Scientists at the US National Institutes of Health (NIH) have introduced an AI tool called Logistic Regression-Based Immunotherapy-Response Score (LORIS). The tool can predict the response to immune checkpoint inhibitors for cancer patients.

According to the findings of a proof-of-concept study, the AI model uses routine clinical data to assist clinicians in determining the suitability of immunotherapy drugs for individual cancer treatment.

Immune checkpoint inhibitors are a kind of immunotherapy medicines that aid immune cells to kill cancer cells.

The proof-of-concept study was spearheaded by the Center for Cancer Research at the National Cancer Institute, an NIH unit and the Memorial Sloan Kettering Cancer Center.

The US Food and Drug Administration (FDA) currently recognises two predictive biomarkers for identifying potential candidates for immune checkpoint inhibitor therapy: tumour mutational burden and PD-L1 protein levels. But these markers do not consistently forecast treatment responses.

Unlike previous models that rely on costly and infrequently collected molecular sequencing data, LORIS uses five standard clinical features: age, type of cancer, prior systemic therapy, blood albumin level and blood neutrophil-to-lymphocyte ratio. It also incorporates tumour mutational burden data obtained through sequencing panels.

The model’s efficacy was validated using data from 2,881 patients with 18 different types of solid tumours, all treated with immune checkpoint inhibitors.

LORIS demonstrated accuracy in predicting both the likelihood of response to treatment and patient survival rates, highlighting its potential to identify patients with low tumour mutational burden who may still benefit from immunotherapy.

Despite the promising results, the team emphasised the need for larger prospective studies to fully assess the AI model’s clinical applicability.