Researchers at the National Institutes of Health (NIH), as part of a proof-of-concept study, have developed an artificial intelligence (AI) tool that detects cancer using routine clinical data.

The AI-based simple blood test is designed to predict how a patient’s cancer will respond to immune checkpoint inhibitors, a type of drugs that help immune cells kill cancer cells.

Its machine-learning model helps doctors determine whether immunotherapy drugs can be effective for the treatment of a patient’s disease.

The study, recently published in Nature Cancer, was led by researchers at the National Cancer Institute’s (NCI) Center for Cancer Research and Memorial Sloan Kettering Cancer Center in New York.

NCI is a part of NIH that leads the National Cancer Program and NIH’s efforts to dramatically reduce the prevalence of cancer and improve the lives of people with cancer.

Currently, two predictive biomarkers have been approved by the US Food and Drug Administration (FDA) to identify patients who may be eligible for immune checkpoint inhibitors.

The first predictive biomarker is tumour mutational burden, which is defined as the number of mutations in the DNA of cancer cells.

The second one is Programmed death-ligand 1 (PD-L1), which is a tumour cell protein that limits the immune response and works as a target for some immune checkpoint inhibitors.

However, the biomarkers may not always precisely predict the immune response.

The NIH study deployed a new type of machine-learning model that makes predictions based on five clinical features that are routinely collected from patients.

Its five features include the patient’s age, cancer type, history of systemic therapy, blood albumin level, and blood neutrophil-to-lymphocyte ratio, a marker of inflammation.

The model also considers tumour mutational burden, assessed through sequencing panels.

In the study, the model was built and evaluated using data from multiple data sets that included 2,881 patients treated with immune checkpoint inhibitors across 18 solid tumour types.

The model accurately predicted a patient’s chances of responding to an immune checkpoint inhibitor and how long they would live, both overall and before the disease returned.

NIH researchers said that the machine learning model also identified patients with low tumour mutational burden who could still be treated effectively with immunotherapy.