Evaluating the predictive potential of Th1 (IFN-γ⁺CD4⁺)/CD4⁺ in rapidly progressive amyotrophic lateral sclerosis

Th1 (IFN-γ⁺CD4⁺)/CD4⁺ cells exacerbate the release of pro-inflammatory cytokines, contributing to neuronal death. It is proposed that the peripheral immune system plays a pivotal role in the pathophysiology of amyotrophic lateral sclerosis (ALS). This study aims to develop an interpretable machine learning model based on blood Th1/CD4⁺ cells to predict rapidly progressive ALS.

We enrolled 564 patients with sporadic ALS who met the eligibility inclusion criteria for further analysis. Immune cells and cytokines were quantified using flow cytometric cell counting and a flow cytometry-based fluorescent bead capture assay. Multivariate Cox proportional hazards models and restricted cubic spline analyses were applied to estimate the correlation between Th1/CD4⁺ cells and rapidly progressive ALS. The important variables identified through LASSO regression analysis were incorporated into the development of the machine learning model.

The multivariate Cox proportional hazards model revealed that, compared to the low Th1/CD4⁺ group (Th1/CD4⁺  < 16.21), the high Th1/CD4⁺ group (Th1/CD4⁺  ≥ 16.21) was positively associated with the rate of ALS progression (HR: 1.90, 95% CI: 1.34-2.70). Th1/CD4⁺ is also associated with the decline in forced vital capacity (r = 0.11, P = 0.01). The machine learning model was built using Th1/CD4⁺ in combination with the other 4 features. Xgboost performed best in the validation cohort, achieving an AUC of 0.804 and a G mean of 0.756.

Th1/CD4⁺ (with an optimal cutoff value of 16.21) was established as an independent risk factor for rapid progression in ALS. The machine learning model incorporating Th1/CD4⁺ demonstrated strong predictive performance.

The prospective cohort study is registered with the Chinese Clinical Trial Registry (ID: ChiCTR2400079885) ( http://www.chictr.org.cn/ ).