Heart Rate Variability Provides Prognostic value in Multiple System Atrophy

Heart Rate Variability Provides Prognostic value in Multiple System Atrophy | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Heart Rate Variability Provides Prognostic value in Multiple System Atrophy Paulo Bastos, Marc Kermongant, Margherita Fabbri, Frederic Roche, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7325643/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 23 Feb, 2026 Read the published version in Clinical Autonomic Research → Version 1 posted 4 You are reading this latest preprint version Abstract Purpose Multiple system atrophy (MSA) is a progressive neurodegenerative disorder characterized by autonomic dysfunction, parkinsonism, and cerebellar impairment. Predicting disease severity and survival remains challenging due to the heterogeneity of disease progression. Heart rate variability (HRV), a non-invasive measure of autonomic nervous system function, has been used as a biomarker of autonomic failure. However, the role of non/linear HRV in MSA remains underexplored, and its prognostic value is yet to be fully established. Methods This study aimed to evaluate the predictive value of HRV features in MSA, identifying HRV features most predictive of mortality and survival, and assessing whether HRV provides unique, complementary insights beyond traditional clinical severity measures (n = 214). Regression models were employed to assess the association between HRV features and disease severity, as assessed by the Unified MSA Rating Scale (UMSARS), or time-to-death. Survival analyses were used to investigate HRV’s prognostic value. Mediation analysis explored the relationship between HRV, UMSARS, and survival. Results HRV features demonstrated negative correlations with disease severity, mirroring clinical deterioration. While no single HRV feature showed strong correlations with the UMSARS, their combination was a significant predictor. HRV alone predicted time-to-death almost as well as the UMSARS and combining HRV with UMSARS significantly improved survival prediction accuracy. HRV maintained a direct effect on survival, independent of the UMSARS, highlighting its distinct physiological relevance. Conclusions HRV provides valuable, complementary information beyond UMSARS in predicting disease severity and survival in MSA. While HRV alone has only moderate predictive power, it captures distinct physiological processes not reflected by traditional clinical scales. HRV MSA neurodegeneration dysautonomia Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Multiple system atrophy (MSA) is a relentlessly progressive neurodegenerative disorder characterized by a complex interplay of autonomic failure, parkinsonism, and cerebellar dysfunction.[ 1 ] Despite advancements in understanding its clinical trajectory,[ 2 , 3 ] accurately predicting disease severity and survival remains a significant challenge, largely due to the heterogeneous and multifaceted expression of MSA. Current clinical assessment, primarily based on the Unified Multiple System Atrophy Rating Scale (UMSARS), offer essential but inherently limited evaluations of disease burden, as they may fail to capture the deeper, underlying pathophysiological mechanisms driving disease progression.[ 4 , 5 ] This underscores a pressing need for novel, complementary biomarkers that can enhance predictive modelling, refine patient stratification, and optimize clinical management in MSA. Autonomic failure is a cornerstone of MSA pathophysiology and clinical management.[ 6 ] The progression of the UMSARS scores and the severity of orthostatic hypotension (OH) - a key feature of autonomic failure - have been shown to be key markers/factors of poor survival.[ 7 ] Heart rate variability (HRV), a non-invasive and dynamic measure of autonomic nervous system function, has been used as a biomarker of dysautonomia (e.g., [ 8 ]). HRV, the variation in the period between consecutive heartbeats over time, is a useful signal to understand the status of the autonomic nervous system (ANS) and provides information about the interaction between the sympathetic and parasympathetic systems, being one of the distinguishing features across many cardiovascular pathologies (e.g., [ 9 ]). HRV reflects complex physiological interactions governing autonomic regulation, cardiovascular integrity, and even neural network stability—domains highly relevant to MSA, where autonomic dysfunction is a defining clinical feature.[ 10 – 14 ] The fact that HRV is reduced in MSA compared to control populations has been previously demonstrated.[ 15 ] Despite its theoretical promise, the clinical utility of HRV in MSA remains underexplored, and its precise relationship with disease severity and survival has yet to be firmly established. In this study, we systematically evaluated the predictive power of HRV in MSA, addressing critical gaps in the field. In addition to commonly used HRV analysis domains such as time-domain and frequency-domain, we have also used non-linear approaches to better understand HRV complexity.[ 16 , 17 ] These extra HRV domains (e.g., fractal analysis, short-term complexity, entropy, regularity, nonlinear dynamics) have been identified as informative for a better understanding of HRV. However, their use as clinical tools has to date been rather limited.[ 18 ] Specifically, we investigated: The relationship between HRV metrics and disease severity, as measured by the clinical gold-standard UMSARS. The prognostic value of HRV for survival prediction, identifying the HRV features most strongly associated with mortality. The independent contribution of HRV beyond traditional clinical assessment (cardiovascular autonomic tests), determining whether it provides non-redundant insights into disease progression. The independent contribution of HRV beyond traditional clinical assessments, including autonomic cardiovascular battery tests, determining whether it provides unique, non-redundant insights into disease progression. Materials and methods Data Collection and Preprocessing MSA patients were enrolled at the Toulouse French MSA Reference Centre between May 2011 and June 2020. Eligible patients met the current consensus criteria for a probable or possible MSA diagnosis, including both parkinsonian (P) and cerebellar (C) variants[ 19 ] and underwent a cardiovascular autonomic reflex assessment, including HRV (time-domain, frequency-domain, and nonlinear metrics) measured at rest in the supine position and autonomic CV testing as initially described by Ewing. This study was conducted in accordance with the Declaration of Helsinki and approved by the Toulouse University Hospital Ethics Committee as part of a broader prospective longitudinal study on natural MSA progression (CNIL 1338780, CCTIRS 10.065).[ 20 ] All patients were evaluated by a movement disorder specialist on the same day as the cardiovascular autonomic assessment, which was considered the baseline visit. Data collection included (on the same day) demographic information, medical history, neurological examination, diagnostic certainty and subtype classification based on consensus criteria,[ 19 ] and the UMSARS I-II-III and IV scores, COMPASS-31 and SCOPA-AUT.[ 21 – 23 ] Patients were followed until death or until censoring on December 31st 2022. Cardiovascular Autonomic Tests Patients underwent a standardized autonomic laboratory evaluation using continuous beat-to-beat digital blood pressure (BP) (Nexfin®, BMEYE, or Finapres® NOVA, FMS, The Netherlands) and electrocardiogram (ECG) recordings (LabChart, ADInstruments, Oxford, United Kingdom). The assessment included four tests performed in a controlled environment in the morning, following a 5-minute rest in a supine position. The tests were conducted in a fixed order under identical conditions and lasted approximately 45 minutes in total. The following tests were performed (as initially described by Ewing)[ 24 ]: → Deep Breathing Test (DB): Six deep breaths per minute in a supine position. → Valsalva Maneuver (VM): Expiratory pressure of 40 mmHg for 15 seconds while supine. → Head-Up Tilt Test (HUTT): Passive tilt to 80° for 10 minutes with BP recorded every minute using a sphygmomanometer (arm cuff). → Stand Test (ST): Five minutes of standing with BP recorded every minute using a sphygmomanometer (arm cuff) following 5 min in supine position. → Isometric Handgrip Test (HG): Three minutes of sustained handgrip exercise performed while seated. The total cardiovascular score[ 25 ] was derived from key parameters, including heart rate (HR) variations during DB (HR-DB), the 30/15 ratio during ST (HR-ST, immediate changes in HR after standing, max RR interval after 30 sec/min RR interval after 15 sec), the Valsalva ratio (HR-VM, max HR in phase II /min HR in phase IV), systolic BP response to ST (BPs-ST), diastolic BP response to ST (BPd-ST), diastolic BP increase during HG (BPd-HG), and systolic/diastolic BP response during HUTT (BPs/d-HUTT). Additional analyses included the maximum drop in systolic BP during VM phase II (BPs-VM-II) and the systolic BP overshoot in VM phase IV (BPs-VM-IV). Changes in BP and HR were assessed against age-specific laboratory normative data, with responses classified as normal (0) or impaired (1). A test result was considered abnormal if it fell below the 5th percentile for age.[ 26 ] The EwS ranged from 0 (normal autonomic function) to 5 (severe autonomic dysfunction). Patients were classified as having cardiovascular autonomic neuropathy (CAN) if they had an EwS of 2 or higher, indicating at least two abnormal test results.[ 24 , 27 ] Heart Rate Variability (HRV) Analysis HRV analysis was performed using the HRVanalysis software, a validated tool for assessing cardiac autonomic activity through non-invasive RR interval variability measurements.[ 28 ] The software incorporates time-domain, frequency-domain, geometrical, and nonlinear methods, adhering to the standards outlined by the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology.[ 29 ] ECG signals were acquired during 5 minutes in the supine position during the cardiovascular test, and the RR intervals were extracted via the HRV analysis software’s built-in R-peak detection algorithm, which utilizes wavelet denoising, adaptive thresholding, and sliding-window summation for high accuracy (> 99.5% detection rate). Ectopic beats, artifacts, and missing intervals were corrected using cubic spline interpolation for sequences ≤ 3 beats and linear interpolation for longer sequences. Segments with excessive noise (> 5% invalid beats) were excluded to ensure signal integrity. Wavelet transform was applied to localize transient autonomic changes, with Morlet wavelets decomposing the RR series into time-resolved LF, HF, and LF/HF ratio components. HRV indices were computed across four domains. A description for each variable and their clinical meaning can be found in Supplementary Table S1 . Baroreflex assessment The baroreflex response was assessed during 5-minute recordings in supine position. The sequence method has been used, based on the scan of beat-to-beat SBP and R-R interval series. Different sequences of 3 or more consecutive heart beats were identified where either the SBP increased and the R-R intervals lengthened, or the SBP decreased and the R-R intervals shortened.[ 30 ] We used at least 3 beats to consider 1 sequence with a minimum threshold of 1 mmHg between 2 BP records and a 5ms interval for the R-R interval. The minimal coefficient correlation used to validate a sequence was r > 0.85. All slopes of the regression line obtained from all sequences were finally averaged to determine the BRS gain.[ 31 ] Statistical Analyses Correlation Analyses Spearman rank correlation coefficients were computed to examine associations between individual HRV features and clinical measures at the same time, including symptom duration (time from symptom onset to HRV assessment), diagnosis duration (time from MSA diagnosis to HRV assessment), and UMSARS scores (total score combining UMSARS Part I activities of daily living and Part II motor examination). A correlation matrix was visualized using a heatmap to identify patterns of association and potential multicollinearity among HRV features. Local weighted regression A nonparametric locally weighted regression (LOESS) analysis was performed to visualize potential nonlinear relationships between HRV features and UMSARS scores. Scatterplots with smoothed LOESS curves were generated to assess trends in the data. Regularized Regression Models for the UMSARS Given the high-dimensional nature of the HRV dataset and potential multicollinearity, regularized regression models were employed to predict total UMSARS (I + II) scores. LASSO regression (Least Absolute Shrinkage and Selection Operator) uses L1 regularization to shrink some coefficients to zero, performing feature selection by retaining only the most relevant predictors. Ridge regression uses L2 regularization to shrink all coefficients toward zero, mitigating collinearity while retaining all predictors. Both models were tuned using 10-fold cross-validation, optimizing the regularization parameter (λ) to minimize mean squared error (MSE). The coefficients from standardized features were interpreted, with positive values indicating predictors associated with higher total UMSARS I + II scores and negative values indicating protective effects. Survival Analysis for Mortality and Regularized Regression for Time-to-death Kaplan-Meier survival analysis was conducted and survival probabilities were estimated/visualized as a function of follow-up time (time from HRV assessment to death or censoring). A Cox proportional hazards model was fitted to assess the association between individual HRV features and mortality. The survival object was defined using time-to-event (time from HRV assessment to death or censoring) and event status (0 = censored, 1 = death). To improve interpretability and multidimensionality, regularized Cox regression models (LASSO and Ridge) were implemented. These models identified the most predictive HRV features for mortality risk while controlling for overfitting. The optimal λ was selected via 10-fold cross-validation, and hazard ratios (HRs) were reported for retained features. To determine whether HRV features improved mortality prediction beyond clinical severity measures, we developed three models: HRV-only model consisting of LASSO regression trained on HRV features to predict time-to-death, UMSARS-only model consisting of a simple linear regression using UMSARS score as the sole predictor, and combined model (HRV + UMSARS) consisting of a LASSO regression including both HRV features and UMSARS scores as possible predictors. Model performance was evaluated using mean squared error (MSE) and R² values. The added predictive value of HRV features was assessed by comparing the combined model to the UMSARS-only model. Mediation Analyses To investigate whether the relationship between HRV and mortality was mediated through UMSARS (a clinical measure of disease severity), a causal mediation analysis was performed. The independent variable was the HRV-derived risk scores (from LASSO/Ridge models), the mediator was the UMSARS total score, the outcome was the time-to-death. These mediation analyses employed a quasi-Bayesian approximation with 1,000 simulations to estimate the Average Causal Mediated Effect (ACME) or indirect effect of HRV on mortality through UMSARS, the Average Direct Effect (ADE) or direct effect of HRV on mortality, independent of UMSARS, the Total Effect or the sum of ACME and ADE, and the Proportion Mediated or the percentage of HRV’s effect on mortality explained by UMSARS. Neural Network Models To explore potential nonlinear relationships, a feed-forward neural network was developed to predict UMSARS scores and systolic blood pressure (SBP) deltas (maximum observed drop), as assessed during the UMSARS-III test (disease severity and orthostatic hypotension proxies). For the model architecture, the input layer consisted of 50 nodes (one for each HRV feature), the hidden layers consisted of ReLU-activated layers with dropout regularization (rate = 0.2–0.5) and the output layer consisted of a single node for prediction (linear activation). For the training process, the Adam (for UMSARS prediction) or RMSprop (for blood pressure prediction) optimizers were chosen. As the loss function, the Mean Absolute Error (MAE) was evaluated. Early stopping was enforced to prevent overfitting by halting training when validation performance plateaued. To interpret model predictions, SHapley Additive exPlanations (SHAP) were used to quantify the contribution of each HRV feature. Performance was evaluated via scatter plots, kernel density plots, correlation analysis, and training/validation loss curves. Best Subset Selection To identify the most relevant HRV predictors when it comes to both disease severity and mortality risk, best subset selection was performed. Models were built incrementally from one predictor to all predictors, with selection guided by the adjusted R² (variance explained, adjusted for model complexity), the Mallow’s Cp (valuates bias-variance trade-off), the Bayesian Information Criterion (BIC, penalizes overfitting), and the Residual Sum of Squares (RSS, Measures total prediction error). Autonomic Cardiovascular Testing and Survival Analysis Using strategies identical to those detailed above employed for HRV, the association between cardiovascular function (global score and individual sub-scores for each test) and the total UMSARS (I + II), the SBP delta (UMSARS III), or survival have been accessed. In addition, Cox proportional hazards models were used to determine their association with survival. Results 214 MSA patients were included in this study, consisting of 56% male, 71% MSA-P and 86% probable MSA. Mean (SD) disease duration at autonomic assessment was 4.5 (2.2) years, and UMSARS IV score 2.5 (1.1) (see Table 1 for all clinical details). As the cardiovascular autonomic testing was performed only once per patient, longitudinal changes in HRV over the disease course could only be inferred at the cohort level, with different patients assessed at various stages of disease progression (Table 1 ). Generally, HRV features showed modest negative correlations with disease severity, mirroring clinical deterioration. Notably, some HRV metrics exhibited correlations with disease progression comparable in magnitude to those observed with the total UMSARS I + II score, underscoring their potential clinical relevance. While classical HRV parameters such as those of the temporal and time domains decrease with disease progression, HRV features from the Entropy (Approximate Entropy, Sample Entropy, Shanon Entropy, Conditional Entropy (CE), Corrected CE, Normalized CCE, ρ, Lempel-Ziv Complexity) and Symbolic Dynamics domains (OV, OV%, 1V, 1V%, 2V, 2V%, 2UV, 2UV%, MP, MP%) tended to increase with disease progression, reflecting domain-specific patterns of change over time ( Supplementary Figs. 1, 2 and 3 ). Table 1 Summary Clinical and Demographic Features (n = 214) Feature Mean (± SD) | Median [Q1-Q3] | Proportion % Clinical Demographic Gender 56% Male MSA Prob vs Pos 86% Prob | 14% Pos MSA-C vs MSA-P 29% MSA-C | 71% MSA-P Age (years) 64.2 (± 7.9) | 64.3 [59.2–70.1] % Died 188 (88%) Follow-up Duration 3.6 (± 2.0) | 3.4 [2.1-5.0] Disease Duration 4.5 (± 2.2) | 4.0 [3.0–6.0] Total UMSARS 1 22.7 (± 7.2) | 22.0 [18.0–27.0] Total UMSARS 2 26.1 (± 7.7) | 25.0 [20.0–31.0] Max D SBP Drop 35 (± 20) | 32 [ 23 – 47 ] Max D DBP Drop 18 (± 13) | 15 [ 10 – 24 ] Total UMSARS 4 2.5 (± 1.0) | 2.0 [2.0–3.0] Total COMPASS 34.3 (± 16.2) | 30.4 [19.7–47.3] Total SCOPA 19.8 (± 11.2) | 20.0 [13.0–27.0] BRS ms/mmHg 4.0 (± 3.3) | 2.8 [1.9–5.3] Temporal Mean RR (ms) 798 (± 116) | 785 [722–871] pNN50 (%) 3.0 (± 9.5) | 0.0 [0.0-0.7] SDNN (ms) 19.5 (± 11.3) | 16.5 [11.9–24.6] rMSSD (ms) 15.4 (± 15.6) | 11.5 [7.5–16.9] Frequential Ptot (ms²) 420 (± 607) | 215 [109–455] VLF (ms²) 186 (± 269) | 107 [46–214] LF (ms²) 98 (± 174) | 40 [17–102] HF (ms²) 83 (± 229) | 25 [11–58] LF/HF 2.6 (± 2.9) | 1.6 [0.9–3.2] Empirical Decomposition pLF1 (ms²) 59.5 (± 89.1) | 22.9 [9.9–68.5] pLF2 (ms²) 71.3 (± 107.6) | 38.7 [14.9–83.0] pHF1 (ms²) 69.0 (± 141.3) | 22.2 [7.5–59.3] pHF2 (ms²) 149.7 (± 487.8) | 36.6 [13.9–92.3] IMAI1 0.57 (± 0.63) | 0.39 [0.18–0.78] IMAI2 0.90 (± 1.22) | 0.50 [0.24–1.03] Geometrical Triangular index 7.2 (± 2.3) | 6.5 [5.6–8.1] TINN (ms) 112 (± 35) | 102 [86–125] X (ms) 794 (± 127) | 789 [719–867] Y (beats) 58 (± 19) | 58 [44–70] M (ms) 855 (± 124) | 836 [773–922] N (ms) 743 (± 111) | 734 [672–805] Entropy Approximate Entropy 1.10 (± 0.13) | 1.13 [1.04–1.19] Sample Entropy 1.34 (± 0.34) | 1.34 [1.10–1.56] Shanon Entropy 3.31 (± 0.58) | 3.43 [2.97–3.69] Conditional Entropy 0.88 (± 0.20) | 0.91 [0.75–1.03] Corrected CE 0.90 (± 0.22) | 0.92 [0.75–1.06] Normalized CCE 0.64 (± 0.12) | 0.64 [0.56–0.72] Ρ 