Computational Tool Predicts Immunotherapy Outcomes for Metastatic Breast Cancer Patients

Identifying which patients will benefit from immunotherapy is crucial, given the high toxicity associated with these therapies. Previous research has investigated whether the presence or absence of specific cells or the expression levels of various molecules within tumors can indicate a patient's likelihood of responding to immunotherapy. These molecules, known as predictive biomarkers, play an important role in selecting appropriate treatments for patients. Unfortunately, the accuracy of current predictive biomarkers in determining who will benefit from immunotherapy is limited. Additionally, conducting a large-scale evaluation of the characteristics that predict treatment response typically requires collecting tumor biopsies and blood samples from numerous patients and performing several assays, which presents significant challenges. Researchers have now leveraged computational tools to create a method for assessing which patients with metastatic triple-negative breast cancer may benefit from immunotherapy.

A team of computational scientists from the Johns Hopkins Kimmel Cancer Center (Baltimore, MD, USA) and the Johns Hopkins University School of Medicine (Baltimore, MD, USA) utilized a mathematical model called quantitative systems pharmacology to generate 1,635 virtual patients with metastatic triple-negative breast cancer and conducted treatment simulations using the immunotherapy drug pembrolizumab. They analyzed this data with advanced computational tools, including statistical and machine learning methods, to identify biomarkers that can accurately predict treatment responses. Their focus was on determining which patients would respond positively to treatment and which would not. By utilizing the partially synthetic data generated from the virtual clinical trial, the researchers evaluated the performance of 90 biomarkers both individually and in combinations of two, three, and four.

The findings revealed that pretreatment biomarkers, which are measurements taken from tumor biopsies or blood samples before treatment begins, had limited effectiveness in predicting treatment outcomes. Conversely, on-treatment biomarkers, which are collected after the initiation of treatment, proved to be more predictive of outcomes. Interestingly, the study found that some commonly utilized biomarker measurements, such as the expression of PD-L1 and the presence of lymphocytes within the tumor, performed better when assessed before treatment commenced rather than after it started. The researchers also investigated the accuracy of non-invasive measurements, such as immune cell counts in the blood, in forecasting treatment outcomes. According to their research published in the Proceedings of the National Academy of Sciences, some blood-based biomarkers were found to be comparably effective as tumor- or lymph node-based biomarkers in identifying patients likely to respond to treatment, suggesting a less invasive predictive approach.

Measurements of changes in tumor size, which can be easily obtained through CT scans, also showed potential as predictive indicators. Notably, these measurements taken within two weeks of initiating treatment demonstrated significant potential in identifying who would respond favorably if the treatment continued. To confirm their findings, the investigators conducted a virtual clinical trial selecting patients based on tumor diameter changes at the two-week mark after starting treatment. Remarkably, the simulated response rates more than doubled—from 11% to 25%. This underscores the potential of noninvasive biomarkers as alternatives when collecting tumor biopsy samples is not feasible. Overall, these new insights highlight the possibility of better patient selection for immunotherapy in metastatic breast cancer cases. The researchers anticipate that these findings will aid in designing future clinical studies, with the methodology potentially applicable to other cancer types.

“Predictive biomarkers are critical as we develop optimized strategies for triple-negative breast cancer, so as to avoid overtreatment in patients expected to do well without immunotherapy, and undertreatment in those who do not respond well to immunotherapy,” said study co-author Cesar Santa-Maria, M.D., an associate professor of oncology and breast medical oncologist at the Johns Hopkins Kimmel Cancer Center. “The complexities of the tumor microenvironment make biomarker discovery in the clinic challenging, but technologies leveraging in-silico [computer-based] modeling have the potential to capture such complexities and aid in patient selection for therapy.”