Overview
In advanced non-small cell lung cancer, patients whose tumors have high levels of PD-L1 are typically treated with either immunotherapy alone or a combination of immunotherapy and chemotherapy. Without clear evidence comparing these two strategies directly, it remains difficult for doctors to know which patients may benefit from the addition of chemotherapy.
Using data from more than 1,400 patients in the Flatiron Health Research Database, researchers applied a machine learning model to categorize patients by their risk level at the start of treatment. The model analyzed clinical features including laboratory values, other health conditions, and cancer characteristics to predict survival outcomes. The study revealed that approximately 31% of patients—those facing the most challenging initial prognosis—received a significant survival boost from the addition of chemotherapy. For the remaining 69% of patients, immunotherapy-only treatment was found to be just as effective as the combination.
Why this matters
This research shows that machine learning can uncover specific patient subgroups that may be overlooked using traditional clinical assessment alone. By using a tool to risk-stratify patients, clinicians can ensure that those at high risk receive the full combination therapy they need, while protecting lower-risk patients from the toxicities of unnecessary chemotherapy. These insights move the field closer to a "risk-adapted" treatment strategy, where every lung cancer patient receives the optimal amount of therapy for their unique situation.