0.36 (± 0.12) | 0.36 [0.28–0.44] Lempel-Ziv Complexity 0.96 (± 0.11) | 0.98 [0.90–1.03] Poincaré plot Centroid (ms) 798 (± 116) | 785 [722–871] SD1 (ms) 10.5 (± 9.9) | 8.1 [5.3–11.1] SD2 (ms) 24.1 (± 12.6) | 21.3 [15.1–30.9] SD1/SD2 0.44 (± 0.29) | 0.37 [0.28–0.52] Symbolic Dynamics OV 123 (± 77) | 109 [64–175] OV% 32 (± 19) | 29 [ 17 – 45 ] 1V 156 (± 39) | 159 [132–180] 1V% 41 (± 9) | 42 [ 37 – 47 ] 2V 21 (± 22) | 15 [ 6 – 27 ] 2V% 6 (± 6) | 4 [ 1 – 8 ] 2UV 79 (± 44) | 69 [46–106] 2UV% 21 (± 12) | 19 [ 12 – 29 ] MP 160 (± 19) | 163 [149–176] MP% 43 (± 8) | 43 [ 39 – 48 ] Fractal Α1 (DFA) 0.98 (± 0.35) | 0.98 [0.76–1.20] Α2 (DFA) 1.05 (± 0.20) | 1.06 [0.94–1.16] H (DFA) 0.99 (± 0.16) | 0.99 [0.89–1.09] H (Higuchi) 1.81 (± 0.18) | 1.82 [1.71–1.93] H (Katz) 1.67 (± 0.45) | 1.59 [1.43–1.74] Hurst 0.21 (± 0.14) | 0.21 [0.11–0.30] Ewing Ewing Valsalva Ratio 1.2 (± 0.2) | 1.2 [1.1–1.3] Ewing respiratory amplitude (bpm) 5.3 (± 3.3) | 4.0 [3.0–7.0] Ewing 30/15 ratio 1.1 (± 0.1) | 1.0 [1.0-1.1] Ewing Iso DBP Handgrip 9.6 (± 7.0) | 9.0 [4.0–14.0] Ewing Iso SBP Handgrip 15.4 (± 12.4) | 13.0 [5.5–22.0] Ewing SBP Tilting Delta 26.4 (± 21.0) | 23.0 [13.0–38.0] Ewing DBP Tilting Delta 11.1 (± 13.0) | 9.0 [3.0–19.0] Ewing Orthost stand SBP 23.8 (± 22.1) | 22.0 [8.0–35.0] Ewing Orthost stand DBP 10.8 (± 14.7) | 9.0 [0.0-18.3] Ewing Valsalva Score 0.7 (± 0.4) | 1.0 [0.5-1.0] Ewing respiratory score (bpm) 0.6 (± 0.5) | 1.0 [0.0–1.0] Ewing 30/15 0.5 (± 0.4) | 0.5 [0.0–1.0] Ewing Iso Score 0.7 (± 0.4) | 1.0 [0.0–1.0] Ewing Orthost Tilt 0.7 (± 0.4) | 1.0 [0.5-1.0] Ewing Total Score 3.0 (± 1.3) | 3.0 [2.0–4.0] Valsalva SBP Delta phase_IIb (mmHg) 35.9 (± 22.7) | 35.0 [19.3–50.7] Valsalva SBP Delta phase_Ivb (mmHg) 7.7 (± 8.7) | 5.0 [0.0–12.0] HRV and the UMSARS Correlation Analyses As expected, initial correlation analysis revealed widespread multicollinearity among HRV features, with many exhibiting high intercorrelations. However, correlations between individual HRV features and total UMSARS I + II scores were uniformly weak, indicating no strong linear relationships. While some HRV features had modest associations with total UMSARS I + II, none demonstrated an extremely strong predictive signal (maximal correlation of ~ 0.2 for the strongest correlation HRV features). These findings suggested that HRV features alone may not directly reflect disease severity as measured by total UMSARS I + II but could still contain clinically relevant information when combined. To explore potential nonlinear trends, locally weighted regression (LOESS) was applied to visualize the association between HRV features and UMSARS scores. Consistent with the correlation findings, no clear patterns emerged across individual HRV features, reinforcing the idea that no single HRV feature displayed a strong predictive relationship with UMSARS ( Supplementary Figs. 1, 2 and 3 ). Regularized Regression Models Given the challenges posed by multidimensionality and weak linear correlations, Ridge (L2) and LASSO (L1) regression were employed to refine the analysis. These methods address overfitting by penalizing coefficient magnitudes, with LASSO selecting a subset of the most relevant features and Ridge shrinking all coefficients to stabilize the model. LASSO identified Lempel-Ziv Complexity and various entropy-based features as the most informative predictors of the total UMSARS I + II scores. Ridge regression retained all predictors but confirmed that complexity-based and entropy-based metrics were the most relevant. Despite these refinements, the retained HRV features still exhibited only modest relationships with total UMSARS I + II scores, aligning with earlier analyses (Fig. 1 ). While biological age and gender themselves were also informative variables at predicting the UMSARS, adjusting for these did not noticeably change which HRV features were the most informative ones ( Supplementary Fig. 4 ). Best Subset Selection To systematically evaluate different combinations of HRV features, best subset selection was employed. This method iteratively tested models of increasing complexity while balancing predictive accuracy. The most informative HRV features were similar to those identified by regularized models, including Lempel-Ziv Complexity and several entropy-based features. Models with ~ 20 HRV features achieved an adjusted R² of 0.4, translating to a correlation of 0.6 between predicted and observed total UMSARS I + II scores using HRV features alone. (once again, primarily of the Entropy and Symbolic Dynamics domains, followed by other domains to a smaller extent - e.g., Fractal domains with H Higuchi, H DFA or Hurst). This suggests that while HRV features contain useful information when it comes to disease severity, their predictive strength remains modest at best (Fig. 1 ). Adjusting for biological age and gender (e.g., if also including these in the models) did not meaningfully increase the predictive performance (correlation remained at ~ 0.6–0.7 for a peak R 2 with ~ 20 variables) and the same (primarily complexity-based and entropy-based) HRV features remained as the most relevant ones ( Supplementary Fig. 4 ). Neural Network Model To capture potential nonlinear relationships, a feed-forward neural network was implemented. The model architecture was optimized for hyperparameters, dropout regularization, and activation functions. SHapley Additive exPlanations (SHAP) analysis revealed that Entropy, Symbolic Dynamics, and Fractal domains were the most contributing towards these predictions (e.g., 1V, Sample Entropy, or Approximate Entropy negatively contributing to the predicted UMSARS scores; Lempel-Ziv Complexity, 2V%, H DFA, positively contributing to the predicted UMSARS scores). Despite capturing potential nonlinearities, the model reaffirmed the limited predictive power of HRV features alone in assessing disease severity as classical defined using the UMSARS ( Supplementary Fig. 5 ). Across models, higher Lempel-Ziv Complexity, reflecting increased sequence complexity, and greater detrended fluctuation analysis (DFA) values, indicating enhanced self-similarity (fractality) of the RR interval signal, were associated with higher UMSARS scores. Conversely, entropy-based measures were inversely associated with UMSARS. HRV and Blood Pressure Regulation Regularized Regression Models To investigate the relationship between HRV features and SBP delta (as assessed during UMSARS III test) Ridge and LASSO regression were applied. LASSO identified HRV features of the Symbolic Dynamics, Fractal, and Empirical Decomposition domains as the most informative features. Entropy HRV features were still relevant, but relatively less so as compared to the UMSARS. Ridge and LASSO achieved an R² of ~ 0.3 (correlation ~ 0.5), suggesting moderate predictive power ( Supplementary Figs. 6 and 7 ). Best Subset Selection The best models correlated modestly (r = 0.6) with observed SBP deltas using ~ 25 different HRV features, with once again the most import ones belonging primarily to the Symbolic Dynamics (1V, 2UV%), Fractal (e.g., H Katz), and Entropy (e.g., Shannon Entropy) domains ( Supplementary Figs. 6 and 7 ). Neural Network Model A neural network was implemented to assess nonlinear associations between HRV and SBP deltas, with modest predictive accuracy (~ 0.4 correlation). SHAP analysis again features of the Symbolic Dynamics (1V, 2V%), Fractal (e.g., H Hurst, A1 DFA), and Entropy (e.g., Approximate Entropy) domains to be the key predictors. Even under more flexible modelling assumptions, HRV alone provided limited predictive value for blood pressure regulation ( Supplementary Fig. 8 ). HRV and Mortality/Time-to-death Kaplan-Meier Survival Analysis A total of 87% of the cohort patients have died, with a median survival time of 40 months (95% C.I. 39–51). Some HRV features were associated with time-to-death, with scores in most features across the Fractal, Symbolic Dynamics, and Entropy domains being associated with higher chances of survival and/or longer time-to-death (Figs. 2 and 3 ). Cox Proportional Hazards Model Original (imbalanced) dataset in LASSO and Ridge models identified deceased patients with 80–90% accuracy [ (true positives + true negatives) / total observations, using > 0.5 binarized risk scores ], but this was biased due to class imbalance. After randomly up-sampling the minority class, the model accuracy dropped to ~ 70%, providing a more realistic estimate of HRV’s predictive power (Figs. 2 and 3 ). Time-to-Death Prediction HRV features alone predicted time-to-death almost as well as total UMSARS I-II scores alone. The LASSO Model with HRV features only retained the 1V% and H Higuchi features, with a correlation coefficient of ~ 0.35, similar to that of UMSARS-only models. When combining all HRV features + total UMSARS I + II scores, improved predictive accuracy was obtained (correlation of 0.5 for the HRV + UMSARS vs. 0.4 for the UMSARS alone), suggesting that HRV and UMSARS capture partially complementary survival-related information. With slightly more complex models (see best subset regression, 10–20 variables, peaking at ~ 15 HRV variables), a higher correlation can be obtained (correlation of ~ 0.6). Of the time domain features, the pNN50(%) was the most important one, a marker of the parasympathetic drive with higher values suggesting higher vagal tone, thus inversely correlated with mortality. When investigating the information gain obtained by also including the maximum delta in SBP (UMSARS III) and symptom duration, these two features were not among the most informative ones at the overall performance at predicting the time-to-death was not meaningfully increased (~ 0.6 correlation between predicted and observed values, Figs. 2 , 3 and 4 , Supplementary Figs. 9 and 10 ). As observed for the UMSARS, when it comes to time-to-death, biological age and gender themselves were herein observed to be very good features at predicting survival but adjusting for these did not meaningfully increase the predictive performance (remaining at ~ 0.6 for a peak R 2 with ~ 15 variables). This observation reflects the fact that the most important HRV features also change with ageing, accounting for a significant degree of redundancy. Regardless of the age/gender adjustment, the most important HRV features remained the same (primarily those of the Fractal, Symbolic Dynamics, and Entropy domains, Supplementary Fig. 11 ). Mediation Analysis In mediation analysis, HRV had a direct contribution towards time-to-death predictions (ADE = 0.0744, p < 2×10⁻¹⁶), with the total UMSARS I + II scores presenting only a minimal residual mediating effect on HRV’s influence on survival (ACME = 0.0007, p = 0.17). As such, HRV features capture significantly complementary physiological dimensions, beyond clinical severity otherwise classically measured by the UMSARS (Fig. 3 ). HRV and Autonomic Cardiovascular Testing Worse autonomic cardiovascular test scores (historically known as Ewing scores) were significantly associated with shorter survivals (HR = 1.32, p = 0.003 for the global “Ewing” score), with a relatively high redundancy between its different sub-scores being observed. Of the different sub-scores, those pertaining to the Valsalva maneuver (both exacts changes and Valsalva ratio) and to respiratory amplitude were the most informative. Moreover, features pertaining to DBP (diastolic blood pressure) are consistently more informative than those pertaining to the SBP (systolic blood pressure). Not surprisingly given that the orthostatic tests are also part of the overall cardiovascular testing, HRV autonomic testing scores had a poor performance at mirroring the total UMSARS I + II but a high accuracy at mirroring the SBP delta (correlation ~ 0.7) (Fig. 5 and Supplementary Figs. 12 and 13 ). The classical cardiovascular assessment tools (as referred to as “Ewing scores”) provide comparable/identical correlatiom value to that of HRV features and their combination is unlikely to provide much added value (Root Mean Squared Error (RMSE) [RMSE] ~ 12–14, MAE ~ 9–10 at regression with the UMSARS for both classical CV alone, HRV alone, or both combined; RMSE ~ 23–24, MAE ~ 18–20 at regression with the time-to-death in months for both classical CV alone, HRV alone, or both combined, Supplementary Fig. 14 , with RMSE being the Root Mean Squared Error that calculates the square root of the average of the squared differences between predicted and actual values, which penalizes larger errors more heavily, and MAE being the Mean Absolute Error that computes the average of the absolute differences, treating all errors equally regardless of size). Of note, the “classical cardiovascular assessment” herein referred to comprises the extended battery of cardiovascular tests and not merely the orthostatic hypotension assessment performed with the UMSARS III (whose performance was in turn considerably inferior to HRV). Discussion In this large cohort of 214 MSA patients, we show that HRV features—particularly those from the Entropy, Symbolic Dynamics, and Fractal domains—provide complementary and partially independent information from traditional clinical assessments. While HRV alone showed only modest associations with disease severity as measured by UMSARS I + II, it demonstrated comparable prognostic power to classical cardiovascular autonomic tests and significantly improved survival prediction when combined with clinical scales. Importantly, HRV features predicted time-to-death nearly as well as UMSARS scores alone, and their combination yielded improved accuracy, supporting the relevance of HRV as a prognostic biomarker. Mediation analysis confirmed that HRV contributes directly to survival prediction, independently of UMSARS. These findings support the use of HRV as a non-invasive, physiologically meaningful marker that could complement clinical scales in MSA research and care. The dynamic responses of HR and BP during autonomic cardiovascular testing, illustrated by the total score and the sub-scores for the different autonomic tests were altered as already described in MSA patients.[ 32 , 33 ] In addition, we observed that the severity of both parasympathetic (heart rate changes during deep breathing and the Valsalva ratio) and sympathetic (blood pressure responses to the Valsalva maneuver and orthostatic tests) autonomic failure, as assessed through autonomic testing, had prognostic value for survival. Notably, the prognostic value of comprehensive autonomic testing was superior to that of orthostatic hypotension severity alone, as measured by clinical evaluation (UMSARS III). The objective of this study was to evaluate the prognostic value of short-term HRV analysis, based on 5-minute recordings in the supine position, for survival prediction, and to identify the HRV features most strongly associated with mortality. Our findings reveal that while HRV alone has only moderate predictive power for disease severity and survival, it captures physiological signals highly complementary to UMSARS. This may suggest that HRV is not merely an auxiliary metric but a valuable, independent biomarker that can refine risk stratification, prognosis, and potentially even therapeutic decision-making. HRV as a Predictor of Disease Severity The weak correlation between HRV features and UMSARS suggests that autonomic dysfunction in MSA does not always parallel clinical severity as measured by conventional motor and functional scales (i.e., UMSARS). This aligns with previous studies showing that while autonomic impairment is a hallmark of MSA, its severity does not necessarily track with motor decline.[ 4 , 5 , 34 – 36 ] In turn, alternative/complementary data such as entropy- and complexity-based HRV measures emerge as more relevant predictors, indicating that HRV dynamics may better capture the subtle physiological alterations associated with disease progression. Higher Lempel-Ziv Complexity, reflecting greater sequence complexity, and increased detrended fluctuation analysis (DFA), indicating stronger self-similarity of the RR interval signal, were associated with higher UMSARS scores. In contrast, lower entropy-based measures, reflecting reduced signal unpredictability, were associated with higher UMSARS scores. Even with advanced regression models and neural networks, HRV alone achieved only moderate predictive accuracy (adjusted R² ~ 0.4, corresponding to a correlation of ~ 0.6). This reinforces the idea that HRV should be viewed as a complementary physiological indicator that enriches clinical assessments. Expanding the Dimensions of Prognostication and Stratification This study has underscored the independent prognostic value of some HRV parameters mainly assessed by non-linear approaches for survival. The combined HRV + UMSARS model outperformed UMSARS alone, demonstrating that these HRV parameters capture critical physiological processes beyond traditional clinical evaluations. Nonlinear methods of signal analysis can be more useful when characterizing complex dynamics. Cox proportional hazards modelling identified specific HRV features—such as Higuchi’s fractal dimension (H Higuchi) and Lempel-Ziv complexity—as significant predictors of survival. Mediation analysis confirmed that HRV’s impact on survival was largely independent of clinical severity, further underscoring its role as a unique autonomic biomarker with real prognostic relevance. In line with the strong positive correlation between the Higuchi’s index and mortality (plus inverse correlation with time-to-death) herein observed in the case of MSA, an increase in the Higuchi’s index has been previously reported in stroke patients,[ 37 ] and in diabetic patients (compared to non-diabetic),[ 38 ] an observation interpreted as impairment of the autonomic nervous system. An increase in Higuchi’s index has been suggested in atrial fibrillation.[ 39 ] In contrast, other indicators of signal complexity such as the Hurst index ( autocorrelation/signal persistence over time) the Lempel-Ziv (diversity and number of signal sub-patterns), or the 1V% (ordinal variability and phase patterns), were herein observed to present a positive correlation with survival. A decrease in Lempel-Ziv complexity has been previously reported during Terbutaline (selective Beta2-adrenoceptor agonist) infusion, which decreases the parasympathetic drive of heart rate.[ 40 ] The decrease in HRV signal complexity and the increase in the fractal dimension may suggest a reduced central control of HRV. To our knowledge, no study has specifically analyzed Higuchi’s fractal dimension during atrial fibrillation. However, by analogy with other non-linear indices - such as Approximate Entropy - which have been shown to vary similarly to Higuchi’s fractal dimension across different populations, the elevated values observed in our study population appear to reflect a pattern consistent with autonomic dysregulation, as typically seen in atrial fibrillation. Supporting this interpretation, Xin et al. (2017) reported higher entropy values during paroxysmal atrial fibrillation episodes compared to periods of sinus rhythm.[ 41 ] Similarly, Yamada et al. (2000) found Approximate Entropy values around 1.85 in patients with atrial fibrillation, whereas typical Approximate Entropy values in healthy individuals in sinus rhythm are approximately 1.05.[ 42 , 43 ] These findings also suggest that HRV could serve as an objective tool for risk stratification, potentially guiding early interventions and improving patient management. Autonomic testing is known to provide an independent tool for both the diagnosis and prognosis of MSA. Likewise, cardiovascular autonomic failure and orthostatic hypotension are both good prognostic markers of poor survival in MSA.[ 44 , 45 ] In our study, the addition of orthostatic hypotension measurements to the HRV added only a marginal increase in information, owning to the high correlation between the two domains as markers of autonomic failure. In turn, HRV alone was superior to orthostatic hypotension assessment alone. However, a single clinical assessment does not capture the substantial BP variability often observed in these patients, which has also been identified as a risk factor for mortality.