B and comorbidity affecting motor performance as captured by the SARA. Grey dots reflect scatter of all 661 SARA assessments across 219 patients; dashed line reflects 90th percentile of their distribution calculated per decade of age. Longitudinal disease progression in SCA27B, as captured by the SARA total score (a) and across ataxia domains by the respective SARA items (b). Lines indicate progression estimate in linear mixed-effect model. Numbers in brackets reflect 95% confidence interval of estimate. Dots in (a) reflect scatter of all 661 SARA assessments across 219 patients with SCA27B, relative to the time since their baseline assessment. Item scores in (b) were normalised to the respective scoring range to allow comparison of progression rates across ataxia domains independent of metric differences of the scale. Modification of progression by repeat size of the shorter allele Six patients of the cohort showed expanded alleles above the established pathogenic threshold of (GAA)≥ 250 repeats also on the shorter allele (range: 251–348), thus carrying bi-allelic pathogenic expansions in FGF14 (Fig. 6A). Additional twelve patients carried repeat expansions of at least “intermediate size” in the range of (GAA)180–250 repeats on the shorter allele (range: 196–242), i.e. above a lower pathogenic threshold of (GAA)> 180 repeats recently proposed based on their enrichment in ataxia cohorts.19 This yielded a total of 18 patients (8%) in our cohort with (GAA)> 180 repeats on the shorter allele, including two patients with also substantial comorbidities (P13: 251/275 repeats; P166: 242/203 repeats). Repeat numbers on the shorter allele were negatively correlated to repeat numbers on the longer allele (Spearman correlation, rho: −0.20, [95% CI: −0.32 to −0.06], p = 0.004; n = 218, i.e. excluding one outlier with 937/10 repeats on the longer/shorter allele), i.e. repeat expansions on the shorter allele were not evenly distributed but enriched in patients with shorter expansions on the longer allele (Fig. 6B). GAA repeat number on the shorter allele was not associated with the age of onset (Spearman correlation, rho = 0.004 [95% CI: −0.13–0.14]. p = 0.956). However, compared to the full cohort, ataxia severity and/or progression as measured by the SARA appeared higher in patients with bi-allelic (GAA)≥ 250 expansions (e.g. P88 or P132), and even also already in patients with (GAA)> 180 expansions on the shorter allele (e.g. P205 or P184; Fig. 6C). In LMEM analysis (excluding patients P13 and P16 with comorbidity as possible alternative cause), GAA repeat number on the shorter allele was associated with faster longitudinal disease progression in SCA27B (+0.48 SARA points/year per 100 repeats, [95% CI: 0.19–0.77], p = 0.001). Dichotomised by a threshold of (GAA)> 180 repeats,19 progression in SCA27B was +1.6 SARA points/year faster [95% CI: 0.9–2.2] in patients with GAA repeat numbers on the shorter allele above than below this threshold (LMEM, p < 0.001). For the established cutoff of (GAA)≥ 250 repeats,1,2,15 progression was +3.2 points/year faster [95% CI: 1.4–5.0] above than below the threshold (p < 0.001). At the SARA item level, the short repeat number was associated with accelerated progression of the most responsive items gait and stance (by 78% and 47% per 100 repeats relative to progression in the overall cohort, LMEM, p < 0.001 and p = 0.003); and, in addition, particularly also of the otherwise less responsive items sitting or speech (by 147% and 174% per 100 repeats; p = 0.002 p = 0.017; see Table S5 for details). Repeat expansions on the shorter allele of 219 patients with SCA27B. (a) Distribution of repeat numbers on the shorter allele. (b) Spearman correlation of repeat numbers on shorter allele vs. longer allele, excluding one outlier with 937/10 repeats on the longer/shorter allele. (c) Ataxia severity and/or progression in patients with SCA27B and >180 repeat expansions on the shorter allele. Numbers in brackets indicate repeat number on both alleles. Grey dots reflect scatter of all 661 SARA assessments across 219 patients; dashed line reflects 90th percentile of their distribution calculated per decade of age. Discussion Leveraging a large cross-sectional and longitudinal multicentre cohort of patients with SCA27B, this study facilitates a deeper understanding of the overall and domain-specific longitudinal disease progression of SCA27B, the metrics of its capture by the SARA, and clinical and in particular genetic modifiers thereof. These results will facilitate a deeper understanding