Scholarlink

Background <p>Schizophrenia (SCZ) patients have high violent and aggressive behavior (VAB) prevalence, causing harm and stigma. The non-high-density lipoprotein cholesterol to high-density lipoprotein cholesterol ratio (NHHR) and triglyceride-glucose index (TyG) show promise for psychiatric disorder prediction, but their links to SCZ patients’ VAB risk remain understudied.</p> Methods <p>A cross-sectional study included 102 hospitalized SCZ patients from May 2023–October 2024, stratified by Broset Violence Checklist (BVC) into low/medium/high VAB risk groups. Positive and Negative Syndrome Scale (PANSS) quantified psychiatric severity; NHHR and TyG were calculated from venous biochemical indices. Analyses included multivariate logistic regression, mediation analysis (PANSS Scores as mediator), 7 machine learning (ML) models with SHAP interpretation, and FDA Adverse Event Reporting System (FAERS) disproportionality analysis.</p> Results <p>In fully adjusted models, NHHR and TyG were both negatively associated with VAB risk (NHHR: OR = 0.43, 95%CI:0.20–0.92, <i>p</i> = 0.0285; TyG: OR = 0.29, 95%CI:0.12–0.67, <i>p</i> = 0.0038); further analysis revealed an atypical inverted U-shaped correlation between NHHR and VAB risk, with the inflection point measured at 2.98. Meanwhile, PANSS scores partially mediated the associations between NHHR, TyG and VAB, and TyG had higher mediation proportions (54.2% for BVC scores, 57.8% for VAB risk) than NHHR. In ML-based VAB prediction, LASSO Logistic Regression was the optimal model (Accuracy = 0.8065, highest F1 Score), and SHAP analysis confirmed NHHR and TyG as core predictive features. Additionally, FAERS analysis identified positive metabolic abnormality signals for olanzapine (early-onset adverse events) and quetiapine (late-onset adverse events).</p> Conclusion <p>NHHR and TyG are potential VAB risk biomarkers for SCZ patients, with psychiatric severity mediating their effects. ML and FAERS provide clinical prediction and medication monitoring references.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploration of the associations between violent and aggressive behaviors risk and NHHR, TyG in patients with schizophrenia: an SHAP interpretable machine learning prediction and FAERS pharmacovigilance analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月16日

查看全文
  • Guangwei Qing,
  • Hanggai Taorigen,
  • Yuxin Zhou,
  • Gang Wang

摘要

Background

Schizophrenia (SCZ) patients have high violent and aggressive behavior (VAB) prevalence, causing harm and stigma. The non-high-density lipoprotein cholesterol to high-density lipoprotein cholesterol ratio (NHHR) and triglyceride-glucose index (TyG) show promise for psychiatric disorder prediction, but their links to SCZ patients’ VAB risk remain understudied.

Methods

A cross-sectional study included 102 hospitalized SCZ patients from May 2023–October 2024, stratified by Broset Violence Checklist (BVC) into low/medium/high VAB risk groups. Positive and Negative Syndrome Scale (PANSS) quantified psychiatric severity; NHHR and TyG were calculated from venous biochemical indices. Analyses included multivariate logistic regression, mediation analysis (PANSS Scores as mediator), 7 machine learning (ML) models with SHAP interpretation, and FDA Adverse Event Reporting System (FAERS) disproportionality analysis.

Results

In fully adjusted models, NHHR and TyG were both negatively associated with VAB risk (NHHR: OR = 0.43, 95%CI:0.20–0.92, p = 0.0285; TyG: OR = 0.29, 95%CI:0.12–0.67, p = 0.0038); further analysis revealed an atypical inverted U-shaped correlation between NHHR and VAB risk, with the inflection point measured at 2.98. Meanwhile, PANSS scores partially mediated the associations between NHHR, TyG and VAB, and TyG had higher mediation proportions (54.2% for BVC scores, 57.8% for VAB risk) than NHHR. In ML-based VAB prediction, LASSO Logistic Regression was the optimal model (Accuracy = 0.8065, highest F1 Score), and SHAP analysis confirmed NHHR and TyG as core predictive features. Additionally, FAERS analysis identified positive metabolic abnormality signals for olanzapine (early-onset adverse events) and quetiapine (late-onset adverse events).

Conclusion

NHHR and TyG are potential VAB risk biomarkers for SCZ patients, with psychiatric severity mediating their effects. ML and FAERS provide clinical prediction and medication monitoring references.