[ 46 ] Further studies are warranted to evaluate the prognostic value of BP fluctuations under sympathetic control, using methods such as ambulatory blood pressure monitoring in combination with HRV analysis. Toward Real-Time, Automated HRV Monitoring One of the promising implications of this study is the potential for real-time, ambulatory HRV monitoring using wearable electronic devices, which is likely more applicable to HRV assessment than to the standard autonomic cardiovascular test battery. HRV can be continuously and non-invasively monitored with modern digital health technologies—offering an unprecedented opportunity for automated, real-time disease tracking. The leveraging of AI-integrated biosensors that continuously capture HRV dynamics is expected to significantly improve the breath and quality of real-time feedback provided to clinicians and researchers. This is expected to enable proactive, data-driven disease management, allowing for earlier detection of disease worsening and more personalized interventions, whenever available. However, the use of this non-linear approach to HRV analysis requires further investigation to better understand the underlying physiological mechanisms. Complementary Endpoints Another putative implication of these findings lies in their potential impact on clinical trials. Traditional MSA trial endpoints often rely on subjective, symptom-based scales such as the patient-completed UMSARS Part I, which may lack sensitivity to subtle physiological changes. While regulatory agencies prioritize patient-centered outcomes, integrating HRV as an exploratory endpoint—particularly in early-phase (e.g., Phase 2) trials—could offer a sensitive measure of physiological response, complementing conventional clinical scales. Over time, if HRV demonstrates consistent association with clinically meaningful outcomes, it could evolve into a surrogate marker. This is especially relevant in a heterogeneous condition like MSA, where current tools may not fully capture the biological impact of interventions. Outside clinical trials, HRV may also serve a valuable prognostic role by improving prediction of mortality and disease trajectory. Future studies could explore HRV-guided patient stratification, identifying autonomic phenotypes that may respond differently to targeted interventions. A Risk Marker or a Modifiable Risk Factor? An important unresolved question is whether altered HRV in MSA is merely a risk marker or also a modifiable risk factor. If HRV is just a marker, it provides valuable prognostic information but does not necessarily influence disease progression. However, if HRV is a true risk factor, then interventions that modulate autonomic function - such as pharmacological agents, neuromodulation - may have the potential to alter disease trajectory. Future interventional studies should explore whether targeting autonomic dysfunction could slow disease progression, improve quality of life, or even extend survival. If so, HRV could move beyond a diagnostic and prognostic biomarker to become a therapeutic target. The Future of Large-Scale, Ambulatory Data Collection To fully harness the power of HRV in MSA, large-scale, ambulatory and stress exposure studies are needed. Current research relies on controlled, in-clinic HRV measurements, which may not fully capture the dynamic nature of autonomic dysfunction in daily life. Expanding HRV analysis to real-world, long-term monitoring through integrated sensors could improve how we study and manage neurodegenerative diseases, including more prevalent conditions such as Alzheimer's and Parkinson’s disease (e.g., [ 47 – 51 ]). A large-scale HRV data repository could enable the development of predictive models to refine risk stratification and provide personalized disease trajectories across a spectrum of neurodegenerative disorders. Conclusion HRV-based prognostic estimation using non-linear features provided performance comparable to classical cardiovascular assessments. Moreover, HRV significantly enhanced the prognostic utility of the UMSARS clinical scale. These findings establish HRV as a valuable autonomic biomarker offering independent prognostic information in MSA. Although HRV alone does not strongly predict clinical severity as traditionally defined, its added value lies in its simplicity, speed, and feasibility—requiring only a brief (e.g., 5-minute) heart rate recording using easy-to-deploy equipment, in contrast to more complex tools or prolonged monitoring. Its potential for real-time, automated data capture further supports its relevance in both research and clinical settings. While the current analysis focused solely on baseline HRV, future work should explore longitudinal changes to better understand its role in disease progression. Building upon these insights will benefit from integrating HRV into multi-modal biomarker frameworks, refining clinical trial endpoints, and scaling up real-world ambulatory data collection. Abbreviations ACME Average Causal Mediated Effect ADE Average Direct Effect ANS Autonomic nervous system BIC Bayesian Information Criterion BP Blood pressure CAN Cardiovascular autonomic neuropathy CE Conditional Entropy DB Deep Breathing Test DFA Detrended fluctuation analysis ECG Electrocardiogram HG Isometric Handgrip Test HR Hazard ratio HUTT Head-Up Tilt Test HRV Heart rate variability LOESS Locally weighted regression MSA Multiple system atrophy OH Orthostatic hypotension RSS Residual Sum of Squares RMSE Root Mean Squared Error SHAP SHapley Additive exPlanations ST Stand Test UMSARS Unified Multiple System Atrophy Rating Scale VM Valsalva Maneuver Declarations Ethics approval and consent to participate This study was conducted in accordance with the Declaration of Helsinki and approved by the Toulouse University Hospital Ethics Committee as part of a broader prospective longitudinal study on natural MSA progression (CNIL 1338780, CCTIRS 10.065). Consent for publication All patients have consented to be part of this study and have their data published. Competing interests Margherita Fabbri received Honoraria to speak from AbbVie, ORKYN, and BIAL, consultancies from BIAL and LVL Medical; Grant from France Parkinson, HORIZON 2022 French Ministry of Health and MSA Coalition. Professor Olivier Rascol has received honorarium for scientific advices from Lundbeck, ONO Pharma and TEVA and scientific grants from ARAMISE, MSA Coalition, Lundbeck, ONO Pharma and Takeda. Authors’contributions (1) Research Project: A. Conception, B. Organization, C. Execution; (2) Data Collection: A. Design, B. Recruitment and Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the First Draft, B. Review and Critique. P.B.: 1A, 1B, 1C, 3A, 3B. M.K.: 1B, 2C, 3B. M.F.: 3B. F.R.:1C, 2C, 3B. V.P.: 1C, 2C, 3B. F.O.M.: 3B. C.L. 3B. O.R.: 3B. W.G.M.: 3B. A.F.S.: 3B. D.B.: 3B. C.P.L: 3B. A.P.T.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B. Funding No funding was received for this work. Acknowledgements Several authors of this publication are members of the European Reference Network for Rare Neurological Diseases - Project ID No 101085584. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Data availability All data can be made available upon reasonable request. References Poewe W, Stankovic I, Halliday G, Meissner WG, Wenning GK, Pellecchia MT et al (2022) Multiple system atrophy. Nat Rev Dis Primer 8:56 Foubert-Samier A, Pavy-Le Traon A, Guillet F, Le-Goff M, Helmer C, Tison F et al (2020) Disease progression and prognostic factors in multiple system atrophy: A prospective cohort study. Neurobiol Dis 139:104813 Eschlboeck S, Goebel G, Eckhardt C, Fanciulli A, Raccagni C, Boesch S et al (2023) Development and Validation of a Prognostic Model to Predict Overall Survival in Multiple System Atrophy. Mov Disord Clin Pract 10:1368–1376 Krismer F, Palma J, Calandra-Buonaura G, Stankovic I, Vignatelli L, Berger A et al (2022) The Unified Multiple System Atrophy Rating Scale: Status, Critique, and Recommendations. Mov Disord 37:2336–2341 Palma J-A, Vernetti PM, Perez MA, Krismer F, Seppi K, Fanciulli A et al (2021) Limitations of the Unified Multiple System Atrophy Rating Scale as outcome measure for clinical trials and a roadmap for improvement. Clin Auton Res 31:157–164 Wenning GK, Stankovic I, Vignatelli L, Fanciulli A, Calandra-Buonaura G, Seppi K et al (2022) The Movement Disorder Society Criteria for the Diagnosis of Multiple System Atrophy. Mov Disord 37:1131–1148 Foubert-Samier A, Pavy-Le Traon A, Guillet F, Le-Goff M, Helmer C, Tison F et al (2020) Disease progression and prognostic factors in multiple system atrophy: A prospective cohort study. Neurobiol Dis 139:104813 Kong SDX, Gordon CJ, Hoyos CM, Wassing R, D’Rozario A, Mowszowski L et al (2023) Heart rate variability during slow wave sleep is linked to functional connectivity in the central autonomic network. Brain Commun 5:fcad129 Orini M, Van Duijvenboden S, Young WJ, Ramírez J, Jones AR, Hughes AD et al (2023) Long-term association of ultra-short heart rate variability with cardiovascular events. Sci Rep 13:18966 Furushima H, Shimohata T, Nakayama H, Ozawa T, Chinushi M, Aizawa Y et al (2012) Significance and usefulness of heart rate variability in patients with multiple system atrophy. Mov Disord 27:570–574 Kong SDX, Gordon CJ, Hoyos CM, Wassing R, D’Rozario A, Mowszowski L et al (2023) Heart rate variability during slow wave sleep is linked to functional connectivity in the central autonomic network. Brain Commun 5:fcad129 Memon AA, George EB, Nazir T, Sunkara Y, Catiul C, Amara AW (2024) Heart rate variability during sleep in synucleinopathies: a review. Front Neurol 14:1323454 Heimrich KG, Lehmann T, Schlattmann P, Prell T (2021) Heart Rate Variability Analyses in Parkinson’s Disease: A Systematic Review and Meta-Analysis. Brain Sci 11:959 Iniguez M, Jimenez-Marin A, Erramuzpe A, Acera M, Tijero B, Murueta-Goyena A et al (2022) Heart-brain synchronization breakdown in Parkinson’s disease. Npj Park Dis 8:64 Furushima H, Shimohata T, Nakayama H, Ozawa T, Chinushi M, Aizawa Y et al (2012) Significance and usefulness of heart rate variability in patients with multiple system atrophy. Mov Disord 27:570–574 Maestri R, Pinna GD, Porta A, Balocchi R, Sassi R, Signorini MG et al (2007) Assessing nonlinear properties of heart rate variability from short-term recordings: are these measurements reliable? Physiol Meas 28:1067–1077 Huikuri HV, Perkiömäki JS, Maestri R, Pinna GD (2009) Clinical impact of evaluation of cardiovascular control by novel methods of heart rate dynamics. Philos Trans R Soc Math Phys Eng Sci 367:1223–1238 Sassi R, Cerutti S, Lombardi F, Malik M, Huikuri HV, Peng C-K et al (2015) Advances in heart rate variability signal analysis: joint position statement by the e-Cardiology ESC Working Group and the European Heart Rhythm Association co-endorsed by the Asia Pacific Heart Rhythm Society. Europace 17:1341–1353 Gilman S, Wenning GK, Low PA, Brooks DJ, Mathias CJ, Trojanowski JQ et al (2008) Second consensus statement on the diagnosis of multiple system atrophy. Neurology 71:670–676 Foubert-Samier A, Pavy-Le Traon A, Guillet F, Le-Goff M, Helmer C, Tison F et al (2020) Disease progression and prognostic factors in multiple system atrophy: A prospective cohort study. Neurobiol Dis 139:104813 Sletten DM, Suarez GA, Low PA, Mandrekar J, Singer W (2012) COMPASS 31: A Refined and Abbreviated Composite Autonomic Symptom Score. Mayo Clin Proc. ;87:1196–201 Wenning GK, Tison F, Seppi K, Sampaio C, Diem A, Yekhlef F et al (2004) Development and validation of the Unified Multiple System Atrophy Rating Scale (UMSARS). Mov Disord 19:1391–1402 Visser M, Marinus J, Stiggelbout AM, Van Hilten JJ (2004) Assessment of autonomic dysfunction in Parkinson’s disease: The SCOPA-AUT. Mov Disord 19:1306–1312 Ewing DJ, Campbell IW, Clarke BF (1980) Assessment of Cardiovascular Effects in Diabetic Autonomic Neuropathy and Prognostic Implications. Ann Intern Med 92:308–311 Ewing DJ, Clarke BF (1986) Autonomic neuropathy: its diagnosis and prognosis. Clin Endocrinol Metab 15:855–888 Mathias CJ, Bannister SR (eds) (2013) Autonomic Failure [Internet]. Oxford University Press; [cited 2025 Apr 27]. Available from: https://academic.oup.com/book/24366 Ewing DJ, Martyn CN, Young RJ, Clarke BF (1985) The Value of Cardiovascular Autonomic Function Tests: 10 Years Experience in Diabetes. Diabetes Care 8:491–498 Pichot V, Roche F, Celle S, Barthélémy J-C, Chouchou F, HRVanalysis: A Free Software for Analyzing Cardiac Autonomic Activity. Front Physiol [Internet]. 2016 [cited 2025 Mar 19];7. Available from: http://journal.frontiersin.org/article/ 10.3389/fphys.2016.00557/full Heart rate variability (1996) : standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Circulation 93:1043–1065 La Rovere MT, Pinna GD, Raczak G (2008) Baroreflex Sensitivity: Measurement and Clinical Implications. Ann Noninvasive Electrocardiol 13:191–207 Di Rienzo M, Castiglioni P, Iellamo F, Volterrani M, Pagani M, Mancia G et al (2008) Dynamic adaptation of cardiac baroreflex sensitivity to prolonged exposure to microgravity: data from a 16-day spaceflight. J Appl Physiol 105:1569–1575 Iodice V, Lipp A, Ahlskog JE, Sandroni P, Fealey RD, Parisi JE et al (2012) Autopsy confirmed multiple system atrophy cases: Mayo experience and role of autonomic function tests. J Neurol Neurosurg Psychiatry 83:453–459 Stankovic I, Fanciulli A, Kostic VS, Krismer F, Meissner WG, Palma JA et al (2021) Laboratory-Supported Multiple System Atrophy beyond Autonomic Function Testing and Imaging: A Systematic Review by the MoDiMSA Study Group. Mov Disord Clin Pract 8:322–340 Golden EP, McCreary M, Vernino S (2022) Responsiveness of UMSARS and other clinical measures in a longitudinal structured care clinic for multiple system atrophy. Clin Auton Res 32:477–484 Hu W-Z, Cao L-X, Yin J-H, Zhao X-S, Piao Y-S, Gu W-H et al (2023) Non-motor symptoms in multiple system atrophy: A comparative study with Parkinson’s disease and progressive supranuclear palsy. Front Neurol 13:1081219 Köllensperger M, Stampfer-Kountchev M, Seppi K, Geser F, Frick C, Del Sorbo F et al (2007) Progression of dysautonomia in multiple system atrophy: a prospective study of self‐perceived impairment. Eur J Neurol 14:66–72 D’Addio G, Corbi G, Accardo A, Russo G, Ferrara N, Mazzoleni MC et al (2009) Fractal behaviour of heart rate variability reflects severity in stroke patients. Stud Health Technol Inf 150:794–798 Gomolka RS, Kampusch S, Kaniusas E, Thürk F, Széles JC, Klonowski W (2018) Higuchi Fractal Dimension of Heart Rate Variability During Percutaneous Auricular Vagus Nerve Stimulation in Healthy and Diabetic Subjects. Front Physiol 9:1162 Gao C, Lim ASP, Haghayegh S, Cai R, Yang J, Yu L et al (2025) Reduced Complexity of Pulse Rate Is Associated With Faster Cognitive Decline in Older Adults. J Am Heart Assoc 14:e041448 Kuusela TA, Jartti TT, Tahvanainen KUO, Kaila TJ (2002) Nonlinear methods of biosignal analysis in assessing terbutaline-induced heart rate and blood pressure changes. Am J Physiol-Heart Circ Physiol 282:H773–H781 Xin Y, Zhao Y (2017) Paroxysmal atrial fibrillation recognition based on multi-scale wavelet α-entropy. Biomed Eng OnLine 16:121 Mäkikallio TH, SeppÄNEN T, Niemelä M, Airaksinen KEJ, Tulppo M, Huikuri HV (1996) Abnormalities in Beat to Beat Complexity of Heart Rate Dynamics in Patients With a Previous Myocardial Infarction11This study was supported by grants from the Finnish Foundation for Cardiovascular Research and the Medical Council of the Academy of Finland, Helsinki, Finland. J Am Coll Cardiol 28:1005–1011 Yamada A, Hayano J, Sakata S, Okada A, Mukai S, Ohte N et al (2000) Reduced Ventricular Response Irregularity Is Associated With Increased Mortality in Patients With Chronic Atrial Fibrillation. Circulation 102:300–306 Coon EA, Sletten DM, Suarez MD, Mandrekar JN, Ahlskog JE, Bower JH et al (2015) Clinical features and autonomic testing predict survival in multiple system atrophy. Brain 138:3623–3631 Vichayanrat E, Valerio F, Koay S, De Pablo-Fernandez E, Panicker J, Morris H et al Diagnosing Premotor Multiple System Atrophy: Natural History and Autonomic Testing in an Autopsy-Confirmed Cohort. Neurology [Internet]. 2022 [cited 2025 Apr 28];99. Available from: https://www.neurology.org/doi/ 10.1212/WNL.0000000000200861 Pavy-Le Traon A, Foubert‐Samier A, Ory‐Magne F, Fabbri M, Senard J, Meissner WG et al (2022) Ambulatory blood pressure and drug treatment for orthostatic hypotension as predictors of mortality in patients with multiple system atrophy. Eur J Neurol 29:1025–1034 Natarajan A, Pantelopoulos A, Emir-Farinas H, Natarajan P (2020) Heart rate variability with photoplethysmography in 8 million individuals: a cross-sectional study. Lancet Digit Health 2:e650–e657 Brown AA, Ferguson BJ, Jones V, Green BE, Pearre JD, Anunoby IA et al (2022) Pilot Study of Real-World Monitoring of the Heart Rate Variability in Amyotrophic Lateral Sclerosis. Front Artif Intell 5:910049 Leng Y, Cotton-Clay A, Baron S, Fava L, Johnson K, Easwar V et al (2024) Real-world Study of Heart Rate Variability During Sleep Using Under-mattress Sensors in over 30,000 Individuals. Sleep 47:A208–A208 Zhou J, Huebner G, Liu KY, Ucci M (2024) Heart rate variability, electrodermal activity and cognition in adults: Association with short-term indoor PM2.5 exposure in a real-world intervention study. Environ Res 263:120245 Srinivasan AG, Smith SS, Pattinson CL, Mann D, Sullivan K, Salmon P et al (2024) Heart rate variability as an indicator of fatigue: A structural equation model approach. Transp Res Part F Traffic Psychol Behav 103:420–429 Supplementary Files SupplementaryFigure1.pptx SupplementaryFigure10.pptx SupplementaryFigure11.pptx SupplementaryFigure12.pptx SupplementaryFigure13.pptx SupplementaryFigure14.pptx SupplementaryFigures.docx SupplementaryFigure2.pptx SupplementaryFigure3.pptx SupplementaryFigure4.pptx SupplementaryFigure5.pptx SupplementaryFigure6.pptx SupplementaryFigure7.pptx SupplementaryFigure8.pptx SupplementaryFigure9.pptx SupplementaryTable1.docx Cite Share Download PDF Status: Published Journal Publication published 23 Feb, 2026 Read the published version in Clinical Autonomic Research → Version 1 posted Reviewers agreed at journal 04 Sep, 2025 Reviewers invited by journal 04 Sep, 2025 Editor assigned by journal 10 Aug, 2025 First submitted to journal 08 Aug, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7325643","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":510126385,"identity":"5db7d2e8-16c4-4b01-a747-5295d5693b79","order_by":0,"name":"Paulo Bastos","email":"data:image/png;base64,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","orcid":"","institution":"CHU Toulouse: Centre Hospitalier Universitaire de Toulouse","correspondingAuthor":true,"prefix":"","firstName":"Paulo","middleName":"","lastName":"Bastos","suffix":""},{"id":510126386,"identity":"de88bbe0-f47a-4f68-93ea-775c0593e332","order_by":1,"name":"Marc Kermongant","email":"","orcid":"","institution":"CHU Toulouse: Centre Hospitalier Universitaire de Toulouse","correspondingAuthor":false,"prefix":"","firstName":"Marc","middleName":"","lastName":"Kermongant","suffix":""},{"id":510126387,"identity":"1037554d-d801-4bdd-a314-2886d5d69e47","order_by":2,"name":"Margherita Fabbri","email":"","orcid":"","institution":"CHU Toulouse: Centre Hospitalier Universitaire de Toulouse","correspondingAuthor":false,"prefix":"","firstName":"Margherita","middleName":"","lastName":"Fabbri","suffix":""},{"id":510126388,"identity":"dcb283e3-fabe-4343-aa4f-1c1c4bfcabb1","order_by":3,"name":"Frederic Roche","email":"","orcid":"","institution":"CHU Saint-Étienne: Centre Hospitalier Universitaire de Saint-Etienne","correspondingAuthor":false,"prefix":"","firstName":"Frederic","middleName":"","lastName":"Roche","suffix":""},{"id":510126389,"identity":"5fdb2f26-08d0-476a-9dc7-c186c2a2956b","order_by":4,"name":"Vincent Pichot","email":"","orcid":"","institution":"CHU Saint-Étienne: Centre Hospitalier Universitaire de Saint-Etienne","correspondingAuthor":false,"prefix":"","firstName":"Vincent","middleName":"","lastName":"Pichot","suffix":""},{"id":510126390,"identity":"cd7e6d63-07b4-4bbc-87bc-847c057901d2","order_by":5,"name":"Fabienne Ory-Magne","email":"","orcid":"","institution":"CHU Toulouse: Centre Hospitalier Universitaire de Toulouse","correspondingAuthor":false,"prefix":"","firstName":"Fabienne","middleName":"","lastName":"Ory-Magne","suffix":""},{"id":510126391,"identity":"2c2ac04e-1feb-4c13-a3d1-a8beb4147b62","order_by":6,"name":"Clémence Leung","email":"","orcid":"","institution":"CHU Toulouse: Centre Hospitalier Universitaire de Toulouse","correspondingAuthor":false,"prefix":"","firstName":"Clémence","middleName":"","lastName":"Leung","suffix":""},{"id":510126392,"identity":"fd53a46e-d7ec-45d4-86a1-fa0a8ef5c77d","order_by":7,"name":"Olivier Rascol","email":"","orcid":"","institution":"CHU Toulouse: Centre Hospitalier Universitaire de Toulouse","correspondingAuthor":false,"prefix":"","firstName":"Olivier","middleName":"","lastName":"Rascol","suffix":""},{"id":510126393,"identity":"c61a1b8e-faee-45b4-8a2c-2b05e76b1192","order_by":8,"name":"Wassilios Meissner","email":"","orcid":"","institution":"CHU de Bordeaux: Centre Hospitalier Universitaire de Bordeaux","correspondingAuthor":false,"prefix":"","firstName":"Wassilios","middleName":"","lastName":"Meissner","suffix":""},{"id":510126394,"identity":"c6d9c085-5e92-46b1-a481-dec492dbc1b1","order_by":9,"name":"Alexandra Foubert-Samier","email":"","orcid":"","institution":"CHU de Bordeaux: Centre Hospitalier Universitaire de Bordeaux","correspondingAuthor":false,"prefix":"","firstName":"Alexandra","middleName":"","lastName":"Foubert-Samier","suffix":""},{"id":510126395,"identity":"2eb73fbf-7d14-456a-a4f5-370e3c2fe583","order_by":10,"name":"David Bendetowicz","email":"","orcid":"","institution":"CHU de Bordeaux: Centre Hospitalier Universitaire de Bordeaux","correspondingAuthor":false,"prefix":"","firstName":"David","middleName":"","lastName":"Bendetowicz","suffix":""},{"id":510126396,"identity":"f3b59aa3-cf05-4bd4-b4e4-2de2b40637e6","order_by":11,"name":"Cécile Proust-Lima","email":"","orcid":"","institution":"INSERM U1219: Bordeaux Population Health","correspondingAuthor":false,"prefix":"","firstName":"Cécile","middleName":"","lastName":"Proust-Lima","suffix":""},{"id":510126397,"identity":"a9514bc8-532d-4239-a8d9-9cd991292101","order_by":12,"name":"Anne Pavy-le-Traon","email":"","orcid":"","institution":"CHU Toulouse: Centre Hospitalier Universitaire de Toulouse","correspondingAuthor":false,"prefix":"","firstName":"Anne","middleName":"","lastName":"Pavy-le-Traon","suffix":""}],"badges":[],"createdAt":"2025-08-08 09:17:44","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7325643/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7325643/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10286-026-01190-8","type":"published","date":"2026-02-23T15:58:53+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":91021712,"identity":"d4d6e5c1-8974-4ff2-8149-14c327d501f4","added_by":"auto","created_at":"2025-09-10 18:47:45","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":142984,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFeature selection results for heart rate variability (HRV) features in predicting the UMSARS.