Variability of the autoregulation index decreases after removing the effect of the very low frequency band

Abstract

Dynamic cerebral autoregulation (dCA) estimates show large between and within subject variability. Sources of variability include low coherence and influence of CO2 in the very low frequency (VLF) band, where dCA is active. This may lead to unreliable transfer function and autoregulation index (ARI) estimates. We tested whether variability of the ARI could be decreased by suppressing the effect of the VLF band through filtering. We also evaluated whether filtering had any effect on mean group differences between healthy subjects and acute stroke patients.Data from a recent mobilization stroke study were re-analyzed. Middle cerebral artery cerebral blood flow velocity (MCA-CBFV), mean arterial blood pressure (MABP) and end tidal PCO2 (PetCO2) were obtained in 16 healthy subjects and 27 acute ischemic stroke patients in the supine position. The ARI index was calculated from the transfer function (TF) by using spontaneous BP fluctuations. Three different filtering strategies were compared; no filtering (NF), a high pass filter at 0.04Hz (Time Domain Filtering: TDF) and a high pass Transfer Function Filter (TFF) at 0.04Hz. In addition, a simulation study was done to obtain further insight into the effects of the applied filters.The variability of the ARI index decreased significantly only with TFF in healthy subjects (standard deviation (left vs. right) after NF 2.28 vs. 2.36, after TDF 2.13 vs. 2.31 after TFF 1.09 vs. 1.19, p<0.001).Variability was not significantly reduced in stroke patients. The mean ARI was significantly lower in stroke patients compared to healthy subjects after TFF (affected hemisphere 5.85±1.96 vs. 7.13±1.09, non-affected hemisphere 5.96±1.64 vs. 7.31±1.19, p<0.01 for both hemispheres), but not after NF or TDF. The simulation study showed that TFF results in an overestimation of the ARI index at low ARI levels (0–3), but in correct estimates at higher ARI levels.Removing the effect of the VLF band with TFF results in less ARI variability in healthy subjects, and in more pronounced group differences between stroke patients and healthy subjects. This will improve diagnostic properties when using TFA for ARI calculation.