Novel method for intraoperative assessment of cerebral autoregulation by paced breathing

Abstract

Background. Cerebral autoregulation (CA) is the mechanism that maintains constancy of cerebral blood flow (CBF) despite variations in blood pressure (BP). Patients with attenuated CA have been shown to have an increased incidence of peri-operative stroke. Studies of CA in anaesthetized subjects are rare, because a simple and non-invasive method to quantify the integrity of CA is not available. In this study, we set out to improve non-invasive quantification of CA during surgery. For this purpose, we introduce a novel method to amplify spontaneous BP fluctuations during surgery by imposing mechanical positive pressure ventilation at three different frequencies and quantify CA from the resulting BP oscillations.

Methods. Fourteen patients undergoing sevoflurane anaesthesia were included in the study. Continuous non-invasive BP and transcranial Doppler-derived CBF velocity (CBFV) were obtained before surgery during 3 min of paced breathing at 6, 10, and 15 bpm and during surgery from mechanical positive pressure ventilation at identical frequencies. Data were analysed using frequency domain analysis to obtain CBFV-to-BP phase lead as a continuous measure of CA efficacy. Group averages were calculated. Values are means (sd), and P<0.05 was used to indicate statistical significance.

Results. Preoperative vs intraoperative CBFV-to-BP phase lead was 43 (9) vs 45 (8)°, 25 (8) vs 24 (10)°, and 4 (6) vs −2 (12)° during 6, 10, and 15 bpm, respectively (all P=NS).

Conclusions. During surgery, cerebral autoregulation indices were similar to values determined before surgery. This indicates that CA can be quantified reliably and non-invasively using this novel method and confirms earlier evidence that CA is unaffected by sevoflurane anaesthesia.

Clinical trial registration. NCT03071432.

Cerebral autoregulation (CA) is a physiological mechanism that maintains constant cerebral blood flow (CBF) despite variations in blood pressure (BP) and therefore safeguards cerebral metabolic needs during hypotension.¹ If the efficacy of CA is compromised, the risk of cerebral hypoperfusion increases when BP is lower, for example during anaesthesia.² Indeed, attenuated CA, as found in patients with diabetes,³ hypertension,⁴ vascular disease,^5,6 or previous cerebral infarction,⁷ has been demonstrated to be associated with an increased incidence of perioperative stroke.^8–10 Nevertheless, studies that quantify CA efficacy in perioperative patients are scarce because a standardized non-invasive measurement technique is lacking.

Studies of CA efficacy involve assessment of the CBF response to a BP modification, and in such studies, CBF is usually assumed to correlate closely with CBF velocity (CBFV), which is more easily measured. After a sudden BP change provoked by the release of inflated thigh cuffs^11,12 or vasoactive medication,¹³ CBFV,¹⁴ measured with a transcranial Doppler system, returned to baseline within a few seconds. This latency has been used as a measure of CA efficacy, and using these techniques, several studies demonstrated that sevoflurane anaesthesia does not affect CA.^15–17 To eliminate the need to induce a BP modification, an alternative non-invasive method quantifies CA from the CBFV response to spontaneous BP fluctuations originating from short-term BP control by the sympathetic nervous system.^18–20 Unfortunately, this method is inappropriate in anaesthetized patients because sevoflurane suppresses sympathetic outflow²¹ and thus the necessary spontaneous BP oscillations.²² Indeed, the only study attempting to quantify CA during surgery using this method shows obliteration of BP oscillations with sufficient amplitude.²³ This renders intraoperative non-invasive CA quantification effectively impossible.

In the present study, we set out to improve non-invasive CA quantification during surgery. For this purpose, we introduce a novel method to amplify spontaneous BP fluctuations by imposing mechanical positive pressure ventilation at three different frequencies. Intraoperative CA efficacy determined from the resulting imposed BP oscillations was then compared with preoperative values during paced breathing at identical frequencies. We hypothesized that CA indices determined with this novel method are unaffected by sevoflurane anaesthesia.

Methods

The study was approved by the Institutional Ethics Committee of the Academic Medical Centre AMC Amsterdam (reference: MEC 2016_116) and registered with ClinicalTrials.gov (ref: NCT03071432). Oral and written informed consent were obtained before inclusion.