\u003c/strong\u003e Ridge (a) and LASSO (b) regression coefficients are shown. Both models were tuned using 10-fold cross-validation, selecting the optimal penalty parameter (lambda) by minimizing mean cross-validated error (MSE). All predictors were standardized before training, a critical step since LASSO and Ridge penalize based on coefficient magnitude. Positive coefficients indicate variables associated with increased UMSARS scores, while negative coefficients indicate a decrease. The magnitude reflects the strength of association, given standardized predictors. LASSO yields sparse solutions, displaying only non-zero coefficients that highlight key variables. Ridge retains all predictors but shrinks coefficients toward zero, making it less interpretable for selection, though still informative about variable importance. Both models assume a linear relationship between predictors and UMSARS, holding other factors constant. Panel (c) shows four diagnostic plots evaluating best subset regression models across varying numbers of predictors. The adjusted R² plot measures explained variance while accounting for model complexity; the peak adjusted R² identifies the most explanatory yet parsimonious model. The Mallow’s Cp plot assesses bias and variance; ideal models have Cp values close to the number of predictors (p), with higher values suggesting overfitting. The BIC curve balances fit and simplicity; the model with the lowest BIC is considered optimal. The Residual Sum of Squares (RSS) plot shows how error declines with more predictors, though gains diminish as model size increases. Panel (d) displays a variable inclusion matrix across best subset models of increasing complexity. Columns represent model sizes (1 variable to full model), and rows represent predictors. Filled cells indicate variable inclusion at each model size, visually summarizing feature importance across models.\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/5009f8acfb113c5f35bac513.jpg"},{"id":91021456,"identity":"83dbc11c-7cfa-4fd7-8f0c-a40bab4e0ee8","added_by":"auto","created_at":"2025-09-10 18:39:45","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":88304,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSurvival analysis and feature selection results for HRV features in predicting mortality. \u003c/strong\u003e(a) Kaplan-Meier survival curve showing the estimated survival probability over time based on follow-up durations. This non-parametric curve includes 95% confidence intervals (shaded area), offering insight into overall survival trends in the cohort. (b) LASSO-regularized Cox model coefficients for HRV features at the optimal lambda (lambda min), selected via 10-fold cross-validation. Non-zero coefficients highlight the most predictive features, with magnitude and direction indicating the strength and nature of their association with mortality risk. (c) Ridge-regularized Cox model coefficients for HRV features, also at the optimal lambda. Unlike LASSO, Ridge retains all variables, shrinking their coefficients to mitigate multicollinearity while preserving relative contribution. Both models demonstrate HRV feature relevance under different regularization strategies. (d) Logistic regression curves depict the probability of death as a function of two HRV features—“Hurst” and “H Higuchi.” Additionally, linear regression lines illustrate the total number of follow-up months predicted by these same features. These visualizations offer insight into individual HRV features’ relationships with both mortality and duration of follow-up. (e) Density plots of LASSO and Ridge Cox model risk scores, stratified by actual death outcome. A clear rightward shift in the distributions for deceased patients indicates higher predicted risk scores. Note that these risk scores reflect relative risk, not absolute probability. Probabilities at specific time points require applying the baseline survival function from the fitted Cox model. To improve model balance and evaluate separation more fairly, the dataset for generating these distributions was up-sampled to equalize the death and censoring classes. This ensured an unbiased assessment of each model’s ability to distinguish between patients who survived and those who did not.\u003c/p\u003e","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/7ce62d7f3edc5ed785b6ce87.jpg"},{"id":91021455,"identity":"48e6c96d-55a5-46c4-80a0-509fa97ad3f7","added_by":"auto","created_at":"2025-09-10 18:39:45","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":96940,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePredictive modeling of time-to-death in MSA patients using UMSARS and HRV features. \u003c/strong\u003e(a) Coefficients from the LASSO model highlight a sparse set of key predictors, including 1V%, Lempel Ziv Complexity, and Higuchi, emphasizing their importance in predicting time-to-death. (b) Coefficients from the Ridge model show all predictors retained, with magnitudes indicating relative importance, though less interpretable due to coefficient shrinkage. (c) Smoothed regression lines compare actual time-to-death with predicted scores from four models: a UMSARS-only linear regression model, Ridge using HRV features, LASSO using HRV features, and LASSO combining HRV and UMSARS. These curves demonstrate the superior fit of the combined model, showcasing the benefit of integrating both clinical and physiological predictors. (d) Time-to-death is plotted against the four selected features from the LASSO combination model using simple linear regressions. These visualizations underscore the individual contributions of key HRV features to survival prediction. (e) A causal mediation analysis diagram illustrates the relationship between LASSO-based time-to-death predictions, the mediator UMSARS, and the outcome. The total effect is decomposed into a direct effect (ADE) and an indirect effect (ACME) via UMSARS. The analysis reveals a significant direct effect of LASSO predictions on time-to-death, while the indirect effect through UMSARS is not statistically significant. The proportion mediated by UMSARS is small and nonsignificant. (f) A similar causal mediation analysis is shown for Ridge-based predictions. Again, the total effect of Ridge predictions on time-to-death is mainly driven by a significant direct effect, with no significant mediation through UMSARS. This suggests that the predictive value of HRV features is largely independent of clinical UMSARS scores.\u003c/p\u003e","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/3c694e2be82ca6ca9e030b66.jpg"},{"id":91022191,"identity":"a6c7c9e1-c9a5-4415-98f0-3475c078dafc","added_by":"auto","created_at":"2025-09-10 18:55:46","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":120008,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTime-to-death best subset regression using HRV and UMSARS.\u003c/strong\u003e (a) This panel presents four diagnostic plots that assess model performance based on the number of variables included. Each plot evaluates a different aspect of model quality. The adjusted R² plot shows the proportion of variance explained, accounting for model complexity. The model with the highest adjusted R² offers the best explanatory power without overfitting. The Mallow’s Cp plot evaluates bias and variance. Ideally, Cp should be close to the number of predictors (p); higher values suggest overfitting. The Bayesian Information Criterion (BIC) curve balances model fit and parsimony. The model with the lowest BIC achieves the best trade-off between simplicity and explanatory power. The Residual Sum of Squares (RSS) curve measures total residual error. RSS decreases as more predictors are added, but diminishing returns appear as model size increases, highlighting the cost of added complexity. (b) The variable inclusion matrix summarizes which variables are selected across models of increasing complexity. Columns represent model sizes, from single-variable models to those using all predictors (-1), while rows correspond to individual variables. Filled cells indicate inclusion of a variable at a given model size. This matrix allows for quick identification of variables that are consistently selected across different model sizes. Variables chosen in small models and retained in larger ones likely offer strong predictive value. Conversely, variables that appear only in large models may offer marginal benefits or be dependent on others for their contribution. The matrix also highlights redundancy or interactions between variables, as some may only appear when specific others are included. This visualization supports interpretability and helps identify the most robust predictors of time-to-death when combining HRV features and UMSARS.\u003c/p\u003e","description":"","filename":"Figure4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/7d54a7bbba01e38bba316664.jpg"},{"id":91021713,"identity":"4b2b49ba-fda9-4950-ab84-2a3021a46089","added_by":"auto","created_at":"2025-09-10 18:47:46","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":43534,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSurvival probability as a function of the Ewing Score\u003c/strong\u003e. (a) Cox proportional hazard models were used to assess the relationship between the Ewing Score and survival time. The survival curves illustrate the probability of survival over follow-up time, stratified by binned Ewing Score values. For the Ewing Total Score, the Cox regression revealed a significant association with survival (HR=1.32, 95% CI [1.10–1.59], p=0.003). Higher Ewing scores, indicative of worse autonomic function, were associated with shorter survival times. The concordance index for the model was 0.586, indicating moderate predictive accuracy. (b) Forest plot depicting the HR for each of the Ewing score sub-scores, calculated using scaled data.\u003c/p\u003e","description":"","filename":"Figure5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/316d974d8a19a76bcf86aab0.jpg"},{"id":103765618,"identity":"2a862ff5-b0d4-4920-8d17-c08609418ed3","added_by":"auto","created_at":"2026-03-02 16:05:56","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2155960,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/232645c8-5e8b-401f-bc7d-572a13e2cecf.pdf"},{"id":91021458,"identity":"1ac63c59-71b7-40f8-aa6b-009409f69a85","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"pptx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":75916,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure1.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/bc6002f1d7f043a2e6f5db89.pptx"},{"id":91021714,"identity":"ec4d62cf-32e6-4b25-9b2b-89815f20075e","added_by":"auto","created_at":"2025-09-10 18:47:46","extension":"pptx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":139411,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure10.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/069bec7efbfdb461d867f01f.pptx"},{"id":91021716,"identity":"460f14f1-f926-44cb-8aa9-1133160375a5","added_by":"auto","created_at":"2025-09-10 18:47:46","extension":"pptx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":143990,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure11.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/b168543969ccb8082172fd3b.pptx"},{"id":91021718,"identity":"19e9b54b-b2cc-4f6e-9508-01ce4a203291","added_by":"auto","created_at":"2025-09-10 18:47:46","extension":"pptx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":141510,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure12.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/772543936b32186eb7da4282.pptx"},{"id":91021465,"identity":"3bf4799c-428e-49be-be95-836e37f891de","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"pptx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":214055,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure13.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/abfb84f57f0b06f904f727b6.pptx"},{"id":91022193,"identity":"33a98508-61d9-4663-8fda-755476e3493a","added_by":"auto","created_at":"2025-09-10 18:55:46","extension":"pptx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":68843,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure14.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/85599d043bc407bfcf2e2d8e.pptx"},{"id":91021461,"identity":"5f87fa3f-73ea-4a10-b476-d8633fcbf132","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":21839,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/a64192ffc425c507c8a134cd.docx"},{"id":91021472,"identity":"d4f44b67-0859-4324-94e4-19a42e575237","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"pptx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":74476,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure2.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/42d39c2ff7f6b5b1fd830f69.pptx"},{"id":91021471,"identity":"c30936cf-1201-496e-9b02-7e11758ebfbe","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"pptx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":374019,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure3.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/65e049ca92cacc83d88f36e6.pptx"},{"id":91022192,"identity":"23e2d7f0-ff9e-457d-b6bd-753436b9c059","added_by":"auto","created_at":"2025-09-10 18:55:46","extension":"pptx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":140113,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure4.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/35e2b8e7ae8c9b585c3f85ae.pptx"},{"id":91021469,"identity":"3760944e-685e-4dfc-81ac-ea4f39f40143","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"pptx","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":470902,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure5.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/e2948548d6ddc76cd4264e75.pptx"},{"id":91021475,"identity":"cfbef3fe-646b-4112-a0e7-5e596a7a4aa3","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"pptx","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":381300,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure6.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/2afdc174ac65b9f882d6e007.pptx"},{"id":91021486,"identity":"c6655532-14eb-4ba7-84b0-57110d3cd1bc","added_by":"auto","created_at":"2025-09-10 18:39:46","extension":"pptx","order_by":13,"title":"","display":"","copyAsset":false,"role":"supplement","size":138895,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure7.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/46d7a730614c8cf6c20668bd.pptx"},{"id":91021721,"identity":"50afe0bf-d1ec-4e33-84ee-2749895ac4f7","added_by":"auto","created_at":"2025-09-10 18:47:46","extension":"pptx","order_by":14,"title":"","display":"","copyAsset":false,"role":"supplement","size":404881,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure8.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/3d83a5a6fae4ee567922eade.pptx"},{"id":91021724,"identity":"9cea9a72-cf7c-4531-89c4-243cfbf09eec","added_by":"auto","created_at":"2025-09-10 18:47:46","extension":"pptx","order_by":15,"title":"","display":"","copyAsset":false,"role":"supplement","size":388824,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure9.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/05e5c8630e35d977cc345100.pptx"},{"id":91021719,"identity":"c5d87548-9456-4a77-bf74-63fb9c0696f7","added_by":"auto","created_at":"2025-09-10 18:47:46","extension":"docx","order_by":16,"title":"","display":"","copyAsset":false,"role":"supplement","size":21595,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7325643/v1/e93d1698ed349fc78f7c1d35.docx"}],"financialInterests":"","formattedTitle":"Heart Rate Variability Provides Prognostic value in Multiple System Atrophy","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMultiple system atrophy (MSA) is a relentlessly progressive neurodegenerative disorder characterized by a complex interplay of autonomic failure, parkinsonism, and cerebellar dysfunction.[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e] Despite advancements in understanding its clinical trajectory,[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] accurately predicting disease severity and survival remains a significant challenge, largely due to the heterogeneous and multifaceted expression of MSA. Current clinical assessment, primarily based on the Unified Multiple System Atrophy Rating Scale (UMSARS), offer essential but inherently limited evaluations of disease burden, as they may fail to capture the deeper, underlying pathophysiological mechanisms driving disease progression.[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] This underscores a pressing need for novel, complementary biomarkers that can enhance predictive modelling, refine patient stratification, and optimize clinical management in MSA.\u003c/p\u003e\u003cp\u003eAutonomic failure is a cornerstone of MSA pathophysiology and clinical management.[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] The progression of the UMSARS scores and the severity of orthostatic hypotension (OH) - a key feature of autonomic failure - have been shown to be key markers/factors of poor survival.[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] Heart rate variability (HRV), a non-invasive and dynamic measure of autonomic nervous system function, has been used as a biomarker of dysautonomia (e.g., [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]). HRV, the variation in the period between consecutive heartbeats over time, is a useful signal to understand the status of the autonomic nervous system (ANS) and provides information about the interaction between the sympathetic and parasympathetic systems, being one of the distinguishing features across many cardiovascular pathologies (e.g., [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]). HRV reflects complex physiological interactions governing autonomic regulation, cardiovascular integrity, and even neural network stability\u0026mdash;domains highly relevant to MSA, where autonomic dysfunction is a defining clinical feature.[\u003cspan additionalcitationids=\"CR11 CR12 CR13\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] The fact that HRV is reduced in MSA compared to control populations has been previously demonstrated.[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] Despite its theoretical promise, the clinical utility of HRV in MSA remains underexplored, and its precise relationship with disease severity and survival has yet to be firmly established.\u003c/p\u003e\u003cp\u003eIn this study, we systematically evaluated the predictive power of HRV in MSA, addressing critical gaps in the field. In addition to commonly used HRV analysis domains such as time-domain and frequency-domain, we have also used non-linear approaches to better understand HRV complexity.[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] These extra HRV domains (e.g., fractal analysis, short-term complexity, entropy, regularity, nonlinear dynamics) have been identified as informative for a better understanding of HRV. However, their use as clinical tools has to date been rather limited.[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] Specifically, we investigated:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eThe relationship between HRV metrics and disease severity, as measured by the clinical gold-standard UMSARS.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThe prognostic value of HRV for survival prediction, identifying the HRV features most strongly associated with mortality.