Scholarlink

Background and objectives <p>The prognosis of patients with ischemic stroke (IS) is highly heterogeneous, influenced by both genetic and environmental factors. Traditional prognostic models for IS are primarily based on clinical variables, while the incremental value of integrating single nucleotide polymorphisms (SNPs) into these models remains unclear. This study aimed to evaluate the added predictive value of genetic polymorphisms beyond established clinical predictors for 90-day poor outcomes after IS.</p> Methods <p>Candidate SNPs associated with IS prognosis were identified through a PubMed search up to December 31, 2023. A retrospective analysis was conducted on a cohort of acute ischemic stroke patients who were admitted within 14&#xa0;days of onset between September 2016 and October 2020. The primary outcome was poor 90-day prognosis, defined as a modified Rankin Scale score of 3–6 or any composite clinical event (stroke recurrence, myocardial infarction, or all-cause death). SNPs were genotyped using multiplex tagged-amplicon deep sequencing. Three prediction models were developed using LASSO logistic regression: SNP model, clinical model and mixed model. Model performance was assessed by area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity on the full dataset and under tenfold-cross-validation. AUCs were compared between models using the DeLong test.</p> Results <p>Sixty-six candidate SNPs were identified from the literature, of which 59 passed quality control and were analyzed. Among 1,322 screened patients, 866 were included; 559 (64.6%) had good outcomes and 307 (35.4%) had poor outcomes at 90&#xa0;days. The SNP model, comprising 35 SNPs, showed moderate discrimination (AUC 0.696; mean cross-validated AUC 0.631 ± 0.046) with relatively high sensitivity but low specificity. The clinical model, including 5 core and 14 candidate clinical variables, achieved an AUC of 0.811 (mean cross-validated AUC 0.773 ± 0.049). The mixed model, incorporating 17 SNPs, 5 core clinical variables and 9 candidate clinical variables, provided the best performance (AUC 0.838; mean cross-validated AUC 0.792 ± 0.048) and a more favorable balance of accuracy, sensitivity and specificity than either model alone. Pairwise DeLong tests showed that the mixed model significantly outperformed both the clinical and SNP models (<i>P</i> &lt; 0.001). Seven SNPs remained independently significant in the mixed model: CYP4A11 rs9333025, NINJ2 rs12425791, PRKCH rs2230500, PTGS1 rs1330344, CYP2C19 rs4986893, VEGFA rs3025039 and MTHFR rs1801133.</p> Conclusions <p>Genetic polymorphisms provide modest but statistically significant incremental value to clinical prognostic models for ischemic stroke, yielding a small improvement in discrimination (AUC on the order of 0.01) over clinical predictors alone. While this gain is limited, it supports incorporating genetic information into outcome prediction, pending further validation in larger cohorts and across different stroke subtypes.</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

Incremental value of genetic polymorphisms in prognostic models for ischemic stroke

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月30日

查看全文
  • Dong Zhang,
  • Rui-Nan Ma,
  • Zhi-Zhang Li,
  • Ying Ding,
  • Xiao-Guang Zhang,
  • Jie Xue,
  • Yun-Qi Hu,
  • Dan-Zhuoma Ci,
  • Yue-Ying Bai,
  • Liang Hu,
  • Dai-Zhan Zhou,
  • Yun-Hua Yue

摘要

Background and objectives

The prognosis of patients with ischemic stroke (IS) is highly heterogeneous, influenced by both genetic and environmental factors. Traditional prognostic models for IS are primarily based on clinical variables, while the incremental value of integrating single nucleotide polymorphisms (SNPs) into these models remains unclear. This study aimed to evaluate the added predictive value of genetic polymorphisms beyond established clinical predictors for 90-day poor outcomes after IS.

Methods

Candidate SNPs associated with IS prognosis were identified through a PubMed search up to December 31, 2023. A retrospective analysis was conducted on a cohort of acute ischemic stroke patients who were admitted within 14 days of onset between September 2016 and October 2020. The primary outcome was poor 90-day prognosis, defined as a modified Rankin Scale score of 3–6 or any composite clinical event (stroke recurrence, myocardial infarction, or all-cause death). SNPs were genotyped using multiplex tagged-amplicon deep sequencing. Three prediction models were developed using LASSO logistic regression: SNP model, clinical model and mixed model. Model performance was assessed by area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity on the full dataset and under tenfold-cross-validation. AUCs were compared between models using the DeLong test.

Results

Sixty-six candidate SNPs were identified from the literature, of which 59 passed quality control and were analyzed. Among 1,322 screened patients, 866 were included; 559 (64.6%) had good outcomes and 307 (35.4%) had poor outcomes at 90 days. The SNP model, comprising 35 SNPs, showed moderate discrimination (AUC 0.696; mean cross-validated AUC 0.631 ± 0.046) with relatively high sensitivity but low specificity. The clinical model, including 5 core and 14 candidate clinical variables, achieved an AUC of 0.811 (mean cross-validated AUC 0.773 ± 0.049). The mixed model, incorporating 17 SNPs, 5 core clinical variables and 9 candidate clinical variables, provided the best performance (AUC 0.838; mean cross-validated AUC 0.792 ± 0.048) and a more favorable balance of accuracy, sensitivity and specificity than either model alone. Pairwise DeLong tests showed that the mixed model significantly outperformed both the clinical and SNP models (P < 0.001). Seven SNPs remained independently significant in the mixed model: CYP4A11 rs9333025, NINJ2 rs12425791, PRKCH rs2230500, PTGS1 rs1330344, CYP2C19 rs4986893, VEGFA rs3025039 and MTHFR rs1801133.

Conclusions

Genetic polymorphisms provide modest but statistically significant incremental value to clinical prognostic models for ischemic stroke, yielding a small improvement in discrimination (AUC on the order of 0.01) over clinical predictors alone. While this gain is limited, it supports incorporating genetic information into outcome prediction, pending further validation in larger cohorts and across different stroke subtypes.