- Atrial fibrillation (AF) is the most common sustained arrhythmia and a major contributor to stroke.
- Wearable ECG devices may improve AF detection, but automated rhythm classification requires validation against clinically adjudicated ECG
- We evaluated the Frontier X Plus (FX+) (Fourth Frontier Technologies LLC, Austin, TX) chest-worn single-lead ECG (Fig. 1) against simultaneous Holter monitoring and additionally assessed interval agreement and waveform lead correspondence
- Design: Prospective validation study of 63 subjects with documented AF or sinus-rhythm controls
- Procedure: Simultaneous 5-minute supine resting recordings were obtained from FX+ and a standard 12-lead Holter (Mortara Instruments, Milwaukee, WI) (Figure 2)
- Rhythm was adjudicated by a cardiologist using the time-synchronized Holter ECG, blinded to FX+ output
- Analyses: Primary - AF detection performance; Secondary - Bland–Altman (BA) agreement for PR, QRSd, QT, and RR intervals, and Pearson correlation between the FF waveform and each Holter lead
- PR agreement was assessed only in analyzable sinus-rhythm recordings; AF, prolonged PR/AV block, and poorly defined P waves were excluded, resulting in a smaller PR sample
- To identify the Holter lead that best matched the wearable waveform, we first performed a lead-matched exploratory analysis in the subset with synchronized wearable files available for all 12 Holter leads (n=9). We then confirmed the strength of correlation for the identified lead in the larger subset with consistently available V5-aligned synchronized wearable files (n=55)
Figure 1. Frontier X Plus (FX+)
(Device provided by Fourth Frontier Technologies LLC, Austin, TX)
Figure 3. Bland–Altman Plots Comparing Fourth Frontier and Holter Intervals
- FX+ showed high AF-detection performance versus cardiologist-adjudicated Holter: Sensitivity 94.4% (95% CI 87.6–100.0) and specificity 97.9% (95% CI 94.7–100.0)
- There were no direct AF↔NSR misclassifications; when FX+ disagreed with cardiologist adjudication, it usually labeled the strip as inconclusive rather than assigning the opposite rhythm
- BA analysis showed excellent agreement for RR. Agreement for PR, QRSd, and QT was poorer - particularly with systematic underestimation of QRS duration - indicating that FF-derived interval measurements should not be considered diagnostically interchangeable with Holter (Fig. 3)
- In the exploratory lead-matched subset (n=9), V5 showed the highest median correlation with the FF waveform (median r=0.41). In the larger V5-aligned cohort (n=55), the FF waveform again showed the strongest correlation with Holter V5 (median r=0.67), confirming V5 as the closest 12-lead analog (Fig. 4).
- FX+ demonstrated high AF-detection performance against cardiologist-adjudicated Holter recordings
- Wearable-derived interval measurements showed closest agreement for RR, whereas PR, QRSd, and QT were not sufficiently concordant with Holter for diagnostic substitution.
- FF morphology most closely resembled Holter V5, supporting V5 as the most appropriate 12-lead analog for visual and interval-based comparisons
University of Rochester School of Nursing and University of Rochester Medical Center, Rochester, NY
Sudhir K Mummidi, MBBS, MS, RN; Daniel J. Lachant, DO; Mark A. Marinescu, MD; Jean-Philippe Couderc, PhD, MBA; Dillon J Dzikowicz, PhD, RN, PCCN
Validation of the Frontier X Plus Wearable ECG for Atrial Fibrillation Detection
Figure 2. Example Time Synchronization Plot Demonstrating Alignment between FX+ and Holter signals prior to adjudication
- Simultaneous recordings and blinded cardiologist adjudication provided a clinically relevant reference standard
- No AF↔NSR swaps were observed
- Recordings were short and resting/supine; ambulatory performance was not evaluated
- PR agreement and the lead-matched sensitivity analysis were performed in smaller subsets based on interval eligibility and synchronized file availability
- Cohort: 63 subjects, 59% male; 56% were >65 years; 67% had BMI ≥25 kg/m²in. Race: 84% White, 10% Black, 6% other; 3% Hispanic/Latino
Figure 4. Primary Common-reference Heatmap of Median Pearson Correlation between the Fourth Frontier Waveform and each Holter Lead (n = 55)