AI Tool Predicts Cancer Patients’ Response to Immunotherapy

By LabMedica International staff writers
Posted on 05 Jun 2024

Image: The AI tool predicts whether someone’s cancer will respond to immune checkpoint inhibitors (Photo courtesy of National Cancer Institute)

Immune checkpoint inhibitors are a form of immunotherapy drug that enables immune cells to target and destroy cancer cells. At present, the Food and Drug Administration has approved two predictive biomarkers for identifying patients who might benefit from immune checkpoint inhibitors. The first biomarker is tumor mutational burden, which measures the number of mutations in the DNA of cancer cells. The second biomarker is PD-L1, a protein found on tumor cells that inhibits the immune response and is targeted by some immune checkpoint inhibitors. However, these biomarkers are not always reliable in predicting a patient's response to immune checkpoint inhibitors. Recent machine-learning models utilizing molecular sequencing data have demonstrated potential in predicting responses, but this data is costly and not routinely collected. Researchers have now created an artificial intelligence (AI) tool that uses standard clinical data, such as results from a basic blood test, to predict if a patient’s cancer will respond to immune checkpoint inhibitors.

The machine-learning model, named Logistic Regression-Based Immunotherapy-Response Score (LORIS), was developed by scientists at the National Cancer Institute (Bethesda, MD, USA). It aims to assist doctors in determining the efficacy of immunotherapy drugs for a patient's cancer treatment. The AI model bases its predictions on five clinical features routinely collected from patients: age, cancer type, history of systemic therapy, blood albumin level, and blood neutrophil-to-lymphocyte ratio, an indicator of inflammation. The model also considers tumor mutational burden, evaluated through sequencing panels.

This model was built and validated using data from multiple independent datasets comprising 2,881 patients treated with immune checkpoint inhibitors across 18 types of solid tumors. The model accurately predicted both a patient’s likelihood of responding to an immune checkpoint inhibitor and their overall survival time, including the period before disease recurrence. Remarkably, the model also identified patients with low tumor mutational burden who could still benefit from immunotherapy. The findings of the study were published in Nature Cancer on June 3, 2024. The researchers emphasized the need for larger prospective studies to further validate the AI model in clinical settings and have made it publicly accessible.

Related Links:
National Cancer Institute
LORIS