NON-INVASIVE VOICE BIOMARKER IS ASSOCIATED WITH INCIDENT CORONARY ARTERY DISEASE EVENTS AT FOLLOW-UP

Abstract

Background: Voice biomarkers have been shown to be associated with coronary artery disease (CAD) diagnosed by angiography, hospitalization in patients with heart failure, and with hemodynamic indices of pulmonary hypertension. In the current study we evaluate the association between a pre-identified voice biomarker and incident CAD events at follow-up. Methods and Results: Patients referred for clinically indicated coronary angiography underwent a total of three 30-second voice recordings using the Vocalis Health smartphone application between January 1, 2015, and February 28, 2017: R1—reading aloud a pre-specified text; R2—describing a positive emotional experience; and R3—describing a negative emotional experience. A pre-established voice biomarker was derived from each individual recording, and the mean biomarker value was calculated for each patient. Individuals were clinically followed through December 31st 2019. The pre-specified primary outcome was a composite of presenting to the emergency department with chest pain and/or being admitted to hospital with chest pain and/or having unstable angina, non-ST-segment elevation myocardial infarction (NSTEMI) or STEMI; the pre-specified secondary outcome was a composite of a positive stress test at follow-up and/or the presence of CAD at follow-up coronary angiography. The association between the baseline voice biomarker and follow-up events was estimated using Cox proportional hazards models. One hundred and eight patients were included in the final analysis (mean age: 59.47 ± 11.44 years, males: 54.6%). The median follow-up time was 24 months (range 1 – 60 months). Overall, 43 (39.8%) patients had the composite primary outcome, and 15 (13.8%) had the composite secondary outcome. Patients were separated into three tertiles, according to the distribution of the mean voice biomarker (T1 lowest tertile, T2 middle tertile, T3 highest tertile), and were divided into two groups based on having a high (T3) versus low (T1 and T2) mean baseline voice biomarker value. Patients with a high mean voice biomarker value at baseline had a significantly higher frequency of the primary composite end-point compared to those with a low mean voice biomarker value at baseline (21 (58.3%) vs. 22 (30.6%), p=0.0056). The frequency of the secondary composite end-point was similar between groups (8 (22.2%) vs. 7 (9.7%), p=0.0848). In multivariable Cox proportional hazards models adjusting for CAD grade on baseline angiography, a high baseline mean voice biomarker was significantly associated with both the primary and secondary composite outcomes (HR 2.611, 95% CI 1.421 – 4.798, p=0.002; and HR 3.132, 95% CI 1.130 – 8.679, p=0.028). Conclusion The current study demonstrates a significant association between a noninvasive voice biomarker and incident CAD events at follow-up. These results may have important clinical implications for telemedicine, and the remote and non-invasive screening of patients to identify those at risk of coronary disease and its complications. Funding: This study was in part supported by Vocalis Health, Tel-Aviv Israel Declaration of Interest: All authors have no conflicts of interest to disclose Ethical Approval: protocol was approved by the institutional review board at Mayo Clinic, and all patients provided informed consent for participation.