NIH-developed AI tool predicts patients’ response to immunotherapy

On June 3, 2024, researchers at the National Institutes of Health published a proof-of-concept study in Nature Cancer, presenting an artificial intelligence (AI) tool that uses machine learning (ML) to assess how an individual cancer patient may respond to immune checkpoint inhibitors. The organization predicts that this tool could be critical in helping providers determine whether to start a patient on immunotherapy drugs.

According to the study, immune checkpoint blockade drugs have revolutionized cancer care; however, one remaining challenge is accurately predicting patient response.

The NIH press release notes that the FDA has only approved two predictive biomarkers for providers to identify immunotherapy candidates: tumor mutational burden and programmed cell death 1 ligand 1 (PD-L1) expression. Tumor mutational burden measures the number of DNA mutations in cancer cells. PD-L1, a protein that interferes with and limits immune response, is also a predictive biomarker for immunotherapy candidates, as many immunotherapy drugs target this protein.

However, the NIH notes that despite their availability, these predictive biomarkers are not always accurate in predicting immunotherapy responses.

The Nature Cancer study analyzed large datasets of patients with 18 types of solid tumors with varying clinical, pathologic, and genomic features. Some patients had been treated with immune checkpoint inhibitors while others had not.

Using the data, the researchers developed an AI tool that provides a clinical score called LORIS, a logistic regression-based immunotherapy-response score. The six-feature logistic regression model looks at multiple factors, including five clinical factors: age, cancer type, history of systemic therapy, blood albumin levels, and blood neutrophil-to-lymphocyte ratio. The model also considers tumor mutational burden, which is assessed with sequencing panels.

Compared to the existing predictive biomarkers, LORIS was better at predicting immunotherapy response and identifying responsive patients irrespective of the tumor mutational burden or PD-L1 expression.

Additionally, researchers in the study note, “LORIS consistently predicts patient objective response and short-term and long-term survival across most cancer types. Moreover, LORIS showcases a near-monotonic relationship with ICB response probability and patient survival, enabling precise patient stratification. As an accurate, interpretable method using a few readily measurable features, LORIS may help improve clinical decision-making in precision medicine to maximize patient benefit.”

While additional studies are needed to validate this tool, the availability of an AI model to predict immunotherapy response may guide providers in prescribing the most effective first-line treatment.