By Dorothy Hagmajer
Chronic obstructive pulmonary disease, or COPD, is a disease that negatively impacts the quality of life of almost 11.7 million people across the United States. In the past, physicians have struggled to identify the tipping point of when lung function begins to decline — and titrating treatment appropriately.
Now, thanks to researchers in the UF College of Public Health and Health Professions, clinicians may be able to uncover which measurements can predict that initial tipping point of lung function decline.
“I’m very interested in developing learning models that not only help with disease management, but also with developing new therapeutic strategies,” said Takis Benos, Ph.D., the William Bushnell Presidential Chair and professor in the Department of Epidemiology. “In that respect, it has been incredibly rewarding to have uncovered factors that seemed to be very important for either development of future general abnormalities or those specific to one kind of disease.”
Though COPD is commonly associated with behaviors like cigarette smoking and inevitabilities like aging, other contributing factors remain unclear. If these mysterious others can be identified, however, a team of providers has the potential to make improved informed clinical decisions by recognizing at-risk individuals sooner rather than later.
“We wanted to identify clinical, genetic, and radiological features that are not only correlated, but can predict, future abnormal lung function,” Benos explained.
In a study published in PLOS Medicine, Benos and his team analyzed data from two groups and found that the presence of emphysema in non-obstructed individuals is a powerful, direct indicator of future obstruction.
One of the key measurement mechanisms the team used was forced expiration volume, or how much air a person can exhale during a forced breath, a number that provided reliable predictions independently of sex and height. Another factor, the thickness of an airway wall, also indicated future lung decline.
“Now, although we can successfully recognize measurements that predict any incident lung function decline, future research will endeavor to also predict specific trajectories of an individual’s lung function decline,” Benos said.
This research builds upon Benos’ lab’s research, which strives to tackle problems in medicine by applying novel machine learning algorithms to issues of chronic disease progression, cancer, and other diseases. At the end of the day, Benos said, machine learning can help clinicians address timeliness in disease progression — and determine how to augment care if they know someone is at risk of developing a chronic illness.
“Ultimately, it’s up the clinicians to decide whether they need to intervene earlier — like instances where someone has yet to develop the lung abnormality,” Benos pointed out. “But the goal is always to provide more information than we have before, in hopes that patient quality of life can be improved.”