Related Experiment Video
Updated: Jan 13, 2026

A Protocol for Comprehensive Assessment of Bulbar Dysfunction in Amyotrophic Lateral Sclerosis ALS
Published on: February 21, 2011
Voice-Based Prediction of Survival in Amyotrophic Lateral Sclerosis (ALS) Patients Using Biomechanical Acoustic
Margarita Pérez-Bonilla1, Paola Díaz Borrego2, Marina Mora-Ortiz3
1Physical Medicine & Rehabilitation, Reina Sofía University Hospital, Córdoba 14004, Spain; Department of Applied Physics, Radiology and Physical Medicine, Faculty of Medicine and Nursing of Córdoba, Córdoba 14004, Spain.
Objective:
To evaluate whether voice-derived acoustic and biomechanical features can serve as non-invasive biomarkers for mortality-risk prediction and survival stratification in patients with amyotrophic lateral sclerosis (ALS).
Methods:
We conducted a retrospective study including 50 ALS patients evaluated in a phoniatrics consultation with available sustained vowel recordings, demographic data, and functional assessments. Nested logistic regression models were developed to predict clinical outcomes, progressively incorporating demographic variables, functional indices (Grade, Roughness, Breathiness, Asthenia, Strain, and Barthel), acoustic features (fundamental frequency, jitter, shimmer, harmonics-to-noise ratio), and biomechanical voice parameters (Pr1-Pr22). Model performance was assessed using receiver operating characteristic curves and area under the curve (AUC) comparisons via DeLong tests. Stepwise Akaike Information Criterion (StepAIC) was applied to optimize the final model. A Cox proportional hazards model was used to evaluate the association between voice parameters and survival time.
Results:
The final StepAIC model, which included a subset of biomechanical features, achieved excellent predictive performance (AUC = 0.903, 95% confidence interval: 0.816-0.989), significantly outperforming baseline and acoustic-only models. Bootstrapping confirmed the model's robustness and generalizability. Cox regression analysis showed that the derived risk scores stratified patients into tertiles with significantly different survival probabilities (log-rank P < 0.0001; hazard ratio for high vs. low-risk group = 11.2).
Conclusion:
Biomechanical voice features are strong predictors of mortality in ALS and outperform traditional clinical and acoustic indices. These findings support the integration of voice analysis into ALS monitoring protocols as a non-invasive, cost-effective, and scalable prognostic tool.

