The relationship between dietary vitamin B1 and stroke: a machine learning analysis of NHANES data.

BACKGROUND

Vitamin B1 deficiency is closely linked to damage in the cardiovascular system. However, the relationship between dietary Vitamin B1 intake and the risk of stroke remains ambiguous and requires further investigation.

METHODS

This study analyzed data from participants in the National Health and Nutrition Examination Survey (NHANES: 2005-2018) to investigate the relationship between dietary vitamin B1 and ischemic stroke. Weighted multivariable logistic regression models and restricted cubic spline (RCS) regression were employed to explore potential nonlinear relationships, and subgroup analyses were conducted to assess the robustness of the results. Additionally, the Least Absolute Shrinkage and Selection Operator (LASSO) was utilized for feature selection. Eight machine learning methods were employed to construct predictive models and evaluate their performance. Based on the best-performing model, we further examined variable importance and model accuracy, employing Shapley Additive Explanations (SHAP) analysis to interpret the model. Finally, a nomogram was created to enhance the readability of the predictive model results.

RESULTS

After controlling for various variables, vitamin B1 exhibited a significant linear negative correlation with stroke risk. In comparison to the lowest quartile, the adjusted odds ratio (OR) for the fourth quartile was notably reduced to 0.66 (95% CI: 0.46, 0.94). Restricted cubic spline (RCS) analysis further confirmed a linear inverse relationship between vitamin B1 levels and stroke risk. Moreover, the Gradient Boosting Machine (GBM) model demonstrated robust predictive efficacy, achieving an area under the curve (AUC) of 91.9%.

CONCLUSION

A large-scale study based on NHANES indicates that as dietary intake of vitamin B1 increases, the risk of stroke shows a gradual decline. Therefore, appropriately increasing dietary intake of vitamin B1 may reduce the risk of stroke occurrence.