LBNP experiment

Subjects were placed on the experimental table in a supine position and received peripheral venous cannulation. Monitoring was attached consisting of ECG, peripheral oxygen saturation, continuous blood pressure, stroke volume (SV) and cardiac output (CO) monitoring via fingercuff (Finapres NOVA, Finapres Medical Systems, Enschede, Netherlands), cerebral Near Infrared Spectroscopy (NIRS, NIRO 200NX, Hamamatsu Photonics, Herrsching, Germany), and right temporal transcranial doppler sonography (TCD) via robotic probe (Delica 9UA, Shenzen Medical, Shenzen, China). Their analog output signals were synchronously recorded at 500 Hz by using an analog-digital-converter and then stored on a laptop using the Windaq data acquisition software (Dataq Instruments, Akron, OH, USA). A custom-made airtight chamber was fastened above the lower body of the subject. A Neoprene skirt attached to the chamber and fastened around the waist of the subject prevented air from leaking into the pressure chamber. The lower body chamber was connected to a calibrated air pump that produced a preset pressure which it adjusted automatically according to pressure feedback from a sensor installed inside the chamber. Once all signals were stable, a 10-minute baseline measurement was initiated. Subsequently, lower body pressure was decreased to -15 mmHg for a time of 5 minutes. Pressure was then reduced to -30 mmHg for another 5 minutes. LBNP was further decreased in -5 mmHg steps every 5 minutes. The maximum pressure was maintained until subjects experienced symptoms of presyncope indicated by dizziness, nausea and a sudden loss of blood pressure or cardiac output. At presyncope, LBNP was immediately turned off and the subject was tilted into Trendelenburg position.

LBPP experiment and hypoxia

For the experiment with positive pressure and hypoxia, subjects were placed on the experimental table as mentioned above. Monitoring was identical to the LBNP experiment. Additionally, subjects received a face mask for in- and expiratory O₂ and etCO₂ measurements which were acquired with an Innocor gas analyzer (Innovision, Glamsbjerg, Denmark). Once stable signals were achieved on all channels, a baseline measurement of 10 minutes was started. LBPP was then increased in 5 mmHg steps every 5 minutes until the last pressure stage where steady LBPP could be maintained (20 or 25 mmHg), then hypoxia was established with LBPP maintained. Hypoxia was achieved with a gas blender attached to pressurized air and pressurized nitrogen. The gas mixture accumulated in a reservoir bag was subsequently inhaled by the subjects. Hypoxia was reached with an SpO₂ of 84%. The experiment was continued for 10 minutes or until abort by the subject. Termination of the experiment was done by turning of the LBPP and detaching the face mask.

Data acquisition and—analysis

Data acquisition for subsequent off-line beat-to beat analysis was performed with the Windaq system at a sample rate of 500 Hz (Dataq Instruments, Akron, USA). QRS complexes from the ECG, heart rate, systolic-, mean-, and diastolic finger blood pressure were automatically detected and marked based on a custom-written software (“Physiowave”) in PV-Wave language (PV-Wave, Visual Numerics, Boulder, USA). Data were manually checked by an experienced observer for wrong or missing values. Further data processing was performed with Matlab (Matlab R2020a, MathWorks Inc. Natick, USA). Cerebral autoregulation indices were calculated using a version of the moving average (Mx) algorithm [3, 6, 10]: A moving Pearson correlation coefficient between mean arterial pressure (MAP) and middle cerebral artery mean flow velocity (MFV, Mx) or MAP and cerebral tissue oxygenation index (TOI, ORx) over 30 ten second averages was used to indicate the state of autoregulation. The resulting Mx and ORx indices are dimensionless values between 0 and 1 representing the degree of linear relationship between MAP and MFV or MAP and TOI, respectively. Increased indices reflect a more pressure passive nature of the underlying TCD or NIRS values. A Pearson coefficient of 0.3 and above was interpreted as disturbed autoregulation for Mx and ORx based on previous reports [6, 11, 12]. Values for cerebral and global hemodynamic variables as well as autoregulation parameters at the first 10 seconds of the last pressure level in LBNP, 5 seconds before and after presyncope as well as at the last pressure level in LBPP and during hypoxia were compared to baseline measurements (mean values over 10 minutes).

