November 06, 2022
2 min read
Yasin H, et al. Abstract 16. Presented at: International Kidney Cancer Symposium: North America; Nov. 4-5, 2022; Austin, Texas.
The authors report no relevant financial disclosures.
Artificial intelligence models can help predict cardiotoxicity risk among patients with renal cell carcinoma treated with VEGF receptor inhibitors, according to study results.
Integration of artificial intelligence (AI) models into electronic medical records can help oncologists and other members of the clinical care team identify those who may benefit from cardio-oncology monitoring and treatment, findings presented at the International Kidney Cancer Symposium: North America showed.
“Further studies comparing differences in outcomes between high-risk … patients who were referred to cardio-oncology versus patients who were not referred are warranted,” Hesham Yasin, MD, clinical fellow at Vanderbilt University Medical Center, and colleagues wrote.
Tyrosine kinase inhibitors that target VEGF receptors are standard components of renal cell carcinoma treatment. These agents generally are effective and safe, but they can cause cardiotoxicity risk for an estimated 3% to 8% of patients, according to study background.
Researchers hypothesized that AI and machine learning may help predict which patients may be at elevated risk for cardiotoxicity, thereby allowing for timely referral for cardio-oncology monitoring or treatment.
Yasin and colleagues used the Vanderbilt University Medical Center EMR to obtain de-identified data related to 2,047 patients with renal cell carcinoma who received any of 10 VEGF receptor TKIs.
Investigators applied random forest and artificial neural network machine learning algorithms to analyze the cohort, categorizing patients into four risk groups — potential, low, moderate or major — based on cardiotoxicity risk factors.
The potential-risk group included any patient treated with VEGF inhibitors.
The low-risk group included those with one or more of the following risk factors: hyperlipidemia, high-density lipoprotein (HDL) cholesterol between 41 and 59, or low-density lipoprotein (LDL) cholesterol between 160 and 189.
The moderate-risk group included those with up to two of the following risk factors: essential hypertension, HDL of 40 or less, diabetes mellitus, age older than 65 years, LDL of at least 190 and/or xanthoma, BMI greater than 35 kg /m2smoking, left ventricular ejection fraction (LVEF) 51% to 54%, blood pressure of 140/90 or higher, N-terminal prohormone of brain natriuretic peptide level of 400 or greater, or brain natriuretic peptide level greater than 100.
The major-risk group included those with three or more of the moderate risk factors, or one or more of the following: radiation, systolic heart failure, ischemic cardiomyopathy or other cardiomyopathy, coronary artery disease, diastolic heart failure, severe aortic stenosis, severe Mitral regurgitation, atrial fibrillation, LVEF of 50% or less, severe pulmonary hypertension, troponin greater than 0.02 or HbA1C greater than 9.
Yasin and colleagues divided patients into a training set (80%) and validation set (20%).
Limited validation analyzes showed 58% of patients who exhibited major risk for cardiotoxicity did not receive referrals to cardio-oncology specialists.
“A pilot project is underway to integrate model predictions in Epic workflow as a report of patients who should be referred to cardio-oncology for monitoring and/or tailored treatments,” Yasin and colleagues wrote.