Predictive nomogram for ischemic stroke risk in clear cell renal cell carcinoma patients

Jie Wen, Yi Rong, Yinbo Kang, Dingyang Lv, Fan Cui, Huiyu Zhou, Mohan Jia, Qiwei Wang, Weibing ShuangScientific 2024 · DOI

TL;DR

Researchers have developed a predictive tool, known as a nomogram, to identify clear cell renal cell carcinoma (ccRCC) patients at high risk for ischemic stroke. By combining traditional risk factors like hypertension and age with tumor-specific data like the WHO/ISUP grade, the model provides a personalized risk score that can help clinicians intervene early and improve patient outcomes.

What Problem Does This Paper Address?

Clear cell renal cell carcinoma (ccRCC) is the most common form of kidney cancer, and patients diagnosed with it face a significantly higher risk of ischemic stroke—a condition where blood flow to the brain is obstructed. While the link between cancer and stroke is well-documented (affecting roughly 14.6% of cancer patients), the specific mechanisms connecting kidney tumors to cerebrovascular events remain elusive.

Existing stroke prediction models are generally designed for the broader population and often fail to account for the unique biological stressors introduced by a malignancy. This paper addresses the lack of a specialized, clinically applicable tool to screen ccRCC patients for stroke risk within one year of their cancer diagnosis.

Methods

The researchers conducted a retrospective analysis of 915 ccRCC patients who underwent surgery. To build a robust predictive model, they employed a multi-step statistical pipeline:

  1. Univariate Logistic Regression: Used for initial screening of potential risk factors.
  2. LASSO Regression: The "Least Absolute Shrinkage and Selection Operator" was used to refine the variable list. This technique is excellent for preventing "overfitting" by penalizing less important variables, ensuring the final model is both simple and accurate.
  3. Multivariate Logistic Regression: The final model was constructed using the most impactful predictors.
  4. Nomogram & Shiny App: The results were translated into a visual nomogram (a graphical calculating device) and a "Shiny" web-based calculator for real-world clinical use.

Datasets

The study utilized a dataset from the First Hospital of Shanxi Medical University, covering patients treated between January 2013 and May 2022.

  • Total Cohort: 915 patients.
  • Training Cohort (70%): 640 patients used to build the model.
  • Validation Cohort (30%): 275 patients used to test the model's accuracy on "unseen" data. The dataset included a wide range of variables, from demographic data (age, BMI) and lifestyle factors (smoking) to laboratory markers (D-dimer, lipids) and tumor characteristics (size, stage, and grade).

Key Insights and Ideas

  • Tumor Grade vs. Size: One of the most striking findings was that while tumor size and stage did not significantly correlate with stroke risk, the WHO/ISUP grade (a measure of how "abnormal" the cancer cells look) did. This suggests that the biological aggressiveness of the tumor—perhaps through the secretion of pro-coagulant substances—is a bigger driver of stroke than the physical mass of the tumor itself.
  • The "Perfect Storm" of Risk: The model identifies a synergy between traditional cardiovascular risks (hypertension, diabetes, high LDL-C) and cancer-specific markers (D-dimer and tumor grade).
  • Logarithmic Transformation: The authors used lgD-dimer (log-transformed D-dimer) to handle the extreme variance in coagulation data, a necessary statistical step to ensure the model wasn't skewed by outliers.

Key Results

The final model identified six independent predictors: WHO/ISUP grade, diabetes, hypertension, LDL-C, age, and D-dimer. * Discrimination: The model achieved an Area Under the ROC Curve (AUC) of 0.816 in the training cohort and 0.775 in the validation cohort. In predictive modeling, an AUC above 0.7 is generally considered good, and above 0.8 is excellent. * Clinical Threshold: The researchers identified an optimal cutoff value of 53.7%. Patients with a predicted probability above this threshold are considered "high-risk." * Calibration: Calibration curves showed a high degree of agreement between the model’s predicted risks and the actual observed stroke events, indicating the model is reliable for clinical decision-making.

Limitations and Future Work

The authors acknowledge several hurdles:

  • Retrospective Nature: Because the study looked at past records, it is susceptible to selection bias and potential inaccuracies in patient-reported data (like smoking history).
  • Single-Center Study: The data comes from one hospital in China. To ensure the model works for diverse global populations, multi-center prospective studies are required.
  • Mechanistic Mystery: While the model proves a correlation between tumor grade and stroke, it doesn't explain the why. Future research is needed to explore whether this is due to tumor-induced hypercoagulability, tumor embolism, or other inflammatory pathways.

Summary updated

Abstract

Clear cell renal cell carcinoma (ccRCC) and ischemic stroke are critical global health challenges with a notable association. This study explores the correlation between tumor-related factors and ischemic stroke risk, aiming to construct a predictive nomogram model for ischemic stroke in ccRCC patients. We retrospectively analyzed data from ccRCC patients who underwent nephrectomy at the First Hospital of Shanxi Medical University between January 1, 2013, and May 31, 2022. The data were randomly divided into a training cohort (70%) and a validation cohort (30%). Predictive factors were identified using univariate logistic regression, least absolute shrinkage and selection operator regression, and multivariate logistic regression. A nomogram and a Shiny local calculator were developed using these predictors. We identified six predictors for the nomogram: WHO/ISUP grade, diabetes, hypertension, LDL-C, age, and D-dimer. The nomogram showed good discrimination, with an area under the ROC curve of 0.816 in the training cohort and 0.775 in the validation cohort. The optimal cutoff value was 53.7%. The model demonstrated excellent calibration and clinical applicability. WHO/ISUP grade correlates with ischemic stroke risk, offering insights into cancer-related ischemic stroke mechanisms. This nomogram aids in identifying high-risk individuals among ccRCC patients, facilitating early management and improved outcomes.
BibTeX
@inproceedings{wen2024predictive,
  author = {Jie Wen and Yi Rong and Yinbo Kang and Dingyang Lv and Fan Cui and Huiyu Zhou and Mohan Jia and Qiwei Wang and Weibing Shuang},
  title = {{Predictive nomogram for ischemic stroke risk in clear cell renal cell carcinoma patients}},
  booktitle = {Scientific},
  year = {2024},
  doi = {10.1038/s41598-024-82072-9},
}
Figure