Testicular cancer has a high survival rate, but the chemotherapy typically used during treatment can induce nephrotoxicity. Researchers out of Denmark have developed a new AI algorithm that identifies patients who may be vulnerable to nephrotoxicity, sharing their findings in JNCI Cancer Spectrum.
“Maintenance of sufficient renal function during treatment with chemotherapy is vital, and identification of patients at risk for developing nephrotoxicity could influence the treatment of choice if alternatives exist,” wrote Sara L. Garcia, MSc, Technical University of Denmark, and colleagues. “Additionally, impaired renal function has been associated with increased risk of cardiovascular disease, which may pose a problem in long-term cancer survivors.”
Garcia et al. explored data from 433 patients, using random forest classification algorithms to calculate each patient’s risk of developing nephrotoxicity. The baseline model performed well, but its ability to predict outcomes “substantially improved” when researchers added key genomic information.
Overall, when categorizing each patient by the severity of their risk, the algorithm correctly moved 67% of the patients who developed nephrotoxicity into the high-risk group. In addition, 92% of patients who did not develop nephrotoxicity were correctly moved into the low-risk group.
“The ability to develop machine learning models for patient stratification in different nephrotoxicity risk groups has the potential to balance aggressive treatment against predicted toxicity risk,” the authors wrote.
In addition, the researchers added, the AI model they developed may also prove valuable for the classification of patients undergoing chemotherapy-based treatment for ovarian, bladder and lung cancer.
The full analysis from Garcia and colleagues is available here.