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThe independent contribution of HRV beyond traditional clinical assessment (cardiovascular autonomic tests), determining whether it provides non-redundant insights into disease progression.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThe independent contribution of HRV beyond traditional clinical assessments, including autonomic cardiovascular battery tests, determining whether it provides unique, non-redundant insights into disease progression.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eData Collection and Preprocessing\u003c/h2\u003e\u003cp\u003eMSA patients were enrolled at the Toulouse French MSA Reference Centre between May 2011 and June 2020. Eligible patients met the current consensus criteria for a probable or possible MSA diagnosis, including both parkinsonian (P) and cerebellar (C) variants[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] and underwent a cardiovascular autonomic reflex assessment, including HRV (time-domain, frequency-domain, and nonlinear metrics) measured at rest in the supine position and autonomic CV testing as initially described by Ewing. This study was conducted in accordance with the Declaration of Helsinki and approved by the Toulouse University Hospital Ethics Committee as part of a broader prospective longitudinal study on natural MSA progression (CNIL 1338780, CCTIRS 10.065).[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] All patients were evaluated by a movement disorder specialist on the same day as the cardiovascular autonomic assessment, which was considered the baseline visit. Data collection included (on the same day) demographic information, medical history, neurological examination, diagnostic certainty and subtype classification based on consensus criteria,[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] and the UMSARS I-II-III and IV scores, COMPASS-31 and SCOPA-AUT.[\u003cspan additionalcitationids=\"CR22\" citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e] Patients were followed until death or until censoring on December 31st 2022.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eCardiovascular Autonomic Tests\u003c/h3\u003e\n\u003cp\u003ePatients underwent a standardized autonomic laboratory evaluation using continuous beat-to-beat digital blood pressure (BP) (Nexfin\u0026reg;, BMEYE, or Finapres\u0026reg; NOVA, FMS, The Netherlands) and electrocardiogram (ECG) recordings (LabChart, ADInstruments, Oxford, United Kingdom). The assessment included four tests performed in a controlled environment in the morning, following a 5-minute rest in a supine position. The tests were conducted in a fixed order under identical conditions and lasted approximately 45 minutes in total. The following tests were performed (as initially described by Ewing)[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]:\u003c/p\u003e\u003cp\u003e\u0026rarr; Deep Breathing Test (DB): Six deep breaths per minute in a supine position.\u003c/p\u003e\u003cp\u003e\u0026rarr; Valsalva Maneuver (VM): Expiratory pressure of 40 mmHg for 15 seconds while supine.\u003c/p\u003e\u003cp\u003e\u0026rarr; Head-Up Tilt Test (HUTT): Passive tilt to 80\u0026deg; for 10 minutes with BP recorded every minute using a sphygmomanometer (arm cuff).\u003c/p\u003e\u003cp\u003e\u0026rarr; Stand Test (ST): Five minutes of standing with BP recorded every minute using a sphygmomanometer (arm cuff) following 5 min in supine position.\u003c/p\u003e\u003cp\u003e\u0026rarr; Isometric Handgrip Test (HG): Three minutes of sustained handgrip exercise performed while seated.\u003c/p\u003e\u003cp\u003eThe total cardiovascular score[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] was derived from key parameters, including heart rate (HR) variations during DB (HR-DB), the 30/15 ratio during ST (HR-ST, immediate changes in HR after standing, max RR interval after 30 sec/min RR interval after 15 sec), the Valsalva ratio (HR-VM, max HR in phase II /min HR in phase IV), systolic BP response to ST (BPs-ST), diastolic BP response to ST (BPd-ST), diastolic BP increase during HG (BPd-HG), and systolic/diastolic BP response during HUTT (BPs/d-HUTT). Additional analyses included the maximum drop in systolic BP during VM phase II (BPs-VM-II) and the systolic BP overshoot in VM phase IV (BPs-VM-IV).\u003c/p\u003e\u003cp\u003eChanges in BP and HR were assessed against age-specific laboratory normative data, with responses classified as normal (0) or impaired (1). A test result was considered abnormal if it fell below the 5th percentile for age.[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] The EwS ranged from 0 (normal autonomic function) to 5 (severe autonomic dysfunction). Patients were classified as having cardiovascular autonomic neuropathy (CAN) if they had an EwS of 2 or higher, indicating at least two abnormal test results.[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/p\u003e\n\u003ch3\u003eHeart Rate Variability (HRV) Analysis\u003c/h3\u003e\n\u003cp\u003eHRV analysis was performed using the HRVanalysis software, a validated tool for assessing cardiac autonomic activity through non-invasive RR interval variability measurements.[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] The software incorporates time-domain, frequency-domain, geometrical, and nonlinear methods, adhering to the standards outlined by the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology.[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/p\u003e\u003cp\u003eECG signals were acquired during 5 minutes in the supine position during the cardiovascular test, and the RR intervals were extracted via the HRV analysis software\u0026rsquo;s built-in R-peak detection algorithm, which utilizes wavelet denoising, adaptive thresholding, and sliding-window summation for high accuracy (\u0026gt;\u0026thinsp;99.5% detection rate). Ectopic beats, artifacts, and missing intervals were corrected using cubic spline interpolation for sequences\u0026thinsp;\u0026le;\u0026thinsp;3 beats and linear interpolation for longer sequences. Segments with excessive noise (\u0026gt;\u0026thinsp;5% invalid beats) were excluded to ensure signal integrity. Wavelet transform was applied to localize transient autonomic changes, with Morlet wavelets decomposing the RR series into time-resolved LF, HF, and LF/HF ratio components. HRV indices were computed across four domains. A description for each variable and their clinical meaning can be found in \u003cb\u003eSupplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e.\u003c/p\u003e\n\u003ch3\u003eBaroreflex assessment\u003c/h3\u003e\n\u003cp\u003eThe baroreflex response was assessed during 5-minute recordings in supine position. The sequence method has been used, based on the scan of beat-to-beat SBP and R-R interval series. Different sequences of 3 or more consecutive heart beats were identified where either the SBP increased and the R-R intervals lengthened, or the SBP decreased and the R-R intervals shortened.[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e] We used at least 3 beats to consider 1 sequence with a minimum threshold of 1 mmHg between 2 BP records and a 5ms interval for the R-R interval. The minimal coefficient correlation used to validate a sequence was r\u0026thinsp;\u0026gt;\u0026thinsp;0.85. All slopes of the regression line obtained from all sequences were finally averaged to determine the BRS gain.[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/p\u003e\n\u003ch3\u003eStatistical Analyses\u003c/h3\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eCorrelation Analyses\u003c/h2\u003e\u003cp\u003eSpearman rank correlation coefficients were computed to examine associations between individual HRV features and clinical measures at the same time, including symptom duration (time from symptom onset to HRV assessment), diagnosis duration (time from MSA diagnosis to HRV assessment), and UMSARS scores (total score combining UMSARS Part I activities of daily living and Part II motor examination). A correlation matrix was visualized using a heatmap to identify patterns of association and potential multicollinearity among HRV features.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eLocal weighted regression\u003c/h3\u003e\n\u003cp\u003e A nonparametric locally weighted regression (LOESS) analysis was performed to visualize potential nonlinear relationships between HRV features and UMSARS scores. Scatterplots with smoothed LOESS curves were generated to assess trends in the data.\u003c/p\u003e\n\u003ch3\u003eRegularized Regression Models for the UMSARS\u003c/h3\u003e\n\u003cp\u003eGiven the high-dimensional nature of the HRV dataset and potential multicollinearity, regularized regression models were employed to predict total UMSARS (I\u0026thinsp;+\u0026thinsp;II) scores. LASSO regression (Least Absolute Shrinkage and Selection Operator) uses L1 regularization to shrink some coefficients to zero, performing feature selection by retaining only the most relevant predictors. Ridge regression uses L2 regularization to shrink all coefficients toward zero, mitigating collinearity while retaining all predictors. Both models were tuned using 10-fold cross-validation, optimizing the regularization parameter (λ) to minimize mean squared error (MSE). The coefficients from standardized features were interpreted, with positive values indicating predictors associated with higher total UMSARS I\u0026thinsp;+\u0026thinsp;II scores and negative values indicating protective effects.\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eSurvival Analysis for Mortality and Regularized Regression for Time-to-death\u003c/h2\u003e\u003cp\u003eKaplan-Meier survival analysis was conducted and survival probabilities were estimated/visualized as a function of follow-up time (time from HRV assessment to death or censoring). A Cox proportional hazards model was fitted to assess the association between individual HRV features and mortality. The survival object was defined using time-to-event (time from HRV assessment to death or censoring) and event status (0\u0026thinsp;=\u0026thinsp;censored, 1\u0026thinsp;=\u0026thinsp;death). To improve interpretability and multidimensionality, regularized Cox regression models (LASSO and Ridge) were implemented. These models identified the most predictive HRV features for mortality risk while controlling for overfitting. The optimal λ was selected via 10-fold cross-validation, and hazard ratios (HRs) were reported for retained features. To determine whether HRV features improved mortality prediction beyond clinical severity measures, we developed three models: HRV-only model consisting of LASSO regression trained on HRV features to predict time-to-death, UMSARS-only model consisting of a simple linear regression using UMSARS score as the sole predictor, and combined model (HRV\u0026thinsp;+\u0026thinsp;UMSARS) consisting of a LASSO regression including both HRV features and UMSARS scores as possible predictors. Model performance was evaluated using mean squared error (MSE) and R\u0026sup2; values. The added predictive value of HRV features was assessed by comparing the combined model to the UMSARS-only model.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eMediation Analyses\u003c/h2\u003e\u003cp\u003eTo investigate whether the relationship between HRV and mortality was mediated through UMSARS (a clinical measure of disease severity), a causal mediation analysis was performed. The independent variable was the HRV-derived risk scores (from LASSO/Ridge models), the mediator was the UMSARS total score, the outcome was the time-to-death. These mediation analyses employed a quasi-Bayesian approximation with 1,000 simulations to estimate the Average Causal Mediated Effect (ACME) or indirect effect of HRV on mortality through UMSARS, the Average Direct Effect (ADE) or direct effect of HRV on mortality, independent of UMSARS, the Total Effect or the sum of ACME and ADE, and the Proportion Mediated or the percentage of HRV\u0026rsquo;s effect on mortality explained by UMSARS.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eNeural Network Models\u003c/h2\u003e\u003cp\u003eTo explore potential nonlinear relationships, a feed-forward neural network was developed to predict UMSARS scores and systolic blood pressure (SBP) deltas (maximum observed drop), as assessed during the UMSARS-III test (disease severity and orthostatic hypotension proxies). For the model architecture, the input layer consisted of 50 nodes (one for each HRV feature), the hidden layers consisted of ReLU-activated layers with dropout regularization (rate\u0026thinsp;=\u0026thinsp;0.2\u0026ndash;0.5) and the output layer consisted of a single node for prediction (linear activation). For the training process, the Adam (for UMSARS prediction) or RMSprop (for blood pressure prediction) optimizers were chosen. As the loss function, the Mean Absolute Error (MAE) was evaluated. Early stopping was enforced to prevent overfitting by halting training when validation performance plateaued. To interpret model predictions, SHapley Additive exPlanations (SHAP) were used to quantify the contribution of each HRV feature. Performance was evaluated via scatter plots, kernel density plots, correlation analysis, and training/validation loss curves.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eBest Subset Selection\u003c/h2\u003e\u003cp\u003eTo identify the most relevant HRV predictors when it comes to both disease severity and mortality risk, best subset selection was performed. Models were built incrementally from one predictor to all predictors, with selection guided by the adjusted R\u0026sup2; (variance explained, adjusted for model complexity), the Mallow\u0026rsquo;s Cp (valuates bias-variance trade-off), the Bayesian Information Criterion (BIC, penalizes overfitting), and the Residual Sum of Squares (RSS, Measures total prediction error).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eAutonomic Cardiovascular Testing and Survival Analysis\u003c/h2\u003e\u003cp\u003eUsing strategies identical to those detailed above employed for HRV, the association between cardiovascular function (global score and individual sub-scores for each test) and the total UMSARS (I\u0026thinsp;+\u0026thinsp;II), the SBP delta (UMSARS III), or survival have been accessed. In addition, Cox proportional hazards models were used to determine their association with survival.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e214 MSA patients were included in this study, consisting of 56% male, 71% MSA-P and 86% probable MSA. Mean (SD) disease duration at autonomic assessment was 4.5 (2.2) years, and UMSARS IV score 2.5 (1.1) (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e for all clinical details). As the cardiovascular autonomic testing was performed only once per patient, longitudinal changes in HRV over the disease course could only be inferred at the cohort level, with different patients assessed at various stages of disease progression (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Generally, HRV features showed modest negative correlations with disease severity, mirroring clinical deterioration. Notably, some HRV metrics exhibited correlations with disease progression comparable in magnitude to those observed with the total UMSARS I\u0026thinsp;+\u0026thinsp;II score, underscoring their potential clinical relevance. While classical HRV parameters such as those of the temporal and time domains decrease with disease progression, HRV features from the Entropy (Approximate Entropy, Sample Entropy, Shanon Entropy, Conditional Entropy (CE), Corrected CE, Normalized CCE, ρ, Lempel-Ziv Complexity) and Symbolic Dynamics domains (OV, OV%, 1V, 1V%, 2V, 2V%, 2UV, 2UV%, MP, MP%) tended to increase with disease progression, reflecting domain-specific patterns of change over time (\u003cb\u003eSupplementary Figs.\u0026nbsp;1, 2 and 3\u003c/b\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSummary Clinical and Demographic Features (n\u0026thinsp;=\u0026thinsp;214)\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFeature\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMean (\u0026plusmn;\u0026thinsp;SD) | Median [Q1-Q3] | Proportion %\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"13\" rowspan=\"14\"\u003e\u003cp\u003eClinical Demographic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e56% Male\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMSA Prob vs Pos\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e86% Prob | 14% Pos\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMSA-C vs MSA-P\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e29% MSA-C | 71% MSA-P\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e64.2 (\u0026plusmn;\u0026thinsp;7.9) | 64.3 [59.2\u0026ndash;70.1]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e% Died\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e188 (88%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFollow-up Duration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.6 (\u0026plusmn;\u0026thinsp;2.0) | 3.4 [2.1-5.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDisease Duration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.5 (\u0026plusmn;\u0026thinsp;2.2) | 4.0 [3.0\u0026ndash;6.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal UMSARS 1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e22.7 (\u0026plusmn;\u0026thinsp;7.2) | 22.0 [18.0\u0026ndash;27.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal UMSARS 2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26.1 (\u0026plusmn;\u0026thinsp;7.7) | 25.0 [20.0\u0026ndash;31.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMax D SBP Drop\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35 (\u0026plusmn;\u0026thinsp;20) | 32 [\u003cspan additionalcitationids=\"CR24 CR25 CR26 CR27 CR28 CR29 CR30 CR31 CR32 CR33 CR34 CR35 CR36 CR37 CR38 CR39 CR40 CR41 CR42 CR43 CR44 CR45 CR46\" citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMax D DBP Drop\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e18 (\u0026plusmn;\u0026thinsp;13) | 15 [\u003cspan additionalcitationids=\"CR11 CR12 CR13 CR14 CR15 CR16 CR17 CR18 CR19 CR20 CR21 CR22 CR23\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal UMSARS 4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.5 (\u0026plusmn;\u0026thinsp;1.0) | 2.0 [2.0\u0026ndash;3.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal COMPASS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e34.3 (\u0026plusmn;\u0026thinsp;16.2) | 30.4 [19.7\u0026ndash;47.3]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal SCOPA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19.8 (\u0026plusmn;\u0026thinsp;11.2) | 20.0 [13.0\u0026ndash;27.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBRS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ems/mmHg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.0 (\u0026plusmn;\u0026thinsp;3.3) | 2.8 [1.9\u0026ndash;5.3]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eTemporal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMean RR (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e798 (\u0026plusmn;\u0026thinsp;116) | 785 [722\u0026ndash;871]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003epNN50 (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.0 (\u0026plusmn;\u0026thinsp;9.5) | 0.0 [0.0-0.7]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSDNN (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19.5 (\u0026plusmn;\u0026thinsp;11.3) | 16.5 [11.9\u0026ndash;24.6]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003erMSSD (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15.4 (\u0026plusmn;\u0026thinsp;15.6) | 11.5 [7.5\u0026ndash;16.9]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e\u003cp\u003eFrequential\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePtot (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e420 (\u0026plusmn;\u0026thinsp;607) | 215 [109\u0026ndash;455]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eVLF (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e186 (\u0026plusmn;\u0026thinsp;269) | 107 [46\u0026ndash;214]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLF (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e98 (\u0026plusmn;\u0026thinsp;174) | 40 [17\u0026ndash;102]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHF (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e83 (\u0026plusmn;\u0026thinsp;229) | 25 [11\u0026ndash;58]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLF/HF\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.6 (\u0026plusmn;\u0026thinsp;2.9) | 1.6 [0.9\u0026ndash;3.2]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eEmpirical Decomposition\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003epLF1 (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e59.5 (\u0026plusmn;\u0026thinsp;89.1) | 22.9 [9.9\u0026ndash;68.5]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003epLF2 (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e71.3 (\u0026plusmn;\u0026thinsp;107.6) | 38.7 [14.9\u0026ndash;83.