Patients and study design

A total of 18 patients were included in this prospective study. Two patients were excluded because of preoperative insufficient adherence to the protocol and another two because exclusion criteria were met (no transcranial window and cardiac arrhythmia during surgery, respectively). The remaining 14 patients [four male (29%); mean age 58 (range 43–73) yr, weight 73 (range 57–99) kg, and height 172 (range 158–185) cm] completed the protocol both before and during surgery. Exclusion criteria were laparoscopic procedures, known diabetes mellitus, Parkinson’s disease, cardiac arrhythmia, central or peripheral autonomic neuropathy, Shy-Drager syndrome, cardiomyopathy (known), and the inability to detect transcranial Doppler signals because of anatomical variances.

Frequency domain analysis of CA

In awake subjects, BP, and consequently CBFV, fluctuate spontaneously around two predominent frequencies. The low frequency (LF; ∼0.1 Hz or a wavelength ∼10 s) is attributable to baroreflex-mediated sympathetic nervous system activity,^18,19 and the high frequency (HF; ∼0.25 Hz or a wavelength of ∼4 s) is attributable to respiration (Fig. 1A and B). Using ‘frequency domain’ analysis,^24,25 the relative power (or amplitude) of these fluctuations can be calculated and analysed in order to determine CA efficacy. Given that CA is a mechanism that requires ∼5 s⁷ to react to BP disturbances, HF BP oscillations pass unaltered into CBFV oscillations.²⁰ In contrast, LF BP oscillations are counter-regulated and damped by CA. Therefore, LF oscillations in CBFV are out of phase with LF oscillations in BP, and the phase lead of CBFV to BP in the LF domain (expressed in degrees, °) is the main measure for CA efficacy. When CA is intact, this is reflected as a ∼50° LF phase lead, whereas impaired CA is characterized by a ∼30° LF phase lead.^3,7

Coherence

The squared coherence function reflects the fraction of CBFV variation that can be linearly attributed to BP variation. In a similar manner to a correlation coefficient, it varies between zero and one and must be sufficiently high to be able to calculate the BP–CBFV relationship (i.e. CA). In awake subjects, spontaneous LF BP oscillations are of sufficient magnitude (Fig. 1A and B), expressed as power. During anaesthesia, however, the power (or amplitude) of LF BP fluctuations is extremely low,²³ owing to suppression of the autonomic nervous system by anaesthetic agents,²¹ and remaining BP variation is solely attributable to the intrathoracic pressure effects of mechanical positive pressure ventilation (Fig. 1C and D).²² Any LF oscillation still present is considered ‘noise’; hence, coherence is low. This presents the researcher with a problem and with a possible solution. When LF BP oscillations are desired for CA calculation, they might be imposed by adjusting mechanical ventilation to different frequencies (Fig. 1E and F). In the present study, we created consistent BP oscillations of a sufficient power at 6, 10, and 15 bpm or 0.1, 0.17, and 0.25 Hz, respectively, and determined the CBFV-to-BP relationship for each breathing frequency. These breathing frequencies may not always be within the physiological range for mechanical ventilation or respiratory rate. Indeed, they are not intended to examine the effects of respiratory rate per se, but these breathing frequencies are imposed to induce an oscillation into the BP signal to examine CA efficacy. Given that CA is calculated from sympathetic system-induced LF BP oscillations at ∼0.1 Hz in awake subjects, we consider the imposed breathing frequency of 6 bpm, creating a BP oscillation around 0.1 Hz, as the most important outcome measure in this study (Fig. 1E and F). Given that higher frequencies are normally regulated less optimally by CA, these may serve as controls to determine whether this method can reconstruct the complete ‘high-pass filter’ that CA mathematically represents.

