Development and validation of a risk assessment tool for ischemic stroke in cancer patients

Figures

• 

  Fig. 1. Results of ROC curve analysis for each model in the development cohort. (A) Model 1 included four parameters (age ≥ 75 years, hypertension, atrial fibrillation, and stage IV cancer). (B) Model 2 included six parameters (the parameters in model 1 plus cancer site [lung, kidney, or pancreas] and adenocarcinoma). (C) Model 3 included the parameters in model 1 plus an NLR of ≥ 4.2 and a D-dimer level of ≥ 1.52 µg/ml. (D) Model 4 included the parameters in model 1 plus cancer site (lung, kidney, or pancreas), adenocarcinoma, an NLR of ≥ 4.28, and a D-dimer level of ≥ 1.52 µg/ml. (E) After stepwise logistic regression analysis, model 5 included atrial fibrillation and a D-dimer level of ≥ 1.52 µg/ml. In these all models, the total risk score was calculated by adding 1 point for each factor present. AUC, area under the curve; NLR, neutrophil-to-lymphocyte ratio; ROC, receiver-operating characteristic.

• 

  Fig. 2. Comparison of the AHANDS and Khorana score in the development cohort. The performance of the AHANDS score (red) was significantly superior to that of the Khorana score (blue) in the development cohort (c-statistic 0.703 vs. 0.54; p for difference < 0.0001).

• 

  Fig. 3. Results of ROC curve analysis in the validation cohort. (A) The c-statistic for the AHANDS score (red) was 0.75. (B) Model 1 showed moderate discrimination (c-statistic 0.70).

• 

  Fig. 4. Performance of the AHANDS score in the development and validation cohorts. Risk was categorized as low (0–2 points), moderate (3–4 points), or high (5–6 points). A higher score indicates a higher risk of developing ischemic stroke in cancer patients.