This study tested whether the velocity reflection index (VRI), a feature derived from carotid continuous-wave Doppler waveforms, is the waveform measure most strongly associated with chronological age after accounting for key confounders. The analysis also evaluated whether that ranking remains stable when using cross-validated machine-learning models and sensitivity analyses of cohort construction.
The dataset comprised 197 participants who met the study eligibility criteria and underwent carotid Doppler measurement using a validated continuous-wave system. Eight waveform-derived features were extracted from the carotid Doppler recordings; the source lists VRI as one of these features but does not enumerate the other seven in the report. Participant recruitment, informed consent, and institutional ethical approval were obtained; the protocol and data collection were approved by the Ethics Committee of Tokushima University Hospital.
Associations between each Doppler-derived feature and chronological age were assessed using Pearson correlations, partial correlations (to adjust for sex and exercise habit), and multivariable regression models. For feature-ranking and predictive performance, the authors trained Random Forest regression models with repeated 10-fold cross-validation. Feature importance was evaluated using both impurity-based importance and held-out permutation importance on unseen data. Bootstrap resampling assessed ranking stability. A cohort-sensitivity analysis tested whether the observed age–feature relationships persisted under alternative cohort constructions.
Across analyses, VRI emerged as the strongest correlate of chronological age. The zero-order Pearson correlation between VRI and age was r = 0.738 with a 95% confidence interval of [0.667, 0.796]. After adjusting for sex and exercise habit, the partial correlation remained strong (partial r = 0.798). In multivariable regression, associations described in the source indicate persistent and robust links between age and VRI.
In Random Forest feature-ranking, VRI held the top position by both impurity-based importance (0.536) and held-out permutation importance (0.765). The cross-validated predictive performance of the Random Forest model was reported as mean absolute error (MAE) = 6.87 ± 1.26 years and R² = 0.572 ± 0.153 across repeated folds.
Repeated 10-fold cross-validation and held-out permutation testing were used to mitigate overfitting and to evaluate generalizable importance estimates. Bootstrap resampling showed that VRI retained the leading ranking in 85.3% of resamples, indicating ranking stability. The cohort-sensitivity analysis reported that the correlation between age and VRI was essentially unchanged when cohort construction was varied, supporting robustness of the age–VRI relationship.
The study examined exercise habit and sex as potential confounders. Exercise habit showed a significant association after adjusting for age in regression models (B = -0.043, p = 0.018), but this association was attenuated and became non-significant after further adjustment for sex (B = -0.026, p = 0.098). Sex remained significantly associated with the outcome after adjustment for age and height. The source does not provide additional effect sizes or subgroup breakdowns beyond these reported coefficients and p-values.
The author concludes that VRI is robustly associated with chronological age and consistently ranks as the leading Doppler-waveform feature in both confounder-adjusted statistical analyses and cross-validated machine-learning models. These results suggest VRI could be a candidate marker of vascular aging derived from carotid Doppler waveforms. However, the report emphasizes that validation against an established arterial-stiffness measure in an independent cohort is required before VRI can be considered a clinical vascular-aging biomarker.
The source notes that de-identified data are not publicly available due to participant privacy and institutional restrictions; requests may be considered subject to institutional and ethics approval. Analysis code is available from the corresponding author upon reasonable request. The report does not provide the full list of the eight waveform-derived features in the text supplied here and does not present subgroup distributions or detailed demographic tables in the extracted content. External validation against established arterial stiffness measures and replication in independent cohorts are required steps identified by the author prior to clinical application.