ASTRO 2019: Predicting Radiation Toxicity in Head/Neck Cancer With Machine-Learning Model
Posted: Thursday, September 19, 2019
Researchers have developed a “machine-learning approach” to predict acute side effects from radiation therapy in patients with head and neck cancer. Based on more than 700 clinical and treatment variables, the model appears to yield clinically valid predictions of feeding tube placement and significant weight loss secondary to radiation therapy. The study findings were presented at the 2019 American Society for Radiation Oncology (ASTRO) Annual Meeting (Abstract 141) and also published in the International Journal of Radiation Oncology • Biology • Physics.
“To our knowledge, this is the first application of a precision oncology approach to predict for these toxicities in a [head and neck cancer] patient cohort undergoing [radiation therapy],” stated Jay P. Reddy, MD, PhD, of The University of Texas MD Anderson Cancer Center in Houston, and colleagues.
Data from 2,121 courses of radiation therapy were extracted from an internal Web-based charting tool; an electronic health record; and a record/verify system to develop predictive models of feeding tube placement, significant weight loss, and unplanned hospitalization. The collected variables included tumor characteristics, prior treatment, and radiation therapy details.
The incidence of feeding tube placement was 23.1%, with an AUC of 0.755, and significant weight loss occurred in 14.2%, with an AUC of 0.751; both of these outcomes met the prespecified threshold of clinical validity. Although the model for unplanned hospitalization did not reach the clinical validity threshold, the researchers noted that this “may change with increasing training data.”
“Further refinement of precision oncology approaches could be transformative by identifying patients who may benefit from early intervention to avert significant weight loss and the need for feeding tube placement due to [radiation therapy] for [head and neck] cancer,” concluded Dr. Reddy and colleagues.
Disclosure: The study authors’ disclosure information may be found at redjournal.org.