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003epHF1 (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e69.0 (\u0026plusmn;\u0026thinsp;141.3) | 22.2 [7.5\u0026ndash;59.3]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003epHF2 (ms\u0026sup2;)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e149.7 (\u0026plusmn;\u0026thinsp;487.8) | 36.6 [13.9\u0026ndash;92.3]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIMAI1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.57 (\u0026plusmn;\u0026thinsp;0.63) | 0.39 [0.18\u0026ndash;0.78]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIMAI2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.90 (\u0026plusmn;\u0026thinsp;1.22) | 0.50 [0.24\u0026ndash;1.03]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eGeometrical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTriangular index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.2 (\u0026plusmn;\u0026thinsp;2.3) | 6.5 [5.6\u0026ndash;8.1]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTINN (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e112 (\u0026plusmn;\u0026thinsp;35) | 102 [86\u0026ndash;125]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eX (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e794 (\u0026plusmn;\u0026thinsp;127) | 789 [719\u0026ndash;867]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eY (beats)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e58 (\u0026plusmn;\u0026thinsp;19) | 58 [44\u0026ndash;70]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eM (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e855 (\u0026plusmn;\u0026thinsp;124) | 836 [773\u0026ndash;922]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eN (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e743 (\u0026plusmn;\u0026thinsp;111) | 734 [672\u0026ndash;805]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"7\" rowspan=\"8\"\u003e\u003cp\u003eEntropy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eApproximate Entropy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.10 (\u0026plusmn;\u0026thinsp;0.13) | 1.13 [1.04\u0026ndash;1.19]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSample Entropy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.34 (\u0026plusmn;\u0026thinsp;0.34) | 1.34 [1.10\u0026ndash;1.56]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eShanon Entropy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.31 (\u0026plusmn;\u0026thinsp;0.58) | 3.43 [2.97\u0026ndash;3.69]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eConditional Entropy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.88 (\u0026plusmn;\u0026thinsp;0.20) | 0.91 [0.75\u0026ndash;1.03]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCorrected CE\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.90 (\u0026plusmn;\u0026thinsp;0.22) | 0.92 [0.75\u0026ndash;1.06]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNormalized CCE\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.64 (\u0026plusmn;\u0026thinsp;0.12) | 0.64 [0.56\u0026ndash;0.72]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eΡ\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.36 (\u0026plusmn;\u0026thinsp;0.12) | 0.36 [0.28\u0026ndash;0.44]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLempel-Ziv Complexity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.96 (\u0026plusmn;\u0026thinsp;0.11) | 0.98 [0.90\u0026ndash;1.03]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003ePoincar\u0026eacute; plot\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCentroid (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e798 (\u0026plusmn;\u0026thinsp;116) | 785 [722\u0026ndash;871]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSD1 (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10.5 (\u0026plusmn;\u0026thinsp;9.9) | 8.1 [5.3\u0026ndash;11.1]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSD2 (ms)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e24.1 (\u0026plusmn;\u0026thinsp;12.6) | 21.3 [15.1\u0026ndash;30.9]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSD1/SD2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.44 (\u0026plusmn;\u0026thinsp;0.29) | 0.37 [0.28\u0026ndash;0.52]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"9\" rowspan=\"10\"\u003e\u003cp\u003eSymbolic Dynamics\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOV\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e123 (\u0026plusmn;\u0026thinsp;77) | 109 [64\u0026ndash;175]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOV%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e32 (\u0026plusmn;\u0026thinsp;19) | 29 [\u003cspan additionalcitationids=\"CR18 CR19 CR20 CR21 CR22 CR23 CR24 CR25 CR26 CR27 CR28 CR29 CR30 CR31 CR32 CR33 CR34 CR35 CR36 CR37 CR38 CR39 CR40 CR41 CR42 CR43 CR44\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1V\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e156 (\u0026plusmn;\u0026thinsp;39) | 159 [132\u0026ndash;180]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1V%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e41 (\u0026plusmn;\u0026thinsp;9) | 42 [\u003cspan additionalcitationids=\"CR38 CR39 CR40 CR41 CR42 CR43 CR44 CR45 CR46\" citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2V\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21 (\u0026plusmn;\u0026thinsp;22) | 15 [\u003cspan additionalcitationids=\"CR7 CR8 CR9 CR10 CR11 CR12 CR13 CR14 CR15 CR16 CR17 CR18 CR19 CR20 CR21 CR22 CR23 CR24 CR25 CR26\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2V%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6 (\u0026plusmn;\u0026thinsp;6) | 4 [\u003cspan additionalcitationids=\"CR2 CR3 CR4 CR5 CR6 CR7\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2UV\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e79 (\u0026plusmn;\u0026thinsp;44) | 69 [46\u0026ndash;106]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2UV%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21 (\u0026plusmn;\u0026thinsp;12) | 19 [\u003cspan additionalcitationids=\"CR13 CR14 CR15 CR16 CR17 CR18 CR19 CR20 CR21 CR22 CR23 CR24 CR25 CR26 CR27 CR28\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e160 (\u0026plusmn;\u0026thinsp;19) | 163 [149\u0026ndash;176]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMP%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e43 (\u0026plusmn;\u0026thinsp;8) | 43 [\u003cspan additionalcitationids=\"CR40 CR41 CR42 CR43 CR44 CR45 CR46 CR47\" citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eFractal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eΑ1 (DFA)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.98 (\u0026plusmn;\u0026thinsp;0.35) | 0.98 [0.76\u0026ndash;1.20]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eΑ2 (DFA)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.05 (\u0026plusmn;\u0026thinsp;0.20) | 1.06 [0.94\u0026ndash;1.16]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eH (DFA)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.99 (\u0026plusmn;\u0026thinsp;0.16) | 0.99 [0.89\u0026ndash;1.09]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eH (Higuchi)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.81 (\u0026plusmn;\u0026thinsp;0.18) | 1.82 [1.71\u0026ndash;1.93]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eH (Katz)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.67 (\u0026plusmn;\u0026thinsp;0.45) | 1.59 [1.43\u0026ndash;1.74]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHurst\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.21 (\u0026plusmn;\u0026thinsp;0.14) | 0.21 [0.11\u0026ndash;0.30]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"16\" rowspan=\"17\"\u003e\u003cp\u003eEwing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Valsalva Ratio\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.2 (\u0026plusmn;\u0026thinsp;0.2) | 1.2 [1.1\u0026ndash;1.3]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing respiratory amplitude (bpm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.3 (\u0026plusmn;\u0026thinsp;3.3) | 4.0 [3.0\u0026ndash;7.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing 30/15 ratio\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.1 (\u0026plusmn;\u0026thinsp;0.1) | 1.0 [1.0-1.1]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Iso DBP Handgrip\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9.6 (\u0026plusmn;\u0026thinsp;7.0) | 9.0 [4.0\u0026ndash;14.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Iso SBP Handgrip\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15.4 (\u0026plusmn;\u0026thinsp;12.4) | 13.0 [5.5\u0026ndash;22.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing SBP Tilting Delta\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26.4 (\u0026plusmn;\u0026thinsp;21.0) | 23.0 [13.0\u0026ndash;38.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing DBP Tilting Delta\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11.1 (\u0026plusmn;\u0026thinsp;13.0) | 9.0 [3.0\u0026ndash;19.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Orthost stand SBP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e23.8 (\u0026plusmn;\u0026thinsp;22.1) | 22.0 [8.0\u0026ndash;35.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Orthost stand DBP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10.8 (\u0026plusmn;\u0026thinsp;14.7) | 9.0 [0.0-18.3]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Valsalva Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7 (\u0026plusmn;\u0026thinsp;0.4) | 1.0 [0.5-1.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing respiratory score (bpm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.6 (\u0026plusmn;\u0026thinsp;0.5) | 1.0 [0.0\u0026ndash;1.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing 30/15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.5 (\u0026plusmn;\u0026thinsp;0.4) | 0.5 [0.0\u0026ndash;1.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Iso Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7 (\u0026plusmn;\u0026thinsp;0.4) | 1.0 [0.0\u0026ndash;1.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Orthost Tilt\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7 (\u0026plusmn;\u0026thinsp;0.4) | 1.0 [0.5-1.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEwing Total Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.0 (\u0026plusmn;\u0026thinsp;1.3) | 3.0 [2.0\u0026ndash;4.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eValsalva SBP Delta phase_IIb (mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35.9 (\u0026plusmn;\u0026thinsp;22.7) | 35.0 [19.3\u0026ndash;50.7]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eValsalva SBP Delta phase_Ivb (mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.7 (\u0026plusmn;\u0026thinsp;8.7) | 5.0 [0.0\u0026ndash;12.0]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eHRV and the UMSARS\u003c/h2\u003e\u003cdiv id=\"Sec18\" class=\"Section3\"\u003e\u003ch2\u003eCorrelation Analyses\u003c/h2\u003e\u003cp\u003eAs expected, initial correlation analysis revealed widespread multicollinearity among HRV features, with many exhibiting high intercorrelations. However, correlations between individual HRV features and total UMSARS I\u0026thinsp;+\u0026thinsp;II scores were uniformly weak, indicating no strong linear relationships. While some HRV features had modest associations with total UMSARS I\u0026thinsp;+\u0026thinsp;II, none demonstrated an extremely strong predictive signal (maximal correlation of ~\u0026thinsp;0.2 for the strongest correlation HRV features). These findings suggested that HRV features alone may not directly reflect disease severity as measured by total UMSARS I\u0026thinsp;+\u0026thinsp;II but could still contain clinically relevant information when combined. To explore potential nonlinear trends, locally weighted regression (LOESS) was applied to visualize the association between HRV features and UMSARS scores. Consistent with the correlation findings, no clear patterns emerged across individual HRV features, reinforcing the idea that no single HRV feature displayed a strong predictive relationship with UMSARS (\u003cb\u003eSupplementary Figs.\u0026nbsp;1, 2 and 3\u003c/b\u003e).\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003eRegularized Regression Models\u003c/h2\u003e\u003cp\u003eGiven the challenges posed by multidimensionality and weak linear correlations, Ridge (L2) and LASSO (L1) regression were employed to refine the analysis. These methods address overfitting by penalizing coefficient magnitudes, with LASSO selecting a subset of the most relevant features and Ridge shrinking all coefficients to stabilize the model. LASSO identified Lempel-Ziv Complexity and various entropy-based features as the most informative predictors of the total UMSARS I\u0026thinsp;+\u0026thinsp;II scores. Ridge regression retained all predictors but confirmed that complexity-based and entropy-based metrics were the most relevant. Despite these refinements, the retained HRV features still exhibited only modest relationships with total UMSARS I\u0026thinsp;+\u0026thinsp;II scores, aligning with earlier analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). While biological age and gender themselves were also informative variables at predicting the UMSARS, adjusting for these did not noticeably change which HRV features were the most informative ones (\u003cb\u003eSupplementary Fig.\u0026nbsp;4\u003c/b\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003eBest Subset Selection\u003c/h2\u003e\u003cp\u003eTo systematically evaluate different combinations of HRV features, best subset selection was employed. This method iteratively tested models of increasing complexity while balancing predictive accuracy. The most informative HRV features were similar to those identified by regularized models, including Lempel-Ziv Complexity and several entropy-based features. Models with ~\u0026thinsp;20 HRV features achieved an adjusted R\u0026sup2; of 0.4, translating to a correlation of 0.6 between predicted and observed total UMSARS I\u0026thinsp;+\u0026thinsp;II scores using HRV features alone. (once again, primarily of the Entropy and Symbolic Dynamics domains, followed by other domains to a smaller extent - e.g., Fractal domains with H Higuchi, H DFA or Hurst). This suggests that while HRV features contain useful information when it comes to disease severity, their predictive strength remains modest at best (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Adjusting for biological age and gender (e.g., if also including these in the models) did not meaningfully increase the predictive performance (correlation remained at ~\u0026thinsp;0.6\u0026ndash;0.7 for a peak R\u003csup\u003e2\u003c/sup\u003e with ~\u0026thinsp;20 variables) and the same (primarily complexity-based and entropy-based) HRV features remained as the most relevant ones (\u003cb\u003eSupplementary Fig.\u0026nbsp;4\u003c/b\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003eNeural Network Model\u003c/h2\u003e\u003cp\u003eTo capture potential nonlinear relationships, a feed-forward neural network was implemented. The model architecture was optimized for hyperparameters, dropout regularization, and activation functions. SHapley Additive exPlanations (SHAP) analysis revealed that Entropy, Symbolic Dynamics, and Fractal domains were the most contributing towards these predictions (e.g., 1V, Sample Entropy, or Approximate Entropy negatively contributing to the predicted UMSARS scores; Lempel-Ziv Complexity, 2V%, H DFA, positively contributing to the predicted UMSARS scores). Despite capturing potential nonlinearities, the model reaffirmed the limited predictive power of HRV features alone in assessing disease severity as classical defined using the UMSARS (\u003cb\u003eSupplementary Fig.\u0026nbsp;5\u003c/b\u003e). Across models, higher Lempel-Ziv Complexity, reflecting increased sequence complexity, and greater detrended fluctuation analysis (DFA) values, indicating enhanced self-similarity (fractality) of the RR interval signal, were associated with higher UMSARS scores. Conversely, entropy-based measures were inversely associated with UMSARS.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003eHRV and Blood Pressure Regulation\u003c/h2\u003e\u003cdiv id=\"Sec23\" class=\"Section3\"\u003e\u003ch2\u003eRegularized Regression Models\u003c/h2\u003e\u003cp\u003eTo investigate the relationship between HRV features and SBP delta (as assessed during UMSARS III test) Ridge and LASSO regression were applied. LASSO identified HRV features of the Symbolic Dynamics, Fractal, and Empirical Decomposition domains as the most informative features. Entropy HRV features were still relevant, but relatively less so as compared to the UMSARS. Ridge and LASSO achieved an R\u0026sup2; of ~\u0026thinsp;0.3 (correlation\u0026thinsp;~\u0026thinsp;0.5), suggesting moderate predictive power (\u003cb\u003eSupplementary Figs.\u0026nbsp;6 and 7\u003c/b\u003e).\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e\u003ch2\u003eBest Subset Selection\u003c/h2\u003e\u003cp\u003eThe best models correlated modestly (r\u0026thinsp;=\u0026thinsp;0.6) with observed SBP deltas using\u0026thinsp;~\u0026thinsp;25 different HRV features, with once again the most import ones belonging primarily to the Symbolic Dynamics (1V, 2UV%), Fractal (e.g., H Katz), and Entropy (e.g., Shannon Entropy) domains (\u003cb\u003eSupplementary Figs.\u0026nbsp;6 and 7\u003c/b\u003e).\u003c/p\u003e\u003cdiv id=\"Sec25\" class=\"Section3\"\u003e\u003ch2\u003eNeural Network Model\u003c/h2\u003e\u003cp\u003eA neural network was implemented to assess nonlinear associations between HRV and SBP deltas, with modest predictive accuracy (~\u0026thinsp;0.4 correlation). SHAP analysis again features of the Symbolic Dynamics (1V, 2V%), Fractal (e.g., H Hurst, A1 DFA), and Entropy (e.g., Approximate Entropy) domains to be the key predictors. Even under more flexible modelling assumptions, HRV alone provided limited predictive value for blood pressure regulation (\u003cb\u003eSupplementary Fig.\u0026nbsp;8\u003c/b\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec26\" class=\"Section3\"\u003e\u003ch2\u003eHRV and Mortality/Time-to-death\u003c/h2\u003e\u003cdiv id=\"Sec27\" class=\"Section4\"\u003e\u003ch2\u003eKaplan-Meier Survival Analysis\u003c/h2\u003e\u003cp\u003eA total of 87% of the cohort patients have died, with a median survival time of 40 months (95% C.I. 39\u0026ndash;51). Some HRV features were associated with time-to-death, with scores in most features across the Fractal, Symbolic Dynamics, and Entropy domains being associated with higher chances of survival and/or longer time-to-death (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec28\" class=\"Section2\"\u003e\u003ch2\u003eCox Proportional Hazards Model\u003c/h2\u003e\u003cp\u003eOriginal (imbalanced) dataset in LASSO and Ridge models identified deceased patients with 80\u0026ndash;90% accuracy [ (true positives\u0026thinsp;+\u0026thinsp;true negatives) / total observations, using\u0026thinsp;\u0026gt;\u0026thinsp;0.5 binarized risk scores ], but this was biased due to class imbalance. After randomly up-sampling the minority class, the model accuracy dropped to ~\u0026thinsp;70%, providing a more realistic estimate of HRV\u0026rsquo;s predictive power (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec29\" class=\"Section2\"\u003e\u003ch2\u003eTime-to-Death Prediction\u003c/h2\u003e\u003cp\u003eHRV features alone predicted time-to-death almost as well as total UMSARS I-II scores alone. The LASSO Model with HRV features only retained the 1V% and H Higuchi features, with a correlation coefficient of ~\u0026thinsp;0.35, similar to that of UMSARS-only models. When combining all HRV features\u0026thinsp;+\u0026thinsp;total UMSARS I\u0026thinsp;+\u0026thinsp;II scores, improved predictive accuracy was obtained (correlation of 0.5 for the HRV\u0026thinsp;+\u0026thinsp;UMSARS vs. 0.4 for the UMSARS alone), suggesting that HRV and UMSARS capture partially complementary survival-related information. With slightly more complex models (see best subset regression, 10\u0026ndash;20 variables, peaking at ~\u0026thinsp;15 HRV variables), a higher correlation can be obtained (correlation of ~\u0026thinsp;0.6). Of the time domain features, the pNN50(%) was the most important one, a marker of the parasympathetic drive with higher values suggesting higher vagal tone, thus inversely correlated with mortality. When investigating the information gain obtained by also including the maximum delta in SBP (UMSARS III) and symptom duration, these two features were not among the most informative ones at the overall performance at predicting the time-to-death was not meaningfully increased (~\u0026thinsp;0.6 correlation between predicted and observed values, Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, \u003cb\u003eSupplementary Figs.