Between-centre variability in transfer function analysis, a widely used method for linear quantification of the dynamic pressure-flow relation: the CARNet study

Abstract

Transfer function analysis (TFA) is a frequently used method to assess dynamic cerebral autoregulation (CA) using spontaneous oscillations in blood pressure (BP) and cerebral blood flow velocity (CBFV). However, controversies and variations exist in how research groups utilise TFA, causing high variability in interpretation. The objective of this study was to evaluate between-centre variability in TFA outcome metrics. 15 centres analysed the same 70 BP and CBFV datasets from healthy subjects (n=50 rest; n=20 during hypercapnia); 10 additional datasets were computer-generated. Each centre used their in-house TFA methods; however, certain parameters were specified to reduce a priori between-centre variability. Hypercapnia was used to assess discriminatory performance and synthetic data to evaluate effects of parameter settings. Results were analysed using the Mann-Whitney test and logistic regression. A large non-homogeneous variation was found in TFA outcome metrics between the centres. Logistic regression demonstrated that 11 centres were able to distinguish between normal and impaired CA with an AUC>0.85. Further analysis identified TFA settings that are associated with large variation in outcome measures. These results indicate the need for standardisation of TFA settings in order to reduce between-centre variability and to allow accurate comparison between studies. Suggestions on optimal signal processing methods are proposed.

Keywords: Cerebral autoregulation; Method comparison; Standardisation; Transfer function analysis.

Conflict of interest statement

Conflict of interest

None declared.

Copyright © 2014 IPEM. Published by Elsevier Ltd. All rights reserved.

Figures

Fig. 1

Average transfer function results (gain (A), phase (B), and coherence (C)) of the centres found for one subject during normocapnia. Results are represented as median (black line), min/max (grey dashed lines), and 1st/3rd quartile (grey solid lines) for the gain in cm/s/mmHg (top), phase in radians (middle), and coherence (bottom).

Fig. 2

Transfer function gain (A) and phase (B) of the artificial datasets ARI 0, 2, 4 and 6, analysed using one set of parameter settings for TFA. The specified parameter settings were: sample frequency = 10 Hz, window length = 95 s, anti-leakage window = Hanning, number of windows = 5, percentage of superposition = 50%, frequency bands were VLF: 0.02–0.07 Hz; LF: 0.07–0.15 Hz; HF: 0.15–0.4 Hz.

Fig. 3

Relation between the average outcomes for phase (y-axis), gain (z-axis), and coherence (x-axis) of the centres for the very low frequency (A), low frequency (B), and high frequency (C). Vertical dashed lines indicate the projection of each point on the x–y plane. Centre 14 is excluded from the graph, because this centre did not report any values for the coherence.

Fig. 4

Phase results in radians (y-axis; A–C) and gain results in cm/s/mmHg (y-axis; D–F) for each centre (x-axis) showing the differences found between normocapnia (dark boxes) and hypercapnia (light boxes). The spread between the results is represented using multiple box plots. On each box, the central mark is the median value, the edges of the box are the 25th and 75th percentiles, and the whiskers extend to the most extreme data points. Asterisks (*) indicate that the centre found a significant difference between normocapnia and hypercapnia (p < 0.05). The graphs show the phase results for the very low frequency (A and D), low frequency (B and E), and high frequency (C and F).

Fig. 5

Receiver operating characteristics (ROC) curves of the centre with the best ability to distinguish (centre 3) between normocapnia and hypercapnia and the centre which is least able to distinguish (centre 14). ROC curves are derived using logistic regression. Logistic regression was performed using the state of CA (normal = baseline, impaired = hypercapnia) as the outcome value and as predictor variables the phase VLF, phase LF, phase HF, gain VLF, gain LF, gain HF, coherence VLF, coherence LF, and coherence HF.

Fig. 6

Gain (A), phase (B), and coherence (C) results (y-axis) for ARI 0, 2, 4, and 6. The spread between the centres is represented using multiple box plots. On each box, the central mark is the median value of the centres of the specific subject, the edges of the box are the 25th and 75th percentiles, and the whiskers extend to the most extreme data points. Results are shown for the very low frequency (VLF), low frequency (LF), and high frequency (HF). Asterisks (*) indicate the values for transfer function phase and gain that were obtained by the software (written in Matlab by DS) that was used to generate the datasets.

Fig. 7

The transfer function results expressed in VLF gain, LF gain, VLF phase, and LF phase of ARI = 6 for three different parameter settings: sample frequency (1, 5, 10, 20, and 50 Hz) (A), window length (25, 50, 75, and 95 s) (B), and different frequency bands (C). Standard indicates the actual values of gain and phase of the 10 ARI models were used as criterion standard which were used as criterion standard value.