– Researchers at the University of Texas MD Anderson Cancer Center have developed a machine learning tool that can accurately predict two of the most challenging side effects of radiation therapy for patients with head and neck cancers, including significant weight loss or the need for a feeding tube.
According to the research team, approximately 53,000 people are diagnosed with head and neck cancer each year in the US. These cancers are more than twice as common in men as in women and are typically diagnosed later in life. While radiation is an effective form of treatment for these cancers, it can also damage the oral tissue and upset the balance of bacteria in the mouth, leading to sore throats, mouth sores, and loss of taste.
As a result, patients with head and neck cancers who receive radiation may experience significant weight loss or require the temporary insertion of a feeding tube.
To build the model, the team examined more than 700 clinical and treatment variables for patients with head and neck cancer who received more than 2,000 courses of radiation therapy across five practice sites.
Researchers used the model to predict three endpoints, including significant weight loss, feeding tube placement, and . The machine learning tool was able to predict weight loss accurately, with an area under the curve (AUC) of 0.751, and the need for feeding tube placement with an AUC of 0.755.
“The models used in this study were consistently good at predicting those two outcomes,” said Jay Reddy, MD, PhD, an assistant professor of radiation oncology at The University of Texas MD Anderson Cancer Center and lead author on the study. “You could rerun those models with a new patient or series of patients and get a number saying this adverse effect is likely to happen or not to happen.”
However, researchers also found that the model fell short of predicting unplanned hospitalizations with sufficient accuracy, achieving an AUC of 0.64. Performing the analyses again with more training data for unplanned hospitalizations could improve this, researchers said.
“As we treat more and more patients, the sample size gets bigger, so every data point should get better. It’s possible we just didn’t have enough information accumulated for this aspect of the model,” said Reddy.
This new approach could help identify patients who may benefit from interventions that prevent significant weight loss after treatment or reduce the need for a feeding tube.
“In the past, it has been hard to predict which patients might experience these side effects,” said Reddy. “Now we have a reliable machine learning model, using a high volume of internal institutional data, that allows us to do so.”
Although this machine learning approach can’t isolate the single-most predictive factor or combination of factors that result in negative side effects, it can provide physicians and patients with a better understanding of what to expect during treatment.
“Being able to identify which patients are at greatest risk would allow radiation oncologists to take steps to prevent or mitigate these possible side effects. If the patient has an intermediate risk, and they might get through treatment without needing a feeding tube, we could take precautions such as setting them up with a nutritionist and providing them with nutritional supplements,” said Reddy.
“If we know their risk for feeding tube placement is extremely high – a better than 50 percent chance they would need one – we could place it ahead of time so they wouldn’t have to be admitted to the hospital after treatment. We’d know to keep a closer eye on that patient.”
Researchers noted that going forward, machine learning tools could help predict which treatment plans are best for each patient, leading to more personalized care and better outcomes.
“Machine learning can make doctors more efficient and treatment safer by reducing the risk of error,” said Reddy. “It has the potential for influencing all aspects of radiation oncology today – anything where a computer can look at data and recognize a pattern.”
Credit: Google News