Statistics

Statistical calculations were performed with SigmaStat 14.0 (Systat Software, San Jose, USA) Comparisons between parameters at baseline, the last pressure level on LBNP/LBPP, and presyncope/hypoxia were done using the One Way Repeated Measured Analysis of Variance (RM ANOVA). A Shapiro-Wilk test was used to test for normal distribution of parameters. Post-hoc-t-tests for multiple pairwise comparisons were performed. P values were automatically corrected for multiple comparisons with Bonferroni correction. Mx and ORx were compared using linear regression analysis. P-values of <0.05 were considered significant.

LBNP and the limits of cerebral hypoperfusion

The results presented here show that cerebral hypoperfusion begins at a reduction of 35% in MFV of the middle cerebral artery as measured by TCD and a reduction of 5.6% in cerebral tissue oxygen saturation as measured by NIRS. This is in contrast to current accepted limits of the respective methods. A higher volatility of MFV in comparison to cerebral saturation has been described previously on several occasions [13, 14]. Currently, widely accepted thresholds for a reduction in cerebral saturation is a decrease of more than 20% or below 50% absolute value and for TCD a reduction of more than 50% in MFV [15]. It has to be noted that differences in cerebral saturation between NIRS devices from different manufacturers exist [16]. Although these thresholds are mostly based on patients during general anesthesia, there is data on awake patients undergoing carotid endarterectomy, reporting cutoff values of 48% reduction in MFV and 20% reduction in NIRS for development of cerebral ischemia [17]. While one still has to consider the differences to our healthy subjects and the difference in outcome, we suggest a reevaluation of critical limits of both NIRS and TCD. Our results also highlight that in our patients, hypoxia had a larger effect on cerebral saturation than had hypoperfusion, again pointing out that several factors contribute to cerebral oxygen saturation.

In our subjects, both MAP and CO parameters such as SV appear to be involved in the development of cerebral hypoperfusion. While SV was significantly decreased at the last LBNP level this was not reflected by CO. In both MAP and CO, a significant reduction only occurred with presyncope which also reflects cerebral parameters. We interpret this as active regulation mechanisms which compensate decreasing SV before the onset of presyncope which is marked by a decrease in CO which also translates to cerebral parameters. Neumann et al. reported a dependence of cerebral blood flow on CO rather than MAP in hypertensive patients during LBNP [18]. A combined MAP and CO dependence of carotid blood flow has also been shown by Skytioti et al. in surgical patients [19]. The same group have demonstrated an association between CO and carotid blood flow in healthy subjects undergoing LBNP, while MAP remained practically unchanged. Skytioti’s LBNP experiment consisted of intermittent LBNP stages of -30 mmHg in contrast to incremental LBNP in our setup that led to presyncope making the results difficult to compare [20].

While the exact threshold for impaired autoregulation remains unclear and is reported to be at index values of 0.3 to 0.5, a threshold of 0.3 has mostly been used recently [6, 11, 21]. Using this threshold, both Mx and ORx adequately represent impairment in autoregulation during presyncope. However, while there was a large increase in ORx, the difference in Mx compared to baseline was less distinct, since Mx was already elevated above 0.3 at baseline. One has to keep in mind the different measurement modalities of TCD and NIRS, the one representing exclusively arterial flow while the other is a mostly venous signal [13, 14]. Although the meaning of an impaired TCD-derived autoregulation at baseline is unclear at this point, we have to assume that this is a physiological phenomenon in our cohort of healthy subjects. Therefore, these results question the ability of the Mx to discriminate between intact and impaired autoregulation in contrast to the ORx. The use of a higher Mx threshold of 0.5 would not have changed this finding [3]. One may suspect from these results that TCD-derived indices are more sensitive, as we have also suspected previously in a clinical context [14]. Our findings may also confirm the hypothesis that the plateau region on the autoregulation curve is not as wide as previously assumed and may be missed by some measurement modalities, or alternatively may not be present at all [8]. At the beginning of the last pressure level at which presyncope occurred we did not observe significant changes in Mx or ORx, while both were below the 0.3 threshold. It seems that cerebral autoregulation was still intact at this point even with imminent presyncope. We suspect that in our collective of young healthy subjects, regulatory mechanisms may be active until the final collapse of hemodynamics in presyncope which is only then reflected by increased Mx and ORx. We also suspect differences to older, morbid patients or patients during general anesthesia. However, more data is needed to investigate the relationship between autoregulation indices, age and medical history.

