Prediction of Cardiovascular Disease Events From the Photoplethysmograph Waveform.

BACKGROUND: A photoplethysmographic pulse waveform (PPG) can be obtained optically by digital health technologies such as a smartphone or wearable devices. Features of the PPG may provide prognostic information additive to blood pressure. METHODS AND RESULTS: We performed comprehensive feature extraction of PPGs recorded during 2009 to 2010 in 114 884 participants in UK Biobank and examined the association of individual and panel features with incident cardiovascular disease (CVD) events over approximately 10 years of follow-up. A total of 9242 participants developed CVD events. There were 3378 deaths, with CVD the primary cause of death in 417 participants. In a penalized Cox proportional hazards model, after adjustment for classical risk factors (age, sex, body mass index, ethnicity, smoking, presence of diabetes, total cholesterol/high-density lipoprotein cholesterol ratio, systolic blood pressure, and medication), PPG indices that were most strongly associated with CVD events included systolic time, the area under the systolic portion of the PPG, and curvature of the mid-late systolic PPG with standardized hazard ratios of 1.43 (95% CI, 1.28-1.61), 0.79 (95% CI, 0.72-0.88), and 0.80 (95% CI, 0.74-0.86), respectively. The addition of PPG indices to classical risk factors increased the prediction of CVD events and this was more marked in younger compared with older people. CONCLUSIONS: These results suggest that the PPG features related to systolic performance and that may be related to preclinical heart failure, offer important prognostic information. Further studies to identify the physiological properties to which they relate and to optimize their use in risk prediction are merited.