Transfer function analysis for clinical evaluation of dynamic cerebral autoregulation—a comparison between spontaneous and respiratory-induced oscillations

Abstract

Oscillations of arterial blood pressure (ABP) and cerebral blood flow velocity (CBFV) can be used for non-invasive assessment of cerebral autoregulation using transfer function analysis. Either spontaneous oscillations (SPO) around 0.1 Hz or respiratory induced oscillations during deep breathing (DB) at a rate of 6/min have been used so far. We investigated 168 patients with severe carotid stenosis or occlusion to evaluate transfer function analysis and compare the SPO and DB approaches. ABP was assessed non-invasively (Finapres), CBFV was measured in the middle cerebral artery using transcranial Doppler sonography. Transfer function phase (P) and gain (G) were extracted from the respective spectra in a low (0.06–0.12 Hz) and high (0.20–0.30 Hz) frequency range (LF, HF) of SPO and from the 0.1 (LF) and 0.2 (HF) Hz peaks induced by DB. For SPO, significant side-to-side differences and differences between groups of severe and critical stenosis were found for P(LF), while P(HF) did not prove to be a significant parameter. G(LF) showed significant side-to-side differences, while G(HF) additionally differed significantly between severe and critical stenosis and occlusion, respectively. For DB, significant side-to-side differences were found for P(LF, HF). Mainly G(HF) differed significantly between the affected and contralateral sides, while both HF and LF gains showed lower values in groups with a higher degree of stenosis. Correlation between G and P values was generally poor. Using Bland–Altman plots a poor inter-method agreement was found mainly for P. Correlations between SPO and DB were higher for G than for P (LF r = 0.64 versus 0.44, HF 0.69 versus 0.28). Analysing reproducibility in 16 patients, only for P(LF, HF) of DB was a highly significant correlation found (Spearman's r up to 0.78). For G(LF, HF) correlations were significant for both SPO and DB with slightly higher r coefficients for SPO. In conclusion, the present study showed that (1) transfer functions P and G represent different information for characterization of dynamic cerebral autoregulation in the frequency domain. (2) Inter-method agreement between DB and SPO is poor for P and moderate for G values. (3) P extracted from DB has a higher reproducibility. (4) The extraction of P and G from the SPO phase spectra is critical and future work on standardizing this process is needed. (5) At present, the DB protocol might be slightly advantageous as a routine diagnostic tool.