of disease evaluation, and in particular trial planning in this novel, frequent and likely treatable genetic ataxia. Ataxia severity in SCA27B, as assessed by the SARA, was primarily attributable to impairment in gait, stance, and lower limb coordination (heel-shin). In contrast, impairment in other ataxia domains contributed substantially less to overall ataxia severity, yielding substantial floor effects and mild subscores ≤1 in SARA score in at least 80% of patients. The strong contribution of gait, stance and lower-limb impairment to overall ataxia severity in SCA27B likely reflects the actual disease phenotype of SCA27B, although capture of gait and balance impairment may be facilitated by the larger number of score levels in items gait and stance than in other SARA items. This might be due to the fact that SCA27B presents as relatively mild ataxia, even across this large cohort of 219 patients (with SARA scores ≤10 in 70% of patients), i.e. in an ataxia disease stage in which genetic ataxias are typically dominated by gait and balance impairment.22 It might also point to a major contribution of vestibulo-cerebellar involvement in gait and balance, as suggested earlier.11,23 In contrast, cerebellar networks underlying speech and upper limb control seem to be relatively spared over a long time of the neurodegenerative disease course in most patients with SCA27B, as indicated by a majority of patients with ≤1 point in other SARA items. Correspondingly, to maximise COA sensitivity, our study indicates that future trials in SCA27B should include COAs with higher granularity in the gait and balance domain, e.g. digital gait analysis,10 the Upright Stability Scale of the Friedreich Ataxia Rating Scale (FARS-E),24 or COAs which focus even stronger on capturing dynamic balance.25 These COAs must now be validated in preparation for clinical trials in SCA27B and are thus part of emerging large international multimodal SCA27B natural history studies (ClinicalTrials.gov: NCT06472557). Our comparison of SARA score distributions between patients with SCA27B and a large cohort of age-stratified controls demonstrated that the effective discrimination between the disease vs. healthy state decreases with age due to ageing-related motor impairment. This is a major caveat for trial planning in SCA27B, given that it is inherently one of the ataxias with the latest, elderly age of onset (average onset in our cohort: 59 years; average onset in published cohorts: 54–68 years1, 2, 3, 4, 5,8). Our analysis does not directly infer treatment effect sizes, as the relevant comparator for estimating treatment effects in interventional trials is the counterfactual disease trajectory in untreated patients with SCA27B, rather than the difference to healthy controls. However, the decreasing difference to healthy age-matched controls delineates the maximal SARA delta that a treatment could realistically achieve, as therapeutic interventions are not expected to improve patients beyond the functional level typically observed in healthy individuals of the same age. In a trial scenario, this poorer discrimination would directly correspond to a smaller metric response window of the SARA for a symptomatic treatment effect, e.g. as presumed for 4-AP.8,15 Our item-level comparison showed that this metric interference by ageing-related motor impairment is particularly conveyed by the appendicular items of the SARA, in which ageing-related variability may be indistinguishable (or even exceed) the mild disease-related motor impairment in this domain in SCA27B. This is important in particular for lower-limb coordination, for which the heel-shin item appears to drive SCA27B-related impairment as measured by the SARA, but which becomes susceptible to ageing-related impairment above age 60 years. However, our item-level LMEM demonstrated that age is also associated with poorer performance of stance. Thus, while gait- and balance-centric COAs that omit appendicular ataxia (like the FARS-E26 or the fSARA27) might probably be overall more responsive than the SARA in SCA27B, they may still be susceptible to age effects in the elderly population suffering from SCA27B. Digital gait analysis may provide better sensitivity in SCA27B if digital measures were specific for ataxia as opposed to healthy ageing, e.g. by capturing gait variability due to cerebellar ataxia as opposed to gait slowing also observed with age,28,29 or dynamic gait aspects like turning30 that might be particularly impaired in SCA27B. This highlights the lack of, and at