Efficacy and neuroprotective effects of tetramethylpyrazine in animal models of spinal cord injury: a systematic review, meta-analysis, and exploratory machine-learning study.

BACKGROUND

Spinal cord injury (SCI) is a severe neurological disorder characterized by persistent motor dysfunction and complex secondary injury processes. Tetramethylpyrazine (TMP) has shown potential neuroprotective effects in experimental SCI; however, the overall efficacy and factors associated with its therapeutic effects remain unclear. This study systematically evaluated the efficacy of TMP in animal models of SCI and explored potential efficacy-related experimental factors using machine-learning methods.

METHODS

PubMed, Embase, Web of Science, CNKI, Wanfang and VIP databases were searched for controlled animal studies investigating TMP treatment in SCI. Hedges' g and corresponding 95% confidence intervals were calculated using random-effects models. Subgroup analyses, sensitivity analyses and assessments of small-study effects were performed where sufficient data were available. An exploratory XGBoost model combined with SHAP analysis was used to evaluate study-level characteristics potentially associated with BBB effect sizes. Model performance was assessed using repeated 10-fold cross-validation with 20 repetitions, and feature-importance stability was evaluated using 1,000 bootstrap resamples.

RESULTS

Forty-two animal studies were included. Compared with control treatment, TMP significantly improved locomotor recovery, with progressively greater improvements in BBB scores from day 1 to day 28 after injury. TMP also improved inclined-plane performance at most evaluated time points. Biochemical analyses indicated that TMP alleviated oxidative stress by reducing malondialdehyde levels and increasing superoxide dismutase activity. TMP treatment was also associated with reduced concentrations of several pro-inflammatory cytokines and improvements in apoptosis-related outcomes. Repeated 10-fold cross-validation showed poor predictive performance (median R2 = -0.314). Bootstrap analyses relatively consistently ranked injury level and sex among the most important features; however, these findings were interpreted as exploratory and hypothesis-generating rather than as validated predictors.

CONCLUSION

The available preclinical evidence indicates that TMP promotes locomotor recovery and exerts antioxidant, anti-inflammatory and anti-apoptotic effects in animal models of SCI. The exploratory machine-learning analysis did not demonstrate reliable predictive performance, and the observed feature-importance patterns should be regarded as hypothesis-generating. Further standardized animal experiments and well-designed translational studies are required.

SYSTEMATIC REVIEW REGISTRATION: https://www.crd.york.ac.uk/PROSPERO/view/CRD420261459838, identifier CRD420261459838.