PPGAF-score: A photoplethysmographic risk prediction tool for developing atrial fibrillation

Luisa Freyer (München)1, A. Krasniqi (München)1, P. Spielbichler (München)1, L. von Stülpnagel (München)1, M. Klemm (München)1, S. Massberg (München)1, A. Bauer (Innsbruck)2, K. Rizas (München)1

1LMU Klinikum der Universität München Medizinische Klinik und Poliklinik I München, Deutschland; 2Tirol Kliniken GmbH Kardiologie und Angiologie Innsbruck, Österreich



Atrial fibrillation (AF) is associated with increased morbidity and mortality. eBRAVE-AF was a siteless, randomized, controlled trial that showed that photoplethysmographic (PPG), smart-device based screening for AF significantly increased the detection rate of treatment-relevant AF, when compared to usual care. In this post-hoc analysis, we hypothesized that information derived from sequential PPG-measurements in individuals being in sinus rhythm, might be predictive for future AF- development and we aimed to develop a dynamic AF-risk score, based on demographic and PPG-derived data (PPGAF-score).



In eBRAVE-AF, individuals who were free of AF at baseline were randomized to digital screening or usual care. For digital screening, participants used a certified app to screen for irregularities by means of PPG, and abnormal findings were confirmed by a 14-day external ECG loop recorder. The primary endpoint was newly diagnosed AF treated with oral anticoagulation (OAC) within 6 months after randomization. After 6 months, participants were invited to cross-over for a second study phase with reverse assignment. In this sub-study, we identified participants who performed PPG-measurements both on day (8 am-6 pm) and at night (10 pm-6 am). PPGAF-score was calculated using the coefficients derived from Cox regression analysis. The multivariable model included age ≥ 65 years, sex and three PPG-derived parameters: (i) mean heart rate (HR) of each single PPG-measurement, which was dichotomized at the median value, (ii) the standard deviation (SD) of the HR between subsequent PPG-measurements, which was also dichotomized at the median value and (iii) the failure to demonstrate nocturnal HR reduction (day HR/night HR < 1). PPGAF-score was dichotomized at the median value.


Of the 2,436 participants included in this analysis, 1,208 (50%) were ≥ 65 years old and 776 (32%) participants were females. Cumulatively 186,551 PPG measurements were performed, corresponding to a median of 28 (IQR 23) daytime measurements and a median of 3 (IQR 5) nighttime measurements per participant. The median HR per measurement was 69 (IQR 11) bpm, the median SD of the HR was 6 (IQR 2) bpm, and 759 (31%) participants failed to show a nocturnal HR reduction. During a follow-up time of 12 months, 53 subjects reached the primary endpoint. Cox regression analysis showed that age ≥ 65 years (HR 3.6; 95% CI 1.9–7.0; p < 0.001), male sex (2.7; 95% CI 1.2–5.9;p = 0.017), HR < 69bpm (HR 1.9; 95% CI 1.1–3.5;p=0.028), SD of HR ≥ 6bpm (HR 2.2; 95% CI 1.2–4.0;p=0.011) and the failure to show nocturnal HR reduction (HR 3.2; 95% CI 1.9–5.5; p < 0.001) were all independent predictors of the primary endpoint. PPGAF-score was calculated from the multivariable model. The median value was 0.22. PPGAF-score > 0.22 compared to a risk score ≤ 0.22 was associated with a significantly higher risk for developing AF requiring OAC (HR 6.9, 95% CI 3.1–15.4;p < 0.001; Figure 1). 



Combining demographic and PPG-derived parameters, we developed a dynamic prediction tool for assessing the one-year risk of developing AF requiring OAC. A risk score of > 0.22 was associated with a 7-fold increased risk for developing treatment-relevant AF. PPG recordings are easy to perform, readily available and might provide a valuable tool not only for diagnosing AF, but also for selecting high-risk individuals, who might benefit from intensified AF-screening. 

Diese Seite teilen