Velocity Reflection Index and Chronological Age: Confounder-Adjusted Statistical and Machine-Learning Analyses of Carotid Doppler Waveforms
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Purpose
To determine whether the velocity reflection index (VRI) is the carotid Doppler waveform feature most strongly associated with chronological age after adjustment for sex and exercise habit, and whether its feature ranking remains stable across cross-validated and cohort-sensitivity analyses.
Methods
Eight waveform-derived features were analysed in 197 participants meeting the study eligibility criteria and measured using a validated continuous-wave carotid Doppler system. Pearson and partial correlations and multivariable regression evaluated associations with chronological age. Random Forest regression with repeated 10-fold cross-validation, held-out permutation importance and bootstrap resampling assessed feature ranking. Sensitivity analysis evaluated the influence of cohort construction.
Results
VRI showed the strongest association with chronological age (r = 0.738, 95% CI [0.667, 0.796]) and remained strongly associated after adjustment for sex and exercise habit (partial r = 0.798). VRI ranked first by both impurity-based (0.536) and held-out permutation (0.765) importance; repeated cross-validation yielded MAE = 6.87 ± 1.26 years and R 2 = 0.572 ± 0.153. Its leading ranking was stable in 85.3% of bootstrap resamples and the age-VRI correlation was essentially unchanged in the cohort-sensitivity analysis. The exercise association was significant after age adjustment (B = -0.043, p = 0.018) but attenuated after additional adjustment for sex (B = -0.026, p = 0.098). The sex association remained significant after adjustment for age and height.
Conclusion
VRI was robustly associated with chronological age and retained the leading feature-importance ranking across adjusted statistical and cross-validated machine-learning analyses. Validation against an established arterial-stiffness measure in an independent cohort is required before VRI can be considered a clinical vascular-aging biomarker.
