August 5, 2025
Please read for 3 minutes
Important takeouts:
- Familial hypercholesterolemia has been diagnosed and has not been treated.
- A new machine learning algorithm has identified individuals with the most likely chance of familial hypercholesterolemia.
BOSTON – New research shows that new machine learning algorithms may improve detection and treatment of individuals with familial hypercholesterolemia.
Spencer V. Carter, MD, Assistant Professor of Clinical Medicine in the Department of Cardiology at UT Southwestern Medical Center presented data at the American Prevention Association Conference on CVD Prevention.
Identify the possibility of familial hypercholesterolemia
“Family hypercholesterolemia significantly increases the lifetime risk of cardiovascular disease, and if not treated, many of these events occur before the age of 50. Early diagnosis of family hypercholesterolemia reveals that important lipid-lowering therapies can be achieved to improve clinical outcomes,” Carter said. “However, even with that knowledge, familial hypercholesterolemia remains undiagnosed and untreated.”
Carter and colleagues evaluated the performance of Find, Identify, Network, and Delivery FH (Find-FH). It was first developed by the Family Heart Foundation. The algorithm utilizes over 60,000 input capabilities to generate numeric FIND-FH scores to identify individuals with the most likely chance of familial hypercholesterolemia, Carter said.
The researchers applied the algorithm to a dataset of over 307,000 patients at UT Southwestern Medical Center. The algorithm identified 340 adults who were considered the most likely to have familial hypercholesterolemia based on multiple criteria. The researchers then conducted manual chart reviews of all patients to assess the accuracy of familial hypercholesterolemia diagnosis by Simon Bloom and the Dutch lipid clinical network (DLCN) criteria.
This study was published in Journal of Clinical Lipidology.
“Radiologous aid” to traditional techniques
Of the 340 patients, the average age was 50 years old, 58% were male, 18% were black, and 55% were white.
Most patients (n = 231) had documented the highest LDL cholesterol levels below 190 mg/dl. The prevalence of atherosclerotic CVD was low (16%). Most people were on statin therapy.
“These properties suggest that this is a group of individuals that may be overlooked in traditional familial hypercholesterolemia screening techniques, and that this algorithm may be a reasonable adjunct to those traditional techniques,” Carter said.
When researchers were divided into quintiles based on numerical FIND-FH scores assigned from the algorithm, they found that the likelihood of familial hypercholesterolemia differed by FIND-FH scores.
“20 [percent] For 32% of individuals identified by this algorithm, at least Simon-Blueume or DLCN criteria allowed for familial hypercholesterolemia. This too is based solely on information found through manual chart reviews of electronic health records,” Carter said.
When they looked at it by DLCN standards, the results were “a little more dramatic,” Carter said. “For those with the highest FIND-FH score, over half… at least familial hypercholesterolemia is possible. 18% have solid familial hypercholesterolemia by DLCN standards,” Carter said.
The researchers then assessed the proportion of individuals identified by machine learning algorithms with sufficient clinical odds for familial hypercholesterolemia, ensuring outreach for further care. Outreach criteria include familial hypercholesterolemia according to clinical criteria. LDL levels above 190 mg/dl. LDL levels above 160 mg/dL on ASCVD. Suboptimal statin response; LDL greater than 130 mg/dL during statin therapy.
“Overall, 56% of our team thought there was adequate clinical suspicion for the familial hypercholesterolemia that our team wanted to reach out to them,” Carter said.
Researchers found that 53% of individuals with high clinical suspicion for familial hypercholesterolemia had LDL levels of 190 mg/dl, which would have been overlooked by LDL-based screening methods.
According to Carter, there is ongoing research into how that outreach is progressing.
“Together, we hope that these data will help guide the application of the FIND-FH algorithm, which will improve the identification and treatment of individuals suspected of familial hypercholesterolemia,” Carter said.
reference:
Source/Disclosure
Collapse
Carter S. Early career research presentation. Presentation: American Prevention Association Conference on CVD Prevention. August 1st-3rd, 2025. Boston.
Disclosure: Carter reports that there are no related financial disclosures.