Instrumentation

Continuous non-invasive, beat-to-beat arterial blood pressure recordings were obtained with a servo-controlled finger photoplethysmograph (^ccNexfin; Edwards Lifesciences, Irvine, CA, USA), with the cuff placed over the middle phalanx of the middle finger of the left hand and a heart-level reference device in place to correct for height differences between the finger and the heart. Validation studies have shown that ^ccNexfin is able to measure and track changes in blood pressure with a high precision and low offset throughout a wide range of blood pressures during cardiac²⁶ and vascular²⁷ surgery. The system performs calibration procedures to adapt the settings of the feedback loop, if necessary, automatically (‘Physiocal’), with increasing time between calibration steps when a stable signal is obtained. We consider blood pressure measurement reliable if Physiocal calibration is separated by ≥30 heart beats.²⁸ The middle cerebral artery was unilaterally insonated through the temporal window with a 2 MHz probe (Compumedics DWL Germany GmbH, Singen, Germany) attached to a headband (DWL Diamon Probe Fixation System; Compumedics DWL Germany GmbH). Insonation depth and signal gain were adjusted to an optimal signal-to-noise ratio. Continuous partial pressure of CO₂ (⁠$PCO2$⁠) from exhaled air (Normocap 200; Datex-Ohmeda) was collected via side-stream sampling through a nasal cannula (before surgery) or tracheal tube (during surgery). End-tidal $PCO2$ (P_etco₂) was determined during real time and displayed throughout the measurements. The signals of BP, the envelope of CBFV, CO₂, and a marker signal were stored with a 200 Hz sample rate for off-line analysis. CBFV_mean and mean arterial pressure (MAP) were the calculated as the integrals of the CBFV and arterial pressure waveforms throughout one heart beat, and heart rate was calculated as the inverse of the inter-beat interval.

Preoperative protocol

Measurements were performed in a temperature-controlled laboratory room at 22 °C, with patients resting in the semi-supine position. Upon arrival, patients were familiarized with the study protocol. After instrumentation and stabilization of all parameters, patients were asked to breath synchronously with a metronome set at a frequency of 6, 10, and 15 bpm for 3 min consecutively. The P_etco₂ was continuously monitored by the researcher, who instructed patients to adjust their inspiration depth to maintain P_etco₂ constant. Also, during a 3 min period of supine rest, BP and CBFV signals were recorded.

Intraoperative protocol

Routine anaesthesia monitoring and study instrumentation were applied. General anaesthesia was induced i.v. with propofol 1.0–3.3 mg kg⁻¹, sufentanil 0.2–0.5 μg kg⁻¹, lidocaine 0–1.0 mg kg⁻¹, and rocuronium 0.0–1.2 mg kg⁻¹. The trachea was intubated and positive pressure ventilation commenced. Anaesthesia was maintained with sevoflurane at an end-tidal concentration of 0.6–1.2 MAC adjusted to age using the formula proposed by Mapleson.^29,30 The fraction of inspired O₂ was 0.4. Arterial blood pressure was supported with a continuous infusion of norepinephrine 0–0.1 μg kg⁻¹min⁻¹. At least 20 min after induction, during stable haemodynamic situations, with the patient in the supine position, the ventilator was adjusted to a frequency of 6, 10, and 15 bpm for 3 min consecutively after values had stabilized. Minute volume was adjusted to maintain P_etco₂ constant.

Statistical analysis and sample size calculation

All data were analysed offline after artefact removal (if applicable). Two minute (120 s) episodes of data per intervention were selected, spline interpolated, and resampled at 4 Hz. With discrete Fourier transform, power, coherence, and phase lead were determined in all subjects for all interventions. Critical coherence was determined from the statistical significance of the squared coherence function and follows passively from the chosen settings for Fourier transform.³¹ A critical coherence value >0.5 was used as cut-off. Group averages and standard deviations per intervention were calculated. Data were tested for statistical significant differences with paired t-tests. We compared preoperative paced breathing interventions with spontaneous blood pressure oscillations and intraoperative paced breathing interventions with their preoperative counterpart. The primary outcome for this study was difference in CBFV_mean-to-MAP phase lead during paced breathing at 6 bpm. At this breathing frequency, BP oscillations around 0.1 Hz are imposed (Fig. 1f). A P-value <0.05 was considered to indicate a statistically significant difference. Earlier papers reported LF phase leads ranging from 52 (sd 10) to 61 (6)° for intact CA and from 26 (6) to 40 (6)° for impaired CA.^3,7 For the present study, we assumed that CA was intact before surgery, with an expected LF phase lead of 50 (10)°. If CA were to be impaired by sevoflurane anaesthesia, this would reduce below 40 (10)°. Assuming normality and a power of 90% at a significance level of 0.05, we need a minimal sample size of 13 patients to detect this difference in two-sided paired t-testing. To allow for drop-out, we included 18 patients.

