Background: Carotid-femoral pulse wave velocity (PWV) is a well-known marker of arterial stiffness and a cardiovascular risk factor. Novel estimations of PWV have been developed, but their ability to improve cardiovascular prediction made by clinical risk tools remains controversial. We hypothesized that PWV estimations can improve cardiovascular risk estimation beyong what is achievable from standard risk factors. Methods: Analysis of the population based CARTaGENE cohort, including participants aged between 40 and 69 years. PWV estimations were obtained using published formulas (fPWV) or algorithmic transformation of tonometry-based radial pulse waveforms captured at baseline (aPWV). 10-year cardiovascular risk for each participant was computed using the Atherosclerotic Cardiovascular Disease (ASCVD) and the SCORE-2 risk tools. All participants were passively followed during 10 years for major adverse cardiovascular events (MACE: cardiovascular death, stroke, myocardial infarction). Associations of fPWV and aPWV with MACEs were obtained using Cox models adjusted for ASCVD or SCORE-2 predictions. Results: 17,548 participants were eligible for the study (51.1% female, median age 53 yo, 8.9% diabetes, 13.9% prior cardiovascular disease), from which 2,263 (12.9%) experienced a MACE during follow-up. Mean PWV values at baseline were 8.4 ± 1.4 m/s (fPWV) and 7.9 ± 1.3 m/s (aPWV), and were closely correlated (correlation coefficient= 0.93). Both fPWV (HR= 1.52, 95% CI [1.47-1.58]) and aPWV (HR=1.60 [1.54-1.66]) were predictive of MACE in unadjusted models. Only aPWV remained significantly associated after adjustments for ASCVD (Hazard ratio [HR]= 1.16 [1.09-1.22]) and SCORE-2 (HR= 1.07 [1.00-1.13]). fPWV was significantly associated with MACE after adjustment for ASCVD, but not SCORE-2, only when this was tested in a population with similar exclusions as in the original fPWV derivation cohort. Conclusions: Algorithm-based PWV, but not formula-based PWV, improves cardiovascular prediction beyond what is achievable with recognized prediction tools.
Read full abstract