What do ALS motor phenotypes measure? A population-based decomposition of their prognostic dimensions

The ALS-OPM 3.3 classification stratifies incident ALS through three axes: onset region (O), propagation time (P) and motor-neuron pattern (M). We assessed whether OPM 3.3 improves prognostic discrimination over established classifications and characterised how it decomposes their prognostic signal in a population-based cohort.

In the prospective Piedmont and Aosta Valley ALS Register (2000-2022), patients were classified on the three axes; M was operationalised as M_3class (M0 balanced, M1d UMN-predominant, M2d LMN-predominant), with the PLS-spectrum M1p analysed separately. Outcome was survival (death or tracheostomy) from onset. Multivariable Cox models (O, P1(n), M_3class, age, sex) underwent bootstrap optimism-correction and fivefold cross-validation, and were compared against alternative taxonomies and clinical staging systems.

Amongst 2,738 non-PLS patients (2,561 events, 93.5%), median survival differed across M classes (M0 27.0, M2d 36.0, M1d 41.9 months). OPM 3.3 outperformed the bulbar/spinal dichotomy but matched the classical phenotype (optimism-corrected C-index 0.703, 95% CI 0.685-0.719 vs 0.701). In the arm-proximal cell (95.7% M2d), the O2p protective effect vanished after M adjustment (HR 0.99, 95% CI 0.79-1.24), whilst M2d retained HR 0.83 (95% CI 0.72-0.95), showing prognosis reflects motor-neuron pattern, not anatomy. Adding King's, Milano-Torino (MiToS), or Fine'til 9 (FT9) clinical staging further improved discrimination.

OPM 3.3 did not improve discrimination over established phenotypes but decomposed their prognostic signal into independently interpretable dimensions. Favourable prognosis of arm-proximal/flail-arm onset reflects the underlying LMN-predominant phenotype, not onset site. The motor-neuron axis is robust to the treatment of the propagation axis and independent of clinical staging.