Photoplethysmography as a tool to assess heart rate variability and predict mortality in the general population: the INVADE study

Abstract

Identifying individuals at risk before symptoms arise remains a central challenge in preventive cardiology. Heart rate variability (HRV) includes a set of parameters derived from beat-to-beat variation in cardiac rhythm reflecting autonomic nervous system activity, which have emerged as powerful prognostic markers. Traditionally its measurement has relied on the electrocardiogram, limiting its use to clinical settings. The photoplethysmographic signal, an optical biosignal that measures light absorption changes caused by pulsatile blood flow, offers a non-invasive and widely accessible alternative. Embedded in millions of consumer smartwatches and wearable devices. This technology places continuous monitoring in the hands of the general public.

This study assessed whether heart rate variability metrics derived from photoplethysmography agree with gold-standard electrocardiography measurements and to determine whether photoplethysmography based measures can serve as feasible prognostic tools in a general population cohort.

Within a population-based survey in Germany, the INVADE study, participants aged ≥60 years were enrolled between August 2013 and February 2015 and followed for a median of 10.2 years. A total of 1,956 subjects were examined using a simultaneous recording of ECG and PPG at the time of enrolment. Heart rate and heart rate variability metrics such as SDNN, RMSSD and Deceleration Capacity (DC) were calculated conventionally from the electrocardiogram and additionally from the photoplethysmographic signal.

Pearson correlation coefficients between ECG-derived and PPG-derived HRV parameters were r=0.89 for SDNN, r=0.94 for rmssd and r=0.74 for Deceleration Capacity (all p<0.01). Given the non-normal distribution of DC, Spearman rank correlation was additionally computed, yielding a higher coefficient of ρ=0.86 (p<0.01) (Figure 1). Bland-Altman analysis demonstrated a negligible mean bias of -0.03 ms for DC with limits of agreement of -7.54 ms to +7.49 ms, confirming absence of systematic bias between ECG- and PPG-derived measures. Kaplan-Meier analysis identified DC as the strongest predictor of all-cause mortality among the investigated parameters, with significantly higher mortality observed in patients with a reduced DC (DC<4.5 ms) (log rank p<0.001) (Figure 2).

PPG-derived HRV parameters demonstrate strong agreement with gold-standard ECG-based measurements, supporting the validity of PPG-derived autonomic indices. Notably, PPG-based DC significantly predicted all-cause mortality over a median of 10.2 years, highlighting its prognostic utility as a widely applicable, non-invasive autonomic parameter for cardiovascular risk stratification in the general population.Figure 1Figure 2