the same time urgent need of, comparative data in healthy aged controls for most clinician-reported and digital outcomes in the ataxia field. Their validation in healthy elderly cohorts will be key for trial planning in SCA27B or similar late-onset ataxias (like e.g. RFC1-disease31). By linear mixed-effect modelling of a large multicentre longitudinal dataset, the present study provides the robust estimate of disease progression in SCA27B. The average progression of 0.54 SARA points/year tends to be higher than previous estimates in smaller, cross-sectional cohorts,4,7,8 but still falls into a rather mild overall progression rate, lower than most other repeat SCAs (even lower than 0.8 points/year estimated for SCA614) or other genetic ataxias (e.g. 0.6–1.6 points/year for most autosomal recessive cerebellar ataxias).22,32 In contrast to Friedreich Ataxia as another GAA-driven ataxia,33 baseline ataxia severity was not a predictor of longitudinal disease progression in SCA27B. Two thirds of this longitudinal progression were contributed by progressive impairment in the gait and stance domain, although the corresponding SARA items gait and stance make up only 35% of the SARA total score. While natural history progression does not directly predict the magnitude or time course of treatment effects, the distinct progression in these domains provides important information on their dynamic range and responsiveness. Domains that show little change over time in untreated disease appear likely to have limited sensitivity to detect change, particularly in disease-modifying trial settings. Our data suggest that gait, stance, and lower-limb components contribute most strongly at least to longitudinal natural history change in SCA27B, thus corroborating and extending their importance from SCA27B ataxia severity (demonstrated by our distribution-based SARA metrics analyses) to their contribution to SCA27B ataxia progression. Given the above mentioned mild overall disease progression—and the poor sensitivity of clinician-reported outcomes (ClinROs) for ataxia in general,14,34,35 proving efficacy of disease-modifying treatments will pose an even greater challenge in SCA27B than in other genetic ataxias.36 This challenge could even be larger for the most trial relevant target group of patients still able to walk independently (SARA gait item ≤3), in whom estimated progression was only 0.40 SARA points/year. A SARA gait score ≤3 was chosen as a proxy for free, i.e. unassisted ambulation, which represents a trial inclusion criterion in most current ataxia treatment trials testing disease-modifying treatments in SCAs (ClinicalTrials.gov: e.g. NCT05160558, NCT06672445, or NCT02960893). The mild diseases progression emphasises the need for future treatment trials to include: ClinROs more focused on the specific dysfunctions in SCA27B; digital-motor outcomes with demonstrated higher sensitivity to change to gait and balance dysfunction than the SARA (e.g. turning parameters30); and/or composite measures combining optimised ClinROs plus optimised digital-motor outcomes.37 Our results furthermore indicate that trial inclusion and treatment response capture in SCA27B might be in particular confounded by an inherently high rate of comorbidities in SCA27B. Fourteen of 29 patients, which had an outlier severity and/or progression rate beyond the 95th percentile in the overall cohort, presented with comorbidities as relevant clinical modifiers of disease severity and progression in SCA27B. Exclusion of these patients aimed at minimising the influence of confounding factors that may transiently or permanently distort SARA-based assessments of ataxia severity; and, at the same time, at ensuring robust attribution of observed progression estimates to the underlying disease per se, in particular in heterogeneous real-world settings such as the one analysed here. The high rate of comorbidities seems already a priori reasonable in late-adult populations like SCA27B, but still largely new to the ataxia trial space where the large majority of other SCAs present with a much younger age, and are thus much less confronted with the necessity to exclude patients from trials based on comorbidity-related functional impairment.38,39 Such comorbidities may (i) mask the actual disease severity and/or progression by exceeding the true disease-related motor impairment (e.g. due to independent dementia and small vessel disease [subject P208], neuromuscular disease [P70, P119], or Parkinsonism [P145]); and/or (ii) interfere with trial outcome