Inter-subject analysis of transfer function coherence in studies of dynamic cerebral autoregulation

Objective: The gain and phase of the arterial blood pressure (BP)-cerebral blood flow velocity (CBFV) relationship, assessed by transfer function analysis (TFA), are widely used dynamic cerebral autoregulation (CA) metrics, but their reliability depend on the statistical significance of the magnitude squared coherence (MSC) function. We tested a new approach, based on inter-subject data, to estimate the confidence limits of MSC. Approach: Five minute beat-to-beat time series of mean arterial BP (MAP, Finometer) and CBFV (transcranial Doppler) were used for intra-subject (MAP and CBFV from same subject) and inter-subject (BP and CBFV swapped between subjects) estimates of MSC. The 95% confidence limit of MSC was obtained by non-parametric methods for the cases of single frequency harmonics in the range [0.02-0.50 Hz], and also from the mean value of all possible frequency intervals in this range. Main results: Intra-subject estimates of MSC were obtained from 100 healthy subjects (48 female, age range: 21-82 years old) allowing calculation of 9,900 inter-subject estimates, with 95% confidence limits in excellent agreement with classical values derived from surrogate random data. Confidence limits of MSC, derived from mean values, decreased asymptotically to around 0.16 with the increasing number of harmonics averaged. Significance: Replacing estimates of MSC at a single frequency harmonic by the mean calculated over the range [0.02-0.30 Hz] could lead to more robust studies of dynamic CA with greater acceptance of recordings, an important consideration in clinical studies where measurements tend to be more susceptible to noise and artefacts. .