Results

Haemodynamic and respiratory variables (Table 1)

Intraoperative BP was ∼25 mm Hg lower than the preoperative values. Likewise, heart rate was ∼7 beats min⁻¹ lower during the intraoperative period compared with the preoperative period. The CBFV and P_etco₂ were similar. Tidal volumes and peak airway pressures decreased with increasing breathing frequency.

Feasibility of the novel method (Fig. 2)

Spontaneous BP variations before surgery had a coherence >0.5, and the phase lead decreased with frequency (Fig. 2a–c). During surgery, LF oscillations in BP were not present; therefore, CBFV_mean-to-MAP phase lead could not be determined (Fig. 2d–f). During paced breathing, before and during surgery, BP oscillations could be imposed at each breathing frequency, resulting in high coherence and reliable estimation of CBFV_mean-to-MAP phase lead (Fig. 2g–l).

Cerebral autoregulation, preoperative vs intraoperative (Table 2)

Before surgery, LF oscillations in MAP as the input variable and CBFV_mean as the output variable were similar either during paced breathing at 15 bpm or during supine resting. Coherence remained >0.5, and calculated phase lead did not differ. During surgery, both MAP and CBFV_mean power disappeared, with low coherence [0.3 (0.2)] rendering quantification of CA no longer possible.

Paced breathing, preoperative vs intraoperative (Table 3)

During paced breathing, before and during surgery, MAP and CBFV_mean powers were adequate, with coherence >0.5 for all situations. Preoperative vs intraoperative CBFV_mean-to-MAP phase lead was 43 (9) vs 45 (8)°, 25(8) vs 24 (10)°, and 4 (6) vs −2 (12)° during 6, 10, and 15 bpm, respectively (all P=NS; Table 3 and Fig. 3).

Discussion

The main finding of this study is that intraoperative CA efficacy can be quantified from BP oscillations amplified by positive pressure ventilation. This is significant because with this technique CA assessment is possible during surgery without the necessity for administration of vasoactive medication or performance of elaborate manoeuvres.

Intraoperative monitoring of cerebral autoregulation is rarely used in clinical practice, because a standardized non-invasive method is lacking. However, patients with compromised autoregulation^3–7 undergo surgery on a daily basis and might be at increased risk for cerebral hypoperfusion and severe complications, including perioperative stroke.^8–10 Although the overall incidence of perioperative stroke is <1% in large series,^32,33 brain magnetic resonance imaging studies suggest that 1 in 10 elderly patients experiences a (subclinical) covert perioperative stroke.³⁴ We have proposed a novel and reliable method to quantify the efficacy of CA and monitor cerebral perfusion with a simple clinical test that could easily be used before and during surgery (i.e. paced breathing). Results from these measurements may be an important step towards guidance for anaesthetists in defining the proper haemodynamic goals for their patients and optimize cerebral perfusion during surgery.

Another method to assess CA during surgery is continuous moving linear regression between BP and CBFV, or the ‘Mx’ method.³⁵ Indeed, much of the insight into intraoperative CA has been gained from this technique, and it has been tested successfully against other indices of autoregulation.³⁶ Still, it offers a semi-quantitative measure at best, and cut-off values for impaired vs intact CA vary in different studies.^35–37 In the present study, we were able to reconstruct CA as a high-pass filter and offer real quantitative information on the efficacy of CA.

In this study, sevoflurane in concentrations <1.2 MAC did not interfere with CA efficacy. This finding confirms our hypothesis and was also found in earlier studies.^15–17 In contrast, one study reported that sevoflurane attenuated cerebral autoregulation.²³ This might be because the fact was not fully appreciated that sevoflurane suppresses sympathetic outflow²¹ and thus spontaneous blood pressure variations during anaesthesia. Therefore, analysis was done on blood pressure variations with very low amplitude, most probably noise. This is further supported by the low coherence between blood pressure and cerebral blood flow velocity reported in that study. Hence, the conclusion that autoregulation was impaired by sevoflurane was imprudent and probably erroneous. In the present study, we have increased the amplitude of blood pressure oscillations by using three different breathing frequencies and were able to report adequate coherence both before and during surgery.