\u0026nbsp;9 and 10\u003c/b\u003e). As observed for the UMSARS, when it comes to time-to-death, biological age and gender themselves were herein observed to be very good features at predicting survival but adjusting for these did not meaningfully increase the predictive performance (remaining at ~\u0026thinsp;0.6 for a peak R\u003csup\u003e2\u003c/sup\u003e with ~\u0026thinsp;15 variables). This observation reflects the fact that the most important HRV features also change with ageing, accounting for a significant degree of redundancy. Regardless of the age/gender adjustment, the most important HRV features remained the same (primarily those of the Fractal, Symbolic Dynamics, and Entropy domains, \u003cb\u003eSupplementary Fig.\u0026nbsp;11\u003c/b\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eMediation Analysis\u003c/h3\u003e\n\u003cp\u003eIn mediation analysis, HRV had a direct contribution towards time-to-death predictions (ADE\u0026thinsp;=\u0026thinsp;0.0744, p\u0026thinsp;\u0026lt;\u0026thinsp;2\u0026times;10⁻\u0026sup1;⁶), with the total UMSARS I\u0026thinsp;+\u0026thinsp;II scores presenting only a minimal residual mediating effect on HRV\u0026rsquo;s influence on survival (ACME\u0026thinsp;=\u0026thinsp;0.0007, p\u0026thinsp;=\u0026thinsp;0.17). As such, HRV features capture significantly complementary physiological dimensions, beyond clinical severity otherwise classically measured by the UMSARS (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cdiv id=\"Sec31\" class=\"Section2\"\u003e\u003ch2\u003eHRV and Autonomic Cardiovascular Testing\u003c/h2\u003e\u003cp\u003eWorse autonomic cardiovascular test scores (historically known as Ewing scores) were significantly associated with shorter survivals (HR\u0026thinsp;=\u0026thinsp;1.32, p\u0026thinsp;=\u0026thinsp;0.003 for the global \u0026ldquo;Ewing\u0026rdquo; score), with a relatively high redundancy between its different sub-scores being observed. Of the different sub-scores, those pertaining to the Valsalva maneuver (both exacts changes and Valsalva ratio) and to respiratory amplitude were the most informative. Moreover, features pertaining to DBP (diastolic blood pressure) are consistently more informative than those pertaining to the SBP (systolic blood pressure). Not surprisingly given that the orthostatic tests are also part of the overall cardiovascular testing, HRV autonomic testing scores had a poor performance at mirroring the total UMSARS I\u0026thinsp;+\u0026thinsp;II but a high accuracy at mirroring the SBP delta (correlation\u0026thinsp;~\u0026thinsp;0.7) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e \u003cb\u003eand Supplementary Figs.\u0026nbsp;12 and 13\u003c/b\u003e). The classical cardiovascular assessment tools (as referred to as \u0026ldquo;Ewing scores\u0026rdquo;) provide comparable/identical correlatiom value to that of HRV features and their combination is unlikely to provide much added value (Root Mean Squared Error (RMSE) [RMSE]\u0026thinsp;~\u0026thinsp;12\u0026ndash;14, MAE\u0026thinsp;~\u0026thinsp;9\u0026ndash;10 at regression with the UMSARS for both classical CV alone, HRV alone, or both combined; RMSE\u0026thinsp;~\u0026thinsp;23\u0026ndash;24, MAE\u0026thinsp;~\u0026thinsp;18\u0026ndash;20 at regression with the time-to-death in months for both classical CV alone, HRV alone, or both combined, \u003cb\u003eSupplementary Fig.\u0026nbsp;14\u003c/b\u003e, with RMSE being the Root Mean Squared Error that calculates the square root of the average of the squared differences between predicted and actual values, which penalizes larger errors more heavily, and MAE being the Mean Absolute Error that computes the average of the absolute differences, treating all errors equally regardless of size). Of note, the \u0026ldquo;classical cardiovascular assessment\u0026rdquo; herein referred to comprises the extended battery of cardiovascular tests and not merely the orthostatic hypotension assessment performed with the UMSARS III (whose performance was in turn considerably inferior to HRV).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this large cohort of 214 MSA patients, we show that HRV features\u0026mdash;particularly those from the Entropy, Symbolic Dynamics, and Fractal domains\u0026mdash;provide complementary and partially independent information from traditional clinical assessments. While HRV alone showed only modest associations with disease severity as measured by UMSARS I\u0026thinsp;+\u0026thinsp;II, it demonstrated comparable prognostic power to classical cardiovascular autonomic tests and significantly improved survival prediction when combined with clinical scales. Importantly, HRV features predicted time-to-death nearly as well as UMSARS scores alone, and their combination yielded improved accuracy, supporting the relevance of HRV as a prognostic biomarker. Mediation analysis confirmed that HRV contributes directly to survival prediction, independently of UMSARS. These findings support the use of HRV as a non-invasive, physiologically meaningful marker that could complement clinical scales in MSA research and care.\u003c/p\u003e\u003cp\u003eThe dynamic responses of HR and BP during autonomic cardiovascular testing, illustrated by the total score and the sub-scores for the different autonomic tests were altered as already described in MSA patients.[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] In addition, we observed that the severity of both parasympathetic (heart rate changes during deep breathing and the Valsalva ratio) and sympathetic (blood pressure responses to the Valsalva maneuver and orthostatic tests) autonomic failure, as assessed through autonomic testing, had prognostic value for survival. Notably, the prognostic value of comprehensive autonomic testing was superior to that of orthostatic hypotension severity alone, as measured by clinical evaluation (UMSARS III).\u003c/p\u003e\u003cp\u003eThe objective of this study was to evaluate the prognostic value of short-term HRV analysis, based on 5-minute recordings in the supine position, for survival prediction, and to identify the HRV features most strongly associated with mortality. Our findings reveal that while HRV alone has only moderate predictive power for disease severity and survival, it captures physiological signals highly complementary to UMSARS. This may suggest that HRV is not merely an auxiliary metric but a valuable, independent biomarker that can refine risk stratification, prognosis, and potentially even therapeutic decision-making.\u003c/p\u003e\u003cdiv id=\"Sec33\" class=\"Section2\"\u003e\u003ch2\u003eHRV as a Predictor of Disease Severity\u003c/h2\u003e\u003cp\u003eThe weak correlation between HRV features and UMSARS suggests that autonomic dysfunction in MSA does not always parallel clinical severity as measured by conventional motor and functional scales (i.e., UMSARS). This aligns with previous studies showing that while autonomic impairment is a hallmark of MSA, its severity does not necessarily track with motor decline.[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e] In turn, alternative/complementary data such as entropy- and complexity-based HRV measures emerge as more relevant predictors, indicating that HRV dynamics may better capture the subtle physiological alterations associated with disease progression. Higher Lempel-Ziv Complexity, reflecting greater sequence complexity, and increased detrended fluctuation analysis (DFA), indicating stronger self-similarity of the RR interval signal, were associated with higher UMSARS scores. In contrast, lower entropy-based measures, reflecting reduced signal unpredictability, were associated with higher UMSARS scores. Even with advanced regression models and neural networks, HRV alone achieved only moderate predictive accuracy (adjusted R\u0026sup2; ~ 0.4, corresponding to a correlation of ~\u0026thinsp;0.6). This reinforces the idea that HRV should be viewed as a complementary physiological indicator that enriches clinical assessments.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec34\" class=\"Section2\"\u003e\u003ch2\u003eExpanding the Dimensions of Prognostication and Stratification\u003c/h2\u003e\u003cp\u003eThis study has underscored the independent prognostic value of some HRV parameters mainly assessed by non-linear approaches for survival. The combined HRV\u0026thinsp;+\u0026thinsp;UMSARS model outperformed UMSARS alone, demonstrating that these HRV parameters capture critical physiological processes beyond traditional clinical evaluations. Nonlinear methods of signal analysis can be more useful when characterizing complex dynamics. Cox proportional hazards modelling identified specific HRV features\u0026mdash;such as Higuchi\u0026rsquo;s fractal dimension (H Higuchi) and Lempel-Ziv complexity\u0026mdash;as significant predictors of survival. Mediation analysis confirmed that HRV\u0026rsquo;s impact on survival was largely independent of clinical severity, further underscoring its role as a unique autonomic biomarker with real prognostic relevance.\u003c/p\u003e\u003cp\u003eIn line with the strong positive correlation between the Higuchi\u0026rsquo;s index and mortality (plus inverse correlation with time-to-death) herein observed in the case of MSA, an increase in the Higuchi\u0026rsquo;s index has been previously reported in stroke patients,[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e] and in diabetic patients (compared to non-diabetic),[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e] an observation interpreted as impairment of the autonomic nervous system. An increase in Higuchi\u0026rsquo;s index has been suggested in atrial fibrillation.[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e] In contrast, other indicators of signal complexity such as the Hurst index ( autocorrelation/signal persistence over time) the Lempel-Ziv (diversity and number of signal sub-patterns), or the 1V% (ordinal variability and phase patterns), were herein observed to present a positive correlation with survival. A decrease in Lempel-Ziv complexity has been previously reported during Terbutaline (selective Beta2-adrenoceptor agonist) infusion, which decreases the parasympathetic drive of heart rate.[\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e] The decrease in HRV signal complexity and the increase in the fractal dimension may suggest a reduced central control of HRV. To our knowledge, no study has specifically analyzed Higuchi\u0026rsquo;s fractal dimension during atrial fibrillation. However, by analogy with other non-linear indices - such as Approximate Entropy - which have been shown to vary similarly to Higuchi\u0026rsquo;s fractal dimension across different populations, the elevated values observed in our study population appear to reflect a pattern consistent with autonomic dysregulation, as typically seen in atrial fibrillation. Supporting this interpretation, Xin et al. (2017) reported higher entropy values during paroxysmal atrial fibrillation episodes compared to periods of sinus rhythm.[\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e] Similarly, Yamada et al. (2000) found Approximate Entropy values around 1.85 in patients with atrial fibrillation, whereas typical Approximate Entropy values in healthy individuals in sinus rhythm are approximately 1.05.[\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e] These findings also suggest that HRV could serve as an objective tool for risk stratification, potentially guiding early interventions and improving patient management. Autonomic testing is known to provide an independent tool for both the diagnosis and prognosis of MSA. Likewise, cardiovascular autonomic failure and orthostatic hypotension are both good prognostic markers of poor survival in MSA.[\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e] In our study, the addition of orthostatic hypotension measurements to the HRV added only a marginal increase in information, owning to the high correlation between the two domains as markers of autonomic failure. In turn, HRV alone was superior to orthostatic hypotension assessment alone. However, a single clinical assessment does not capture the substantial BP variability often observed in these patients, which has also been identified as a risk factor for mortality.[\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e] Further studies are warranted to evaluate the prognostic value of BP fluctuations under sympathetic control, using methods such as ambulatory blood pressure monitoring in combination with HRV analysis.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eToward Real-Time, Automated HRV Monitoring\u003c/h3\u003e\n\u003cp\u003eOne of the promising implications of this study is the potential for real-time, ambulatory HRV monitoring using wearable electronic devices, which is likely more applicable to HRV assessment than to the standard autonomic cardiovascular test battery. HRV can be continuously and non-invasively monitored with modern digital health technologies\u0026mdash;offering an unprecedented opportunity for automated, real-time disease tracking. The leveraging of AI-integrated biosensors that continuously capture HRV dynamics is expected to significantly improve the breath and quality of real-time feedback provided to clinicians and researchers. This is expected to enable proactive, data-driven disease management, allowing for earlier detection of disease worsening and more personalized interventions, whenever available. However, the use of this non-linear approach to HRV analysis requires further investigation to better understand the underlying physiological mechanisms.\u003c/p\u003e\n\u003ch3\u003eComplementary Endpoints\u003c/h3\u003e\n\u003cp\u003eAnother putative implication of these findings lies in their potential impact on clinical trials. Traditional MSA trial endpoints often rely on subjective, symptom-based scales such as the patient-completed UMSARS Part I, which may lack sensitivity to subtle physiological changes. While regulatory agencies prioritize patient-centered outcomes, integrating HRV as an exploratory endpoint\u0026mdash;particularly in early-phase (e.g., Phase 2) trials\u0026mdash;could offer a sensitive measure of physiological response, complementing conventional clinical scales. Over time, if HRV demonstrates consistent association with clinically meaningful outcomes, it could evolve into a surrogate marker. This is especially relevant in a heterogeneous condition like MSA, where current tools may not fully capture the biological impact of interventions. Outside clinical trials, HRV may also serve a valuable prognostic role by improving prediction of mortality and disease trajectory. Future studies could explore HRV-guided patient stratification, identifying autonomic phenotypes that may respond differently to targeted interventions.\u003c/p\u003e\u003cdiv id=\"Sec37\" class=\"Section2\"\u003e\u003ch2\u003eA Risk Marker or a Modifiable Risk Factor?\u003c/h2\u003e\u003cp\u003eAn important unresolved question is whether altered HRV in MSA is merely a risk marker or also a modifiable risk factor. If HRV is just a marker, it provides valuable prognostic information but does not necessarily influence disease progression. However, if HRV is a true risk factor, then interventions that modulate autonomic function - such as pharmacological agents, neuromodulation - may have the potential to alter disease trajectory. Future interventional studies should explore whether targeting autonomic dysfunction could slow disease progression, improve quality of life, or even extend survival. If so, HRV could move beyond a diagnostic and prognostic biomarker to become a therapeutic target.\u003c/p\u003e\u003cdiv id=\"Sec38\" class=\"Section3\"\u003e\u003ch2\u003eThe Future of Large-Scale, Ambulatory Data Collection\u003c/h2\u003e\u003cp\u003eTo fully harness the power of HRV in MSA, large-scale, ambulatory and stress exposure studies are needed. Current research relies on controlled, in-clinic HRV measurements, which may not fully capture the dynamic nature of autonomic dysfunction in daily life. Expanding HRV analysis to real-world, long-term monitoring through integrated sensors could improve how we study and manage neurodegenerative diseases, including more prevalent conditions such as Alzheimer's and Parkinson\u0026rsquo;s disease (e.g., [\u003cspan additionalcitationids=\"CR48 CR49 CR50\" citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]). A large-scale HRV data repository could enable the development of predictive models to refine risk stratification and provide personalized disease trajectories across a spectrum of neurodegenerative disorders.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eHRV-based prognostic estimation using non-linear features provided performance comparable to classical cardiovascular assessments. Moreover, HRV significantly enhanced the prognostic utility of the UMSARS clinical scale. These findings establish HRV as a valuable autonomic biomarker offering independent prognostic information in MSA. Although HRV alone does not strongly predict clinical severity as traditionally defined, its added value lies in its simplicity, speed, and feasibility\u0026mdash;requiring only a brief (e.g., 5-minute) heart rate recording using easy-to-deploy equipment, in contrast to more complex tools or prolonged monitoring. Its potential for real-time, automated data capture further supports its relevance in both research and clinical settings. While the current analysis focused solely on baseline HRV, future work should explore longitudinal changes to better understand its role in disease progression. Building upon these insights will benefit from integrating HRV into multi-modal biomarker frameworks, refining clinical trial endpoints, and scaling up real-world ambulatory data collection.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eACME\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eAverage Causal Mediated Effect\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eADE\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eAverage Direct Effect\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eANS\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eAutonomic nervous system\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eBIC\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eBayesian Information Criterion\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eBP\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eBlood pressure\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eCAN\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eCardiovascular autonomic neuropathy\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eCE\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eConditional Entropy\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eDB\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDeep Breathing Test\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eDFA\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDetrended fluctuation analysis\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eECG\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eElectrocardiogram\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eHG\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eIsometric Handgrip Test\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eHR\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHazard ratio\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eHUTT\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHead-Up Tilt Test\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eHRV\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHeart rate variability\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eLOESS\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLocally weighted regression\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eMSA\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMultiple system atrophy\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eOH\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eOrthostatic hypotension\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eRSS\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eResidual Sum of Squares\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eRMSE\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eRoot Mean Squared Error\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eSHAP\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSHapley Additive exPlanations\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eST\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eStand Test\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eUMSARS\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eUnified Multiple System Atrophy Rating Scale\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eVM\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eValsalva Maneuver\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003ch2\u003eEthics approval and consent to participate\u003c/h2\u003e\n\u003cp\u003eThis study was conducted in accordance with the Declaration of Helsinki and approved by the Toulouse University Hospital Ethics Committee as part of a broader prospective longitudinal study on natural MSA progression (CNIL 1338780, CCTIRS 10.065).