Researchers at Northwestern Medicine have developed a blood test that could transform how doctors predict which head and neck cancer patients will actually benefit from immunotherapy—potentially sparing thousands from months of ineffective treatment and its accompanying side effects. The study, published in the Journal of Clinical Investigation, analyzes patterns in cell-free DNA circulating in the bloodstream to predict treatment response with striking accuracy, offering hope to patients facing one of cancer medicine's most stubborn puzzles.
Head and neck squamous cell carcinoma affects hundreds of thousands of people worldwide each year. While immunotherapy has emerged as one of the most promising cancer treatments in the past decade, the reality remains sobering: only about one in five patients actually respond to these drugs. The rest endure months of treatment, managing side effects and uncertainty, with doctors currently unable to predict who will benefit before starting therapy.
Yaping Liu, Ph.D., an assistant professor of Biochemistry and Molecular Genetics at Northwestern Medicine's Ken and Ruth Davee Department of Neurology, led the team in developing a new approach that sidesteps the limitations of existing tests. Rather than focusing solely on tumor mutations, the blood test examines patterns in how DNA is fragmented across the entire genome. This distinction matters enormously because cell-free DNA in the bloodstream comes not just from dying cancer cells, but also from immune cells—capturing signals from both sides of the cancer-immune interaction that determines whether immunotherapy will work.
"Cell-free DNA comes from both tumor cells and immune cells," Liu explained. "Since immunotherapy depends on the interaction between those two systems, we were able to capture signals from both."
The team analyzed 185 blood samples collected over time from 68 patients enrolled in a multi-institutional phase II clinical trial of pembrolizumab, an immune checkpoint inhibitor. By tracking changes in cfDNA patterns before and after surgery throughout the course of therapy, researchers could develop a scoring system that reliably distinguished responders from non-responders. The results were striking: when the data was visualized after correcting for technical factors, responders and non-responders separated naturally into distinct groups—even without machine learning models.
When incorporated into predictive algorithms, the scoring system demonstrated high accuracy across multiple settings. Patients classified as likely responders based on the blood test achieved significantly better outcomes, including improved disease-free survival compared to those predicted to be non-responders. The new biomarker consistently outperformed existing approaches.
Liu emphasized that validation in larger, independent clinical trials will be essential before the test can be widely adopted in clinical practice. The current study, while including multiple centers, involved a relatively small sample size for such groundbreaking work. "The next step is to validate these results in independent clinical trials," he said.
Yet the implications extend far beyond head and neck cancer. Liu's team is already exploring whether the cfDNA pattern-analysis approach could predict responses in other cancers or even non-cancer diseases where tissue-immune interactions play a central role. If validated, the blood test could offer patients a minimally invasive way to make smarter treatment decisions early, transforming outcomes for millions facing immunotherapy decisions.