Quantitative brainstem and spinal MRI in amyotrophic lateral sclerosis: implications for predicting noninvasive ventilation needs

BackgroundRespiratory complications resulting from motor neurons degeneration are the primary cause of death in amyotrophic lateral sclerosis (ALS). Predicting the need for non-invasive ventilation (NIV) in ALS is important for advance care planning and clinical trial design. The aim of this study was to assess the potential of quantitative MRI at the brainstem and spinal cord levels to predict the need for NIV during the first six months after diagnosis.MethodsForty-one ALS patients underwent MRI and spirometry shortly after diagnosis. The need for NIV was monitored according to French health guidelines for 6 months. The performance of four regression models based on: clinical variables, brainstem structures volumes, cervical spinal measurements, and combined variables were compared to predict the need for NIV within this period.ResultsBoth the clinical model (R2 = 0.28, AUC = 0.85, AICc = 42.67, BIC = 49.8) and the brainstem structures' volumes model (R2 = 0.30, AUC = 0.85, AICc = 40.13, BIC = 46.99) demonstrated good predictive performance. In addition, cervical spinal cord measurements model similar performance (R2 = 0.338, AUC = 0.87, AICc = 37.99, BIC = 44.49). Notably, the combined model incorporating predictors from all three models yielded the best performance (R2 = 0.60, AUC = 0.959, AICc = 36.38, BIC = 44.8). These findings are supported by observed positive correlations between brainstem volumes, cervical (C4/C7) cross-sectional area, and spirometry-measured lung volumes.ConclusionsOur study shows that brainstem volumes and spinal cord area are promising measures to predict respiratory intervention needs in ALS.

Automated summary

This study demonstrates that brainstem volumes and spinal cord area are promising indicators for predicting respiratory intervention needs in ALS patients, with the combined model incorporating predictors from multiple models yielding the best predictive performance.