\u003c/p\u003e\n\u003ch2\u003eConsent for publication\u003c/h2\u003e\n\u003cp\u003eAll patients have consented to be part of this study and have their data published.\u003c/p\u003e\n\u003ch2\u003eCompeting interests\u003c/h2\u003e\n\u003cp\u003eMargherita Fabbri received Honoraria to speak from AbbVie, ORKYN, and BIAL, consultancies from BIAL and LVL Medical; Grant from France Parkinson, HORIZON 2022 French Ministry of Health and MSA Coalition. Professor Olivier Rascol has received honorarium for scientific advices from Lundbeck, ONO Pharma and TEVA and scientific grants from ARAMISE, MSA Coalition, Lundbeck, ONO Pharma and Takeda.\u003c/p\u003e\n\u003ch2\u003eAuthors\u0026rsquo;contributions\u003c/h2\u003e\n\u003cp\u003e(1) Research Project: A. Conception, B. Organization, C. Execution; (2) Data Collection: A. Design, B. Recruitment and Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the First Draft, B. Review and Critique.\u003c/p\u003e\n\u003cp\u003eP.B.: 1A, 1B, 1C, 3A, 3B.\u003c/p\u003e\n\u003cp\u003eM.K.: 1B, 2C, 3B.\u003c/p\u003e\n\u003cp\u003eM.F.: 3B.\u003c/p\u003e\n\u003cp\u003eF.R.:1C, 2C, 3B.\u003c/p\u003e\n\u003cp\u003eV.P.: 1C, 2C, 3B.\u003c/p\u003e\n\u003cp\u003eF.O.M.: 3B.\u003c/p\u003e\n\u003cp\u003eC.L. 3B.\u003c/p\u003e\n\u003cp\u003eO.R.: 3B.\u003c/p\u003e\n\u003cp\u003eW.G.M.: 3B.\u003c/p\u003e\n\u003cp\u003eA.F.S.: 3B.\u003c/p\u003e\n\u003cp\u003eD.B.: 3B.\u003c/p\u003e\n\u003cp\u003eC.P.L: 3B.\u003c/p\u003e\n\u003cp\u003eA.P.T.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eNo funding was received for this work.\u003c/p\u003e\n\u003ch2\u003eAcknowledgements\u003c/h2\u003e\n\u003cp\u003eSeveral authors of this publication are members of the European Reference Network for Rare Neurological Diseases - Project ID No 101085584.\u003c/p\u003e\n\u003ch2\u003eAvailability of data and materials\u003c/h2\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003ch2\u003eData availability\u003c/h2\u003e\n\u003cp\u003eAll data can be made available upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003ePoewe W, Stankovic I, Halliday G, Meissner WG, Wenning GK, Pellecchia MT et al (2022) Multiple system atrophy. Nat Rev Dis Primer 8:56\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFoubert-Samier A, Pavy-Le Traon A, Guillet F, Le-Goff M, Helmer C, Tison F et al (2020) Disease progression and prognostic factors in multiple system atrophy: A prospective cohort study. Neurobiol Dis 139:104813\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEschlboeck S, Goebel G, Eckhardt C, Fanciulli A, Raccagni C, Boesch S et al (2023) Development and Validation of a Prognostic Model to Predict Overall Survival in Multiple System Atrophy. Mov Disord Clin Pract 10:1368\u0026ndash;1376\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKrismer F, Palma J, Calandra-Buonaura G, Stankovic I, Vignatelli L, Berger A et al (2022) The Unified Multiple System Atrophy Rating Scale: Status, Critique, and Recommendations. Mov Disord 37:2336\u0026ndash;2341\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePalma J-A, Vernetti PM, Perez MA, Krismer F, Seppi K, Fanciulli A et al (2021) Limitations of the Unified Multiple System Atrophy Rating Scale as outcome measure for clinical trials and a roadmap for improvement. Clin Auton Res 31:157\u0026ndash;164\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWenning GK, Stankovic I, Vignatelli L, Fanciulli A, Calandra-Buonaura G, Seppi K et al (2022) The Movement Disorder Society Criteria for the Diagnosis of Multiple System Atrophy. Mov Disord 37:1131\u0026ndash;1148\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFoubert-Samier A, Pavy-Le Traon A, Guillet F, Le-Goff M, Helmer C, Tison F et al (2020) Disease progression and prognostic factors in multiple system atrophy: A prospective cohort study. Neurobiol Dis 139:104813\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKong SDX, Gordon CJ, Hoyos CM, Wassing R, D\u0026rsquo;Rozario A, Mowszowski L et al (2023) Heart rate variability during slow wave sleep is linked to functional connectivity in the central autonomic network. Brain Commun 5:fcad129\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOrini M, Van Duijvenboden S, Young WJ, Ram\u0026iacute;rez J, Jones AR, Hughes AD et al (2023) Long-term association of ultra-short heart rate variability with cardiovascular events. Sci Rep 13:18966\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFurushima H, Shimohata T, Nakayama H, Ozawa T, Chinushi M, Aizawa Y et al (2012) Significance and usefulness of heart rate variability in patients with multiple system atrophy. Mov Disord 27:570\u0026ndash;574\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKong SDX, Gordon CJ, Hoyos CM, Wassing R, D\u0026rsquo;Rozario A, Mowszowski L et al (2023) Heart rate variability during slow wave sleep is linked to functional connectivity in the central autonomic network. Brain Commun 5:fcad129\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMemon AA, George EB, Nazir T, Sunkara Y, Catiul C, Amara AW (2024) Heart rate variability during sleep in synucleinopathies: a review. Front Neurol 14:1323454\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHeimrich KG, Lehmann T, Schlattmann P, Prell T (2021) Heart Rate Variability Analyses in Parkinson\u0026rsquo;s Disease: A Systematic Review and Meta-Analysis. Brain Sci 11:959\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIniguez M, Jimenez-Marin A, Erramuzpe A, Acera M, Tijero B, Murueta-Goyena A et al (2022) Heart-brain synchronization breakdown in Parkinson\u0026rsquo;s disease. Npj Park Dis 8:64\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFurushima H, Shimohata T, Nakayama H, Ozawa T, Chinushi M, Aizawa Y et al (2012) Significance and usefulness of heart rate variability in patients with multiple system atrophy. Mov Disord 27:570\u0026ndash;574\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaestri R, Pinna GD, Porta A, Balocchi R, Sassi R, Signorini MG et al (2007) Assessing nonlinear properties of heart rate variability from short-term recordings: are these measurements reliable? Physiol Meas 28:1067\u0026ndash;1077\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuikuri HV, Perki\u0026ouml;m\u0026auml;ki JS, Maestri R, Pinna GD (2009) Clinical impact of evaluation of cardiovascular control by novel methods of heart rate dynamics. Philos Trans R Soc Math Phys Eng Sci 367:1223\u0026ndash;1238\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSassi R, Cerutti S, Lombardi F, Malik M, Huikuri HV, Peng C-K et al (2015) Advances in heart rate variability signal analysis: joint position statement by the e-Cardiology ESC Working Group and the European Heart Rhythm Association co-endorsed by the Asia Pacific Heart Rhythm Society. Europace 17:1341\u0026ndash;1353\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGilman S, Wenning GK, Low PA, Brooks DJ, Mathias CJ, Trojanowski JQ et al (2008) Second consensus statement on the diagnosis of multiple system atrophy. Neurology 71:670\u0026ndash;676\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFoubert-Samier A, Pavy-Le Traon A, Guillet F, Le-Goff M, Helmer C, Tison F et al (2020) Disease progression and prognostic factors in multiple system atrophy: A prospective cohort study. Neurobiol Dis 139:104813\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSletten DM, Suarez GA, Low PA, Mandrekar J, Singer W (2012) COMPASS 31: A Refined and Abbreviated Composite Autonomic Symptom Score. Mayo Clin Proc. ;87:1196\u0026ndash;201\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWenning GK, Tison F, Seppi K, Sampaio C, Diem A, Yekhlef F et al (2004) Development and validation of the Unified Multiple System Atrophy Rating Scale (UMSARS). Mov Disord 19:1391\u0026ndash;1402\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVisser M, Marinus J, Stiggelbout AM, Van Hilten JJ (2004) Assessment of autonomic dysfunction in Parkinson\u0026rsquo;s disease: The SCOPA-AUT. Mov Disord 19:1306\u0026ndash;1312\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEwing DJ, Campbell IW, Clarke BF (1980) Assessment of Cardiovascular Effects in Diabetic Autonomic Neuropathy and Prognostic Implications. Ann Intern Med 92:308\u0026ndash;311\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEwing DJ, Clarke BF (1986) Autonomic neuropathy: its diagnosis and prognosis. Clin Endocrinol Metab 15:855\u0026ndash;888\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMathias CJ, Bannister SR (eds) (2013) Autonomic Failure [Internet]. Oxford University Press; [cited 2025 Apr 27]. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://academic.oup.com/book/24366\u003c/span\u003e\u003cspan address=\"https://academic.oup.com/book/24366\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEwing DJ, Martyn CN, Young RJ, Clarke BF (1985) The Value of Cardiovascular Autonomic Function Tests: 10 Years Experience in Diabetes. Diabetes Care 8:491\u0026ndash;498\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePichot V, Roche F, Celle S, Barth\u0026eacute;l\u0026eacute;my J-C, Chouchou F, HRVanalysis: A Free Software for Analyzing Cardiac Autonomic Activity. Front Physiol [Internet]. 2016 [cited 2025 Mar 19];7. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://journal.frontiersin.org/article/\u003c/span\u003e\u003cspan address=\"http://journal.frontiersin.org/article/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fphys.2016.00557/full\u003c/span\u003e\u003cspan address=\"10.3389/fphys.2016.00557/full\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHeart rate variability (1996) : standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Circulation 93:1043\u0026ndash;1065\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLa Rovere MT, Pinna GD, Raczak G (2008) Baroreflex Sensitivity: Measurement and Clinical Implications. Ann Noninvasive Electrocardiol 13:191\u0026ndash;207\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDi Rienzo M, Castiglioni P, Iellamo F, Volterrani M, Pagani M, Mancia G et al (2008) Dynamic adaptation of cardiac baroreflex sensitivity to prolonged exposure to microgravity: data from a 16-day spaceflight. J Appl Physiol 105:1569\u0026ndash;1575\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIodice V, Lipp A, Ahlskog JE, Sandroni P, Fealey RD, Parisi JE et al (2012) Autopsy confirmed multiple system atrophy cases: Mayo experience and role of autonomic function tests. J Neurol Neurosurg Psychiatry 83:453\u0026ndash;459\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eStankovic I, Fanciulli A, Kostic VS, Krismer F, Meissner WG, Palma JA et al (2021) Laboratory-Supported Multiple System Atrophy beyond Autonomic Function Testing and Imaging: A Systematic Review by the MoDiMSA Study Group. Mov Disord Clin Pract 8:322\u0026ndash;340\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGolden EP, McCreary M, Vernino S (2022) Responsiveness of UMSARS and other clinical measures in a longitudinal structured care clinic for multiple system atrophy. Clin Auton Res 32:477\u0026ndash;484\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu W-Z, Cao L-X, Yin J-H, Zhao X-S, Piao Y-S, Gu W-H et al (2023) Non-motor symptoms in multiple system atrophy: A comparative study with Parkinson\u0026rsquo;s disease and progressive supranuclear palsy. Front Neurol 13:1081219\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eK\u0026ouml;llensperger M, Stampfer-Kountchev M, Seppi K, Geser F, Frick C, Del Sorbo F et al (2007) Progression of dysautonomia in multiple system atrophy: a prospective study of self‐perceived impairment. Eur J Neurol 14:66\u0026ndash;72\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eD\u0026rsquo;Addio G, Corbi G, Accardo A, Russo G, Ferrara N, Mazzoleni MC et al (2009) Fractal behaviour of heart rate variability reflects severity in stroke patients. Stud Health Technol Inf 150:794\u0026ndash;798\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGomolka RS, Kampusch S, Kaniusas E, Th\u0026uuml;rk F, Sz\u0026eacute;les JC, Klonowski W (2018) Higuchi Fractal Dimension of Heart Rate Variability During Percutaneous Auricular Vagus Nerve Stimulation in Healthy and Diabetic Subjects. Front Physiol 9:1162\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGao C, Lim ASP, Haghayegh S, Cai R, Yang J, Yu L et al (2025) Reduced Complexity of Pulse Rate Is Associated With Faster Cognitive Decline in Older Adults. J Am Heart Assoc 14:e041448\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKuusela TA, Jartti TT, Tahvanainen KUO, Kaila TJ (2002) Nonlinear methods of biosignal analysis in assessing terbutaline-induced heart rate and blood pressure changes. Am J Physiol-Heart Circ Physiol 282:H773\u0026ndash;H781\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXin Y, Zhao Y (2017) Paroxysmal atrial fibrillation recognition based on multi-scale wavelet α-entropy. Biomed Eng OnLine 16:121\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eM\u0026auml;kikallio TH, Sepp\u0026Auml;NEN T, Niemel\u0026auml; M, Airaksinen KEJ, Tulppo M, Huikuri HV (1996) Abnormalities in Beat to Beat Complexity of Heart Rate Dynamics in Patients With a Previous Myocardial Infarction11This study was supported by grants from the Finnish Foundation for Cardiovascular Research and the Medical Council of the Academy of Finland, Helsinki, Finland. J Am Coll Cardiol 28:1005\u0026ndash;1011\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYamada A, Hayano J, Sakata S, Okada A, Mukai S, Ohte N et al (2000) Reduced Ventricular Response Irregularity Is Associated With Increased Mortality in Patients With Chronic Atrial Fibrillation. Circulation 102:300\u0026ndash;306\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCoon EA, Sletten DM, Suarez MD, Mandrekar JN, Ahlskog JE, Bower JH et al (2015) Clinical features and autonomic testing predict survival in multiple system atrophy. Brain 138:3623\u0026ndash;3631\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVichayanrat E, Valerio F, Koay S, De Pablo-Fernandez E, Panicker J, Morris H et al Diagnosing Premotor Multiple System Atrophy: Natural History and Autonomic Testing in an Autopsy-Confirmed Cohort. Neurology [Internet]. 2022 [cited 2025 Apr 28];99. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.neurology.org/doi/\u003c/span\u003e\u003cspan address=\"https://www.neurology.org/doi/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1212/WNL.0000000000200861\u003c/span\u003e\u003cspan address=\"10.1212/WNL.0000000000200861\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePavy-Le Traon A, Foubert‐Samier A, Ory‐Magne F, Fabbri M, Senard J, Meissner WG et al (2022) Ambulatory blood pressure and drug treatment for orthostatic hypotension as predictors of mortality in patients with multiple system atrophy. Eur J Neurol 29:1025\u0026ndash;1034\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNatarajan A, Pantelopoulos A, Emir-Farinas H, Natarajan P (2020) Heart rate variability with photoplethysmography in 8 million individuals: a cross-sectional study. Lancet Digit Health 2:e650\u0026ndash;e657\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBrown AA, Ferguson BJ, Jones V, Green BE, Pearre JD, Anunoby IA et al (2022) Pilot Study of Real-World Monitoring of the Heart Rate Variability in Amyotrophic Lateral Sclerosis. Front Artif Intell 5:910049\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLeng Y, Cotton-Clay A, Baron S, Fava L, Johnson K, Easwar V et al (2024) Real-world Study of Heart Rate Variability During Sleep Using Under-mattress Sensors in over 30,000 Individuals. Sleep 47:A208\u0026ndash;A208\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou J, Huebner G, Liu KY, Ucci M (2024) Heart rate variability, electrodermal activity and cognition in adults: Association with short-term indoor PM2.5 exposure in a real-world intervention study. Environ Res 263:120245\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSrinivasan AG, Smith SS, Pattinson CL, Mann D, Sullivan K, Salmon P et al (2024) Heart rate variability as an indicator of fatigue: A structural equation model approach. Transp Res Part F Traffic Psychol Behav 103:420\u0026ndash;429\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"clinical-autonomic-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"autr","sideBox":"Learn more about [Clinical Autonomic Research](http://link.springer.com/journal/10286)","snPcode":"10286","submissionUrl":"https://www.editorialmanager.com/autr/default2.aspx","title":"Clinical Autonomic Research","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"HRV, MSA, neurodegeneration, dysautonomia","lastPublishedDoi":"10.21203/rs.3.rs-7325643/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7325643/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e\u003cp\u003eMultiple system atrophy (MSA) is a progressive neurodegenerative disorder characterized by autonomic dysfunction, parkinsonism, and cerebellar impairment. Predicting disease severity and survival remains challenging due to the heterogeneity of disease progression. Heart rate variability (HRV), a non-invasive measure of autonomic nervous system function, has been used as a biomarker of autonomic failure. However, the role of non/linear HRV in MSA remains underexplored, and its prognostic value is yet to be fully established.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eThis study aimed to evaluate the predictive value of HRV features in MSA, identifying HRV features most predictive of mortality and survival, and assessing whether HRV provides unique, complementary insights beyond traditional clinical severity measures (n\u0026thinsp;=\u0026thinsp;214). Regression models were employed to assess the association between HRV features and disease severity, as assessed by the Unified MSA Rating Scale (UMSARS), or time-to-death. Survival analyses were used to investigate HRV\u0026rsquo;s prognostic value. Mediation analysis explored the relationship between HRV, UMSARS, and survival.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eHRV features demonstrated negative correlations with disease severity, mirroring clinical deterioration. While no single HRV feature showed strong correlations with the UMSARS, their combination was a significant predictor. HRV alone predicted time-to-death almost as well as the UMSARS and combining HRV with UMSARS significantly improved survival prediction accuracy. HRV maintained a direct effect on survival, independent of the UMSARS, highlighting its distinct physiological relevance.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eHRV provides valuable, complementary information beyond UMSARS in predicting disease severity and survival in MSA. While HRV alone has only moderate predictive power, it captures distinct physiological processes not reflected by traditional clinical scales.\u003c/p\u003e","manuscriptTitle":"Heart Rate Variability Provides Prognostic value in Multiple System Atrophy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-10 18:39:41","doi":"10.21203/rs.3.rs-7325643/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2025-09-04T13:29:27+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-04T08:57:36+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-10T14:16:23+00:00","index":"","fulltext":""},{"type":"submitted","content":"Clinical Autonomic Research","date":"2025-08-08T05:17:33+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"clinical-autonomic-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"autr","sideBox":"Learn more about [Clinical Autonomic Research](http://link.springer.com/journal/10286)","snPcode":"10286","submissionUrl":"https://www.editorialmanager.com/autr/default2.aspx","title":"Clinical Autonomic Research","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"c5d9a510-fb03-4a41-a3b1-c78b656990e3","owner":[],"postedDate":"September 10th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-02T16:02:23+00:00","versionOfRecord":{"articleIdentity":"rs-7325643","link":"https://doi.org/10.1007/s10286-026-01190-8","journal":{"identity":"clinical-autonomic-research","isVorOnly":false,"title":"Clinical Autonomic Research"},"publishedOn":"2026-02-23 15:58:53","publishedOnDateReadable":"February 23rd, 2026"},"versionCreatedAt":"2025-09-10 18:39:41","video":"","vorDoi":"10.1007/s10286-026-01190-8","vorDoiUrl":"https://doi.org/10.1007/s10286-026-01190-8","workflowStages":[]},"version":"v1","identity":"rs-7325643","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7325643","identity":"rs-7325643","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}