Abstract
Abstract
Background
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder with substantial clinical heterogeneity, complicating the design and interpretation of clinical trials. Two strategies have been proposed to address this variability: phenotype-based prognostic modelling and biomarker-based stratification using neurofilaments. We aimed to determine whether a widely used phenotype-based method, the TRICALS risk score, could substitute for neurofilament biomarkers in a clinical trial.
Methods
We used data from the MIROCALS phase 2b randomised controlled trial of low-dose interleukin-2 in ALS (
n
= 220). The TRICALS risk score was calculated for each participant. The published biomarker model includes CSF phosphorylated neurofilament heavy chain (pNFH) and a pNFH × treatment interaction. We analysed the effects of replacing CSF pNFH with the TRICALS risk score and of adding TRICALS score as an additional covariate to the original model. Model performance was assessed using likelihood ratio χ
2
statistics and − 2 log likelihood (− 2LL). The primary outcome was survival.
Results
TRICALS risk score was moderately correlated with CSF pNFH (
r
= 0.357), explaining 12.7% of shared variance. Substituting CSF pNFH with TRICALS score reduced model performance (− 2LL 928.25 vs 793.36). The TRICALS × treatment interaction was statistically significant (hazard ratio [HR] 0.72, 95% CI 0.54–0.96
p
= 0.025), but contributed modestly to model fit (Δχ
2
= 4.76), compared with the pNFH × treatment interaction (Δχ
2
= 14.24). Incorporating the pNFH × treatment interaction significantly improved model fit and identified a treatment effect (HR 0.28, 95% CI 0.12–0.63;
p
= 0.002).
Conclusions
The TRICALS risk score does not substitute for CSF neurofilament biomarkers as a prognostic or stratification variable but provides complementary information.