Limitations

First, this study was conducted exclusively with non-invasive measurement techniques. We assumed that BP measurement using ^ccNexfin resembles invasive arterial blood pressure. This is confirmed in extensive validation studies of ^ccNexfin vs intra-arterial blood pressure in anaesthetized patients,^26,27 but validation in awake patients is lacking. One study compared CA determined from both invasive and non-invasive blood pressure and CO₂ measurement techniques and found similar values.³⁸ Additionally, CBFV remains a surrogate for CBF and depends on the assumption that the diameter of the middle cerebral artery is unchanged in physiological circumstances. Validation studies have shown that flow velocity changes accurately reflect flow changes during hypercapnia.^39,40

Second, preoperative measurements were performed on a different occasion from when the operation took place, albeit during the same admission (usually the evening before surgery). This could be a confounder because circadian variations in CA efficacy are known,^41,42 and also Doppler flow probe placement may have been slightly different during surgery compared with before surgery. Nonetheless, studies have shown that the repeatability of autoregulation assessment on different days in the same patient is reliable when using either spontaneous blood pressure oscillations⁴³ or paced breathing.⁴⁴

Third, blood pressure oscillations during paced breathing preoperative vs intraoperative resulted from negative vs positive pressure ventilation, respectively. Although the frequency of these oscillations is identical, the underlying shifts in cardiac output and vascular resistance are not. Also, P_etco₂ might be different during paced breathing because of an unnatural breathing frequency,⁴⁵ and this might influence autoregulation determination.^12,46 Therefore, we instructed patients to adjust their inspiration depth in order to maintain CO₂ concentrations constant and found almost identical values before and during surgery. Decreased tidal volumes during higher breathing frequencies resulted in decreased power (amplitude) of the induced blood pressure oscillations and vice versa. We think that this has not influenced our results, because coherence between MAP and CBFV during all interventions was well above the predefined critical value of 0.5. This indicates that the amplitude of the blood pressure oscillations was sufficiently high to be able to calculate phase lead and gain accurately.

Fourth, during anaesthesia, apart from anaesthetic drugs, other medication was used that could affect the cerebral circulation. Most importantly, this includes vasopressor agents. Boluses of vasopressor agents, such as the α₁-adrenergic agent phenylephrine, may induce cerebral vasoconstriction by either an indirect or a direct mechanism.⁴⁷ For obvious reasons, omission of vasopressor agents as part of the study procedures could not be justified. To avoid confounding, no paced breathing interventions were performed during haemodynamically unstable situations that required bolus vasopressor therapy.

Lastly, we determined the efficacy of CA in relatively healthy subjects without cerebrovascular co-morbidity or generalized vessel disease and found a slightly lower than expected mean phase lead. Looking at the haemodynamics, values for CBFV were similar before and during surgery, although blood pressure was significantly lower. Although the study was not designed is this way, this could indicate that patients in our study group were on the ‘plateau’ part of the autoregulation curve.

Conclusion

During surgery, CA can be quantified reliably from blood pressure oscillations amplified by positive pressure ventilation. Sevoflurane anaesthesia did not affect CA efficacy.

Authors’ contributions

Study conception and design: N.H.S.W., J.H., B.P., W.J.S., J.J.L., R.V.I.

Data acquisition and analysis: N.H.S.W.

Interpretation of data: M.W.H.

Drafting the manuscript: N.H.S.W.

Revising the manuscript critically, final approval of the version to be published, and agreement to be accountable for all aspects of the work: N.H.S.W., J.H., M.W.H., B.P., W.J.S., J.J.L., R.V.I.

Declaration of interest

J.H. received Study funding (NovoNordisk) and a lecture fee (Eli Lilly). M.W.H. provided consultancy for ECHO BV and Eurocept BV and received lecture honoraria for Baxter, BBraun, Fresenius, CSL Behring, Edwards; M.W.H. is also Executive Section Editor Pharmacology with Anesthesia & Analgesia and Section Editor Pro Con Sessions with the Dutch Journal of Anesthesia (NTvA). B.P. has received 60.000 Euro from GE Healthcare, 15.000 Euro from Edwards Life Science and 40.000 Euro from Air Liquide France for contracted research projects; B.P. has received lecture fees from Abbott, Abbvie Netherlands; Orion Pharma BVBA Netherlands, Philips Healthcare Germany, and grants from SCA, ESA, ZonMW; B.P. is Editor of the Netherlands Journal for Anaesthesiology. All other authors have no conflict of interest to declare.

Funding

European Society of Anaesthesiology (https://www.esahq.org/research/esa-research-grants/about-esa-grants/results/ accessed 15 September 2017).

References