assessment by unexpected singular events (e.g. hospitalisation or surgery, eventually leading to dropouts). Preparation of trials in SCA27B will thus require rigorous exclusion criteria and in-depth screening for the presence of comorbidities to increase statistical power while maintaining external validity for elderly populations. Finally, our study also supports GAA repeat expansions also on the cis shorter allele, i.e. non-dominant FGF14 allele, as genetic modifier of SCA27B disease progression. These repeat alleles were estimated to increase longitudinal progression by +0.48 SARA points/year per 100 repeats. This association reflects a relationship in the natural disease trajectory but does not imply it would moderate or predict treatment effects. Rather, it strongly suggests that the short allele must be considered for stratification, adjustment, or enrichment strategies in upcoming trials in SCA27B, similar to what has now become state of the art for Friedreich Ataxia, - interestingly also a GAA repeat expansion (yet autosomal-recessive) disease.13,26 As in Friedreich Ataxia,40 larger repeat sizes may be associated with greater transcriptional impairment also in SCA27B, which may contribute to faster disease progression—in an additive manner with the larger allele. At the domain level, expansions on the shorter allele were associated with faster progression in those ataxia domains that dominated ataxia severity in the overall cohort (gait, stance, and lower-limb involvement). Yet, they were particularly associated with greater annual progression also in the domains of sitting and speech by 147% and 174% per 100 repeats, respectively. Based on the distribution of repeat numbers on the shorter allele, 8% of patients with SCA27B carried a second repeat expansion of (GAA)> 180 repeats. This lower threshold, here independently suggested by the distribution of repeat numbers on the shorter allele, falls into the range of a lower threshold of (GAA)> 180 repeats recently suggested for the longer allele.19 The presence of such a shorter allele was associated with an estimated +1.56 SARA points/year faster progression as compared to patients with shorter alleles of (GAA)< 180 repeats, thus adding to the pathogenicity of the longer allele. For patients with bi-allelic repeats above the well-established pathogenic cutoff of (GAA) ≥ 250 repeats, our study estimated even faster progression (+3.2 points/year). This might indicate gradually increasing and synergistic pathogenicity of both alleles in the range of (GAA)180–250 repeats. Additional studies are required to characterise the frequency and interaction of bi-allelic repeats in this “intermediate” range, as patients with SCA27B and longer alleles below 250 repeats were not included in our study, and the frequency of bi-allelic repeat expansion carriers is thus probably even higher than our estimate. Limitations of our study include the retrospective aggregation of SARA assessments; while each of them had been per se prospectively assessed (i.e. not post hoc reconstructed), they had not been prospectively aggregated upfront in a centralised registry. Accordingly, prior rater training was also not a formal requirement in this study, although SARA assessment had been performed predominantly by experienced clinical raters at ataxia expert centres with standardised assessment as part of recruitment into international ataxia registries. Moreover, the SARA does not assess downbeat nystagmus, which is a relevant impairment in SCA27B. Further validation is thus warranted, which is now ongoing in another recently initiated large multimodal international SCA27B natural history study (NCT06472557). While our study includes a large SCA27B cohort, the number of datasets were still too limited to allow for validation of our LMEM by influence in residual diagnostic or by variable selection. The model was thus fitted based on clinical hypotheses based on SCA27B literature and expert experience about likely relevant exploratory factors. Potential attrition bias, assuming a missing-at-random mechanism, was reduced by the LMEM estimation approach. Potential distortion of SARA-based ataxia severity estimates was minimised by screening the available data for outliers in disease severity and/or progression and reviewing the respective clinical records by local SCA27B neurologists for confirmation of the data and/or identification of interfering factors. Future studies in even larger SCA27B cohorts are warranted to confirm our findings, including the frequency of bi-allelic