Machine learning model uses clinical and genomic data to predict immunotherapy effectiveness

The forecasting tool assesses multiple patient-specific biological and clinical factors to predict the degree of response to immune checkpoint inhibitors and survival outcomes. It markedly outperforms individual biomarkers or other combinations of variables developed so far, according to findings published in Nature Biotechnology.

With further validation, the tool may help oncologists better identify patients most likely to benefit from ICB. Discerning, prior to treatment, patients for whom ICB would be ineffective could reduce unnecessary expense and exposure to potential side effects. It could also indicate the need to pursue alternate treatment strategies, such as combination therapies.

"It's important to know which treatment modalities patients are most suited for," said Dr. Chan, director of Cleveland Clinic's Center for Immunotherapy & Precision Immuno-Oncology. "Our model provides a more comprehensive understanding of the diversity of responses among patients to immune checkpoint blockade. It's the first to assemble such a large-scale set of clinical and genomic variables that have predictive value for immunotherapy across numerous cancer types."

These latest findings build on earlier work from Dr. Chan, who discovered that patients with high tumor mutation burden and DNA repair deficiencies respond well to immune checkpoint therapy. These findings have been validated by clinical trials and the FDA approved as the first tumor type-agnostic approvals for any cancer therapy.

Immune checkpoints are proteins on specific immune cells (T cells) that when activated, or "turned on," prevent immune responses from being too strong and destroying healthy cells. Some cancer cells are able to hijack checkpoint signaling in order to disguise themselves and avoid being targeted by a patient's immune system. Checkpoint inhibitors are a class of immunotherapy drugs that prevent cancer cells from activating these checkpoints.

However, ICB is not effective in all cancer types. Even in cancers responsive to ICB, half or more of all patients treated with ICB do not derive clinical benefit. Previous research has identified some biomarkers and genomic features associated with ICB efficacy, but no single factor can be considered an optimal predictor of treatment outcomes.

In this study, Dr. Chan and his colleagues developed their model using a dataset containing clinical, tumor and genetic sequencing information from nearly 1,500 patients with 16 different cancer types who were treated with two different types of immune checkpoint inhibitors (specifically PD-1/PD-LI inhibitors and CTLA-4 blockade) or a combination of both. They then applied an algorithm that incorporated many genetic, molecular, clinical and demographic variables, some of which have been shown to be associated with ICB response.

Interestingly, the researchers found that the variable with the greatest influence on ICB response is tumor mutational burden (the frequency of certain mutations within a tumor's genes), followed closely by a patient's chemotherapy history. Levels of three blood markers -- hemoglobin, platelets and albumin -- also had strong predictive value, not only for forecasting patients' overall survival, but also the actual radiographic response to ICB treatment.

"How these variables all work together is really the key here," said Dr. Chan. "This model shows that, rather than a single predictive biomarker, we're headed toward a multifactor nomogram for clinical use."

The team's fully integrated model proved to be highly accurate, significantly outperforming two other forecasting tools, including tumor mutational burden, which the FDA approved in 2020 as a biomarker to predict anti-PD-1 ICB efficacy in solid tumors.

"The model works well, despite what type of cancer is being assessed, which shows that these commonalities are what's important," Dr. Chan explain. "These are primary factors that affect ICB response. The factors may be weighted a little bit differently from cancer to cancer, but it's almost like a common language for response prediction."

Taken together, the positive results support moving forward to test the model in a clinical trial with a large, diverse cohort of cancer patients, which would provide a more accurate assessment of its performance in a real-world setting.

The study was funded in part by the National Cancer Institute (part of the National Institutes of Health).