Health News
Cardiologs is using Holter recordings instead of ECGs to catch more cases of AFib – MedCity News

Cardiologs is using Holter recordings instead of ECGs to catch more cases of AFib – MedCity News

heart, doctor, cardiac

Cardiologs, a company that uses ECG data to detect potential cardiac arrhythmias, is developing a new way to predict atrial fibrillation.

On Tuesday, the European Heart Journal – Digital Health published a study that tested Cardiologs’ new algorithm. The study, which was sponsored by the Paris-based company, found that the model can accurately predict the near-term presence or absence of AFib using only the first 24 hours of a Holter device recording.

Founded in 2014, Cardiologs offers a device-agnostic software platform for cardiac arrhythmia detection. The platform is FDA-approved to screen for AFib and other arrhythmias as an aid to physicians, not a standalone device. 

The company develops its software using cardiologist-uploaded data from a variety of devices, including Holter monitors, smartwatches and ECG patches. Last November, Philips announced that it was expanding its portfolio of cardiac solutions for hospital and ambulatory settings by acquiring Cardiologs.

Dr. Jagmeet Singh, a cardiologist at Massachusetts General Hospital, led the company’s newest study. He and his team collected and de-identified Holter recordings from six independent diagnostic testing facilities in the U.S., European Union, India, South Africa and United Kingdom. They identified a training set of recordings, each lasting 7 to 15 days, in which no AFib could be detected in the first 24 hours.

Using the first 24 hours of these recordings, the research team trained their algorithm to predict the presence or absence of AFib in the 15 following days. Using an external dataset not used during its development, they tested the algorithm and found it could predict whether AFib would occur in the near future with an area under the receiver operating curve, sensitivity and specificity of 79.4%, 76% and 69%, respectively. The study also found that the model outperformed those that used 12-lead ECGs to predict near-term AFib.

The algorithm gives “hope to high-risk patients who would benefit from proactive treatment and AFib mitigation strategies,” Dr. Singh said in a news release.

Cardiologs developed the tool to help physicians identify more cases of AFib — a condition that is often undetected and untreated, despite CDC data showing that the condition contributes to about 158,000 deaths each year. Patients are usually required to undergo a 24-hour ECG to be diagnosed, but this recording has a low diagnostic yield due to its short duration and tendency to miss patients who have infrequent AFib episodes. 

There are plenty of other companies looking to tap into ECG data to improve AFib diagnosis, including Apple and iRhythm

In 2019, Apple released study findings showing that 84% of people who received an irregular heart rhythm notification on their Apple Watch were found to be in AFib at the time of the notification. Now, Apple is investigating whether its watch’s ECG feature can be used to detect other types of arrhythmias. 

In February, iRhythm released results from a study in which Kaiser Permanente researchers tested its Zio patch, a device that provides 3-14 days of uninterrupted heart monitoring. The findings showed the Zio patch did a better job of detecting AFib than a 30-day event monitor.

Cardiologs’ study differs from those conducted by Apple and other companies because Cardiologs is the first firm to test AI’s ability to predict AFib in the short-term using a 24-hour Holter recording. Resting 12-lead ECGs, which record the heart’s electrical activity from 12 electric patches throughout the body for a short period, are the standard test for measuring the heart’s electrical function. These ECGs give physicians access to a larger view of the heart’s activity than a Holter recording, but the latter provides longer-duration signals, which therefore gives the AI model more input.

Photo: BrianAJackson, Getty Images