repeat expansion carriers, and in particular to confirm and further delineate clinical and mechanistic effects of shorter alleles >180–200 repeats on SCA27B disease phenotype and progression. In conclusion, by compiling and analysing a large international cohort of patients with SCA27B, this study provides insights into disease progression trajectories, SARA performance metrics, and genetic modifiers. These findings add to the clinical understanding of SCA27B and offer a basis to inform the design of now incipient treatment trials in this likely treatable ataxia. Contributors AT, RDH, and MS contributed to the conception and design of the study. AT, FE, CDep, TW, CDel, AN, CA, DP, EI, FH, MR, MB, MGE, JF, TK, LS, BB, MA, DT, and MS contributed to the acquisition of data. AT, RDH, and MS contributed to the analysis or interpretation of the data. AT and MS contributed to draughting the text and preparing the figures. All authors critically reviewed the manuscript for important intellectual content: AT and MS have directly accessed and verified the underlying data reported in the manuscript. All authors confirm that they had full access to the data in the study and accept responsibility to submit for publication. All authors read and approved the final version of the manuscript. Data sharing statement Anonymised data underlying this article will be made available upon reasonable request by qualified investigators. Declaration of interests AT, RDH, FE, CDep, TW, CDel, AN, CA, DP, EI, FH, MR, and BB have nothing to disclose. MB received honoraria from Bial Germany, unrelated to the present manuscript. MGE received consulting fees from Biogen Germany, and travel support from Aparito, a wholly owned subsidiary of Eli Lilly and Company, all unrelated to this project. JF has received consultancy honoraria from Vico therapeutics and Biogen, unrelated to the present manuscript. TK is receiving research support from the Bundesministerium für Forschung, Technologie und Raumfahrt (BMFTR), Servier, and UCB, all unrelated to this project. Within the last 24 months, he has received consulting fees from BottNeuro, Arrowhead, Bristol-Myers Squibb, and UCB, all unrelated to this project. LS served as consultant for Alexion and Reata Pharmaceuticals unrelated to the present manuscript. MA received honoraria from IPSEN, Abbvie, Orkyn, Teva, Reata Pharmaceuticals, Biogen, Aguettant, Merz, Ever Pharma, Asdia, all unrelated to the present manuscript. DT received funding from the DFG, EU and Bernd Fink Foundation, all unrelated to this project. MS has received consultancy honoraria from Ionis, UCB, Prevail, Orphazyme, Biogen, Servier, Reata, GenOrph, AviadoBio, Biohaven, Zevra, Lilly, Quince, and Solaxa, all unrelated to the present manuscript. Acknowledgements This work was supported, in part, by the Clinician Scientist program “PRECISE.net” funded by the Else Kröner-Fresenius-Stiftung (AT, MS); by the project European Rare Disease Research Alliance (ERDERA), GA n°101156595, funded under call HORIZON-HLTH-2023-DISEASE-07 (MS, RDH) by the European Union; and by the Canadian Institutes of Health Research (grant 189963, BB). MA thanks the association “Connaître les Syndromes Cérébelleux” (CSC) that supported the study. FE received funding from the Deutsche Forschungsgemeinschaft (DFG) in the framework of the DFG Clinician Scientist Programme UMEA, FU 356/12–2. D.P. is supported by a Fellowship award from the Canadian Institutes of Health Research (CIHR). MGE received research support from the German Ministry of Education and Research (BMBF) within the European Joint Program for Rare Diseases (EJP-RD) 2021 Transnational Call for Rare Disease Research Projects (Funding number 01GM2110), from the National Ataxia Foundation (NAF), and from Ataxia UK. JF was funded within the Advanced Clinician Scientist Programme (ACCENT, funded by the Federal Ministry of Research, Technology and Space (BMFTR) under the funding code (FKZ): 01EO2107) and as a PI of the iBehave Network, sponsored by the Ministry of Culture and Science of the State of North Rhine-Westphalia and received funding from the National Ataxia Foundation (NAF). Several authors are members of the European Reference Network for Rare Neurological Diseases. We thank Michael Strupp and Guillemette Clement for their support in data acquisition. Footnotes Supplementary data related to this article can be found at https://doi.org/10.1016/j.ebiom.2026.106437. Supplementary data Supplementary Material References Associated Data Supplementary Materials Supplementary Material