Comparison Between Bivariate Phase-Rectified Signal Averaging and Sequence Method in Assessing the Baroreflex Sensitivity

Massimo W Rivolta1 and Roberto Sassi2
1Dipartimento di Informatica, Università degli Studi di Milano, 2Università degli Studi di Milano, Dipartimento di Informatica


Aims: Baroreflex sensitivity (BRS) measures the rate of changes in inter-beat time intervals (RR) with respect to changes in blood pressure (BP). Recently, Bivariate Phase-Rectified Signal Averaging (BPRSA) was proposed as possible BRS estimator and was compared with the validated sequence method (SM). However, the two methods differ substantially, questioning whether BPRSA might be considered as an estimator of BRS. In this study, we investigated the role that the coupling between RR, BP and respiration has on BRS estimates provided by BPRSA and SM.

Methods: Multivariate autoregressive models (MVAR) were fitted to the data of 10 healthy subjects that underwent a tilt test. MVAR models were then used to generate synthetic signals while artificially varying the coupling between RR, BP and respiration. We also investigated on the behavior of the two methodologies when applied to two independent White Gaussian Noises (WGN). In this case, we found a closed-form solution of the SM algorithm (one of its variants).

Results: Both BPRSA and SM provided higher BRS values during supine with respect to head-up phase. Computerized simulations showed little influence of the coupling between respiration and RR on both estimators, whereas a major difference appeared when the coupling between BP and RR was removed. In this latter case, BPRSA almost vanished, while SM increased of about 10%, regardless the phase. In case of WGN, we found that BRSA converged to zero BRS, while SM was proportional to the ratio of the standard deviations of RR and BP.

Conclusions: SM and BPRSA seemed to carry similar information, likely due to the strong coupling between SBP and RR, while respiration played a little role. However, we do not exclude that other physiological interactions may produce differences between the estimates of the two methodologies.