LBPP/hypoxia

While LBNP has been studied extensively, the same does not apply to LBPP. We observed a decrease in MFV while TOI remained unchanged at the last pressure level of LBPP. CO and SV decreased while there was a substantial increase in MAP. These hemodynamic findings are in line with previous head-down tilt studies representing the same physiological mechanism [22]. Perry et al. reported unchanged cardiac output, but increased MCA flow velocity in moderate LBPP of 20 mmHg, but unchanged MFV in high LBPP of 40 mmHg which is explained by alterations in baroreflex sensitivity [23]. Despite lower LBPP levels in our subjects, our findings are still in line with Perry et al, although we would explain the decrease in MFV rather mechanistically with the higher postcapillary resistance once a level of stressed volume of the venous system is reached. Mx and ORx calculations did not indicate impaired autoregulation during LBPP, nor was this the case for hypoxia, despite a significant increase in MAP and increased CO during hypoxia. This is somewhat surprising since impaired autoregulation has been reported to occur in acute hypoxia on several occasions [9, 24–26]. Whether the hypoxia stimulus was insufficient to cause impaired autoregulation or indices did not adequately show impaired autoregulation is unclear. TCD values appeared unaffected by LBPP or hypoxia which may indicate cerebral hemodynamic changes in hypoxia were not sufficient to cause autoregulation impairment. However, hypoxia acts as a powerful vasodilatator which has been shown to alter TCD readings [27, 28]. Hypocapnia may also play a role through hyperventilation [29]. TOI naturally showed a substantial decrease during hypoxia. This is important for clinical practice since changes in oxygen saturation may be more pronounced in NIRS than changes in cerebral circulation. In NIRS, cerebral malperfusion may thus be masked by changes in oxygen settings. Therefore, with the results presented here we cannot confirm the use of Mx or ORx for the determination of the upper autoregulation limit. For future studies we suggest different methods to reach the upper autoregulation limit, such as exhaustive exercise or a combination of exercise and LBPP [30, 31].

Agreement between autoregulation indices

Both Mx and ORx represent autoregulation impairment in presyncope using a threshold of 0.3 to 0.5, although the impairment is more distinct in ORx. As shown in Fig 3, Mx and ORx show poor agreement throughout the LBNP and LBPP experiment, excluding hypoxia. It appears that Lassen’s curve and its relationship to global hemodynamic parameters depends on the location and modality of measurement. What can be stated at this point is that Mx and ORx are not interchangeable and measurements from different modalities may only be compared with caution. This is in contrast to a report by Steiner et al. on intensive care patients with sepsis, where an almost perfect agreement between the two methods has been shown [32]. The difference in our findings may be caused by awake subjects with spontaneous breathing unlike Steiner’s patients or the underlying disease which may be associated in compromised regulation mechanisms of the cerebral vasculature. Cortical activation in awake subjects may further influence results, while cortical activity should remain relatively unchanged in sedated patients. While the exact cause of this disparity remains unclear at this point, it can be stated that Mx and ORx methods yield different results in awake subjects. However, further validation studies are necessary to confirm these results.

Limitations

Several limitations are associated with this study. First, this is a pilot study with few subjects, therefore it is difficult to generalize the presented results. Second, NIRS and TCD both represent surrogate parameters for cerebral blood flow which both have inherent technical limitations affecting measurement precision and accuracy such as an unknown vascular diameter and dependence on the insonification angle in TCD and for NIRS the possibility of extracranial contamination of the signal. The same is the case for non-invasive CO and MAP measurements which may present inaccuracies, especially in the more extreme ranges.