The technology, developed through a partnership with the Mayo Clinic, gained FDA's breakthrough device designation in December. Its regulatory review was further expedited in light of its applications during the pandemic, the company said.
In an EUA letter issued Monday, FDA framed the algorithm as a tool to help identify more patients who may be at increased risk of bad outcomes with the coronavirus due to cardiac comorbidities. The algorithm is designed to identify such patients by using electrocardiogram data to show if they have a left ventricular ejection fraction (LVEF) of 40% or less, indicating that they may have heart failure or cardiomyopathy.
Analyses of the comorbidities of people who died from COVID-19 have identified cardiac dysfunction as a risk factor. The analyses suggest there is value in identifying patients who also have cardiac conditions, leading the American College of Cardiology to recommend echocardiograms for people infected with the coronavirus who have certain comorbidities.
However, FDA sees gaps in the screening program, stating in its EUA for the Eko algorithm that echo scans are “not indicated for all patients” and “may not be an immediately available option.” As such, the agency contends there is scope to improve the screening process.
FDA awarded an EUA for the Eko ECG Low Ejection Fraction Tool, known as ELEFT for short. ELEFT is a machine learning algorithm trained on electronic health records containing ECG and echo scans taken from the same patients. By comparing the linked tests, the algorithm learnt to use ECGs to make predictions about the findings of echo scans.
Eko validated the algorithm in retrospective and prospective tests. The retrospective study looked at a dataset featuring almost 53,000 patients who underwent a 12-lead ECG and echocardiogram within two weeks of each other, revealing the sensitivity and specificity of the algorithm to be 83.3% and 84.9%, respectively. In the dataset, in which 11.1% of people had a LVEF of 40% or lower, the positive and negative predictive values were 40.8% and 97.6%, respectively, suggesting the algorithm rarely delivers false positives that could cause the screen to fail to detect high-risk patients.
Based on the data, FDA said “it is reasonable to believe that the ELEFT may be effective” in screening for cardiac complications. As there are no available alternatives, the belief was strong enough to support an EUA.
The EUA comes with some caveats. FDA wants physicians to only use the ELEFT algorithm “when echocardiography is not yet available or is not indicated.” In situations when ELEFT is suitable for use, physicians will upload 10 seconds of data from a 12-lead ECG. The algorithm will then analyze the data “within seconds” and makes a yes/no prediction of the likelihood that the patient has a LVEF of less than or equal to 40%. The result is displayed on a smartphone, tablet or PC.
Reflecting the high rate of false positives seen in the retrospective study, FDA wants physicians to run further tests to confirm a patient has a LVEF of 40% or less before starting treatment for the possible cardiac condition.
Eko CEO Connor Landgraf said in a statement the technology will be available at Mayo Clinic immediately, with plans to expand the technology to other providers.