Fact checked byRichard Smith

Read more

September 18, 2023
4 min read
Save

AI algorithms show moderate to severe aortic stenosis may portend poor outcomes

Fact checked byRichard Smith
You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

Key takeaways:

  • Patients with moderate to severe aortic stenosis appear to have outcomes just as poor as those with severe aortic stenosis.
  • Low aortic valve area may be a sign that aortic stenosis is truly severe.

Patients with moderate to severe aortic stenosis had similar outcomes to patients with severe aortic stenosis, whereas those with moderate aortic stenosis were comparable to those with mild aortic stenosis, researchers reported.

The researchers conducted the KP-VALVE project using artificial intelligence algorithms to determine the association between physician-assessed aortic stenosis severity, quantitative parameters and clinical outcomes.

Graphical depiction of source quote presented in the article

AI algorithms to read echo reports

Aortic stenosis is the most common valvular heart condition in the world, and the treatment for severe aortic stenosis has been revolutionized with the advent of transcatheter aortic valve replacement,” Matthew D. Solomon, MD, PhD, physician-researcher with the Kaiser Permanente Northern California Division of Research in Oakland, California, told Healio. “Now, valve replacement can be offered to many more patients, and at much lower risk. This advance opened up a new question: Could TAVR help patients with stages of disease that precede severe, symptomatic aortic stenosis, the current indication for valve replacement? Several recent studies have had conflicting results, with some indicating that moderate aortic stenosis was as bad as severe aortic stenosis, and others suggesting it was not. We realized our data would allow us to shed light on this question, and provide granular data across the wide category of moderate aortic stenosis. There also has been greater recognition of unusual aortic stenosis phenotypes, such as low-flow low-gradient aortic stenosis, which is more difficult to identify by echocardiogram, so there were many areas we wanted to investigate in the valvular heart disease space.

“To do this, we built AI algorithms that could read echo reports and create discrete, standardized variables from our 1 million echocardiograms over the past 10 years,” he said. “This was critical, foundational work, because echo reports are usually free-text, and not easily searchable for research and evaluation. Then, we evaluated our aortic stenosis grading and long-term outcomes. The ultimate goal is to build a comprehensive population management program for our aortic stenosis patients, to continuously improve our care for this group. This research study is the first step toward achieving that goal.”

The researchers evaluated 546,769 patients with no previous aortic valve repair or replacement who had echocardiograms from 2008 to 2018, of whom 49,604 had aortic stenosis (mean age, 77 years; 48% women; 17.6% with moderate aortic stenosis; 3.6% with moderate to severe aortic stenosis; 9.4% with severe aortic stenosis). Median follow-up was 3.7 years.

Quantitative echocardiographic data were consistent with physician-assessed aortic stenosis severity categories, with mean aortic valve gradient, aortic valve max velocity and aortic valve area all meeting criteria for the aortic stenosis severity categories, according to the researchers.

Aortic stenosis severity and outcomes

Greater aortic stenosis severity correlated with worse outcomes for death, CV-related hospitalization and syncope, whereas for death and CV-related hospitalization, those with moderate aortic stenosis tracked more closely with those with mild aortic stenosis or mild to moderate aortic stenosis, whereas those with moderate to severe aortic stenosis tracked more closely with those with severe aortic stenosis, Solomon and colleagues found.

The number of quantitative severe aortic stenosis parameters predicted outcomes (HR for death for one severe aortic stenosis parameter = 1.4; 95% CI, 1.34-1.46; HR for death for two severe AS parameters = 1.7; 95% CI, 1.56-1.85; HR for death for three severe aortic stenosis parameters = 1.78; 95% CI, 1.63-1.94), the researchers wrote.

The severe aortic stenosis parameter most associated with poor outcomes was aortic valve area less than 1 cm2, which had the largest relative contribution (67%) and occurred in 21% of patients with moderate aortic stenosis and 56% of patients with moderate to severe aortic stenosis, according to the researchers.

“Our study showed that many cardiologists graded aortic stenosis as moderate or moderate-severe when only one severe aortic stenosis parameter was present on the echocardiogram, particularly if that parameter was an aortic valve area (AVA) less than 1 cm2,” Solomon told Healio. “Specifically, 56% of echocardiograms graded as moderate-severe aortic stenosis had AVA less than 1 cm2, but with other quantitative parameters that did not meet severe aortic stenosis criteria. This was not too surprising, since cardiologists have been trained historically to make sure all the aortic stenosis parameters ‘match up’ — meaning that if you are going to label aortic stenosis as ‘severe,’ all three quantitative parameters must line up. But they often don’t — and it is usually AVA that is low when the other parameters are not. Cardiologists often attribute the discrepancy to a potential error in calculation, since AVA is a derived measure based on three parameters. ... But our results were completely surprising. We found AVA to have the strongest association with adverse outcomes, and by a significant margin. This finding tells cardiologists that we should raise our antennae when we see these discordant values, as that low AVA may in fact be real and be the tipoff that the aortic stenosis is truly severe.”

Solomon said the findings show that “better identification of nonclassical severe aortic stenosis phenotypes could improve outcomes. At a minimum, it should alert health systems and heart teams to the need to implement comprehensive population management for aortic stenosis patients and to provide support to physicians to help them identify these higher-risk patients so that they can be reviewed and potentially treated by expert valvular heart teams at a TAVR center. These data show the need to implement comprehensive population management for aortic stenosis patients. As we move farther into the big data and AI-fueled age, there is a great opportunity to improve care for our patients by using technology to support our practice, and help provide the best care for our patients.”