Non-pulsatile blood flow is associated with enhanced cerebrovascular carbon dioxide reactivity and an attenuated relationship between cerebral blood flow and regional brain oxygenation

In the field of cardiac surgery and at many ICUs and emergency departments of tertiary care centres, it is daily practice to subject patients to a non-physiological condition where systemic blood flow either completely lacks pulsatility or pulsatility is greatly diminished [1]. Non-pulsatile flow, however, has been linked to increased vascular resistance, raised muscular sympathetic nervous activity, decreased oxygen consumption, endothelial dysfunction with loss of endothelial integrity and deterioration of microvascular perfusion. All of these factors supposedly contribute to increased incidences of gastrointestinal and intracerebral bleeding, strokes, and increased mortality on veno-arterial extracorporeal membrane oxygenation (VA-ECMO) [2‐8].

In rats, endothelial-dependent vascular reactivity is dampened in less than 30 min on non-pulsatile cardiopulmonary bypass (CPB) [9]. Furthermore, within such a short period of absent pulsatility, heterogenous flow in human capillaries ensues. It is marked by vessels with absent being closely adjacent to those with very fast perfusion [2]. On the other hand, pulsatile blood flow is associated with improved cerebral microvascular circulation, a more homogenous tissue perfusion, an increased release of nitric oxide (NO) by the vascular endothelium and a higher ATP content in vital organs [4, 10]. Pulsatile CPB flow proved to be advantageous after acute cerebral ischaemia in dogs, in the course of conventional heart surgery as well as in surgery requiring deep hypothermic cardiac arrest [11‐13].

Neurological complications such as stroke, bleeding, cognitive dysfunction and delirium still remain a major concern in patients on mechanical circulatory support systems that provide complete non-pulsatile or minimal pulsatile flow (e.g. CPB, VA-ECMO) or continuous flow left ventricular assist devices (CF LVAD) [14‐17]. Flow-induced tissue malperfusion, for example, could therefore potentially further augment ischaemia/reperfusion injury in patients in the course of extracorporeal cardiopulmonary resuscitation (ECPR) and loss of pulsatile cerebral perfusion might contribute to worse neurological outcome [4, 7, 18]. In fact, perfused vessel density, an established microcirculatory parameter, was negatively correlated with survival in patients requiring VA-ECMO due to cardiogenic shock [8].

There are a number of factors that determine cerebrovascular myogenic tone and consequently gross cerebral blood flow (CBF) and its local distribution. CBF is regulated by the metabolic demand of the brain, by cerebral perfusion pressure that underlies autoregulation, by the autonomic nervous system, and by hypoxaemia. Arterial partial pressure of carbon dioxide (PaCO₂) is another pivotal factor. The final effect of each of these CBF modulators results from the intricate interplay with other factors. For a detailed overview on this topic, we would like to refer the reader to recently published in-depth reviews [19, 20]. In brief, changes in PaCO₂ alter the diameter of cerebral arterioles by ways similar to those operative during changes of pulsatility thereby affecting nutritive blood flow through the capillary bed [10, 21]. Cerebrovascular CO₂ reactivity (CVR) reflects this ability of the cerebral arterioles to modify CBF by dilating or constricting in response to changes in PaCO₂ [22].

In instances when there is dual CO₂ removal (e.g. via the oxygenator on partial cardiopulmonary bypass or ECMO as well as via the respirator), extremes in PaCO₂ levels are not infrequent. Interestingly in this context, rapid decreases in PaCO₂ immediately after onset of VA-ECMO treatment were significantly associated with intracerebral bleeds [23]. Although cerebral autoregulation is frequently severely compromised after cardiopulmonary resuscitation (CPR) [24], CVR is still intact and hypocapnia can easily set in especially when shock and therapeutic hypothermia have already diminished CO₂ production. Hypocapnia post CPR, however, has been associated with poor neurological outcome, at least when it occurred within the first few hours after CPR, which unfortunately is not uncommon [25, 26].

We hypothesised that CO₂-driven regulation of cerebral oxygenation during non-pulsatile blood flow is altered. Therefore, our objective was to compare the effect of non-physiological non-pulsatile blood flow on CVR and rSO₂.

Patient characteristics and type of surgery is depicted in Table 1. For each condition (i.e. pulsatile or non-pulsatile flow), MAP, haemoglobin, BIS, and body temperature did not change over the measurement period. MAP and BIS values, however, were significantly lower on CPB (63 ± 8 vs. 73 ± 10 mmHg and 40 ± 6 vs. 48 ± 7, p < 0.0001) while PaO₂ was significantly higher (175 ± 56 vs. 135 ± 32 mmHg, p < 0.05). Isovolaemic haemodilution resulted in a decreased haemoglobin level on CPB (8.4 ± 1.1 g/dL vs. 11.1 ± 1.1 g/dL, p < 0.0001). Irrespective of flow mode, we neither observed increases in MAP during hypercapnia (Table 2) nor instances of hypoxaemia.

Absolute values of rSO₂, arterial haemoglobin oxygen saturation, PaO₂, patient temperature, MAP, and, where available, intra- and postoperatively determined cardiac indeces after CO₂ step changes during non-pulsatile and pulsatile flow are listed in Table 2. The course of absolute MCAv during alterations of PaCO₂ within each flow mode is depicted in Fig. 1. MCAv was significantly lower during hypocapnia compared to normocapnia (p < 0.0001), while MCAv was significantly higher during hypercapnia compared to normocapnia (p < 0.0001). There was also a statistically significant difference between flow modes (p < 0.033). Relative MCAv changes of individual patients clearly show that greater shifts from normocapnia occurred when flow was non-pulsatile (Fig. 2, p < 0.05). The MCAv/PaCO₂ slope was 14.4 cm/s/mmHg [CI 11.8–16.9], during non-pulsatile and 10.4 cm/s/mmHg [CI 7.9–13.0] during pulsatile blood flow (p = 0.032).

Under hypocapnia and non-pulsatile flow, CVR was significantly increased (4.3 ± 1.7%/mmHg) compared to pulsatile flow (3.1 ± 1.3%/mmHg, p = 0.017). During hypercapnia, however, non-pulsatile CVR was only nominally but not statistically greater in relation to pulsatile CVR (3.5 ± 1.8%/mmHg vs. 2.6 ± 1.2%/mmHg, p = 0.197, Fig. 3).

Independent of the flow mode, we saw a decline in rSO₂ during hypocapnia and a corresponding rise during hypercapnia (p < 0.0001). However, despite more pronounced CO₂-driven flow velocity changes during non-pulsatile flow, there was no equivalent difference in ΔrSO₂ between corresponding PaCO2 levels when compared to pulsatile flow. The rSO₂/PaCO₂ slope was 2.8%/mmHg [CI 1.5–4.0] on CPB and 3.1%/mmHg [CI 1.9–4.3], postoperatively (p = 0.72). As shown in Fig. 4, ΔMCAv and ΔrSO₂ changes were closely related to each other during non-pulsatile (r = 0.724, p < 0.001) and pulsatile blood flow (r = 0.796, p < 0.001). However, during non-pulsatile flow, the slope of the regression line was less steep and cerebral CVR more accentuated as compared to pulsatile perfusion.

In the present study, short-term non-pulsatile perfusion was associated with significantly enhanced cerebrovascular CVR compared to pulsatile blood flow resulting in exaggerated responses in CBF when non-physiologically low PaCO₂ values were reached. These changes in CBF, however, were not accompanied by equivalent changes in rSO₂.

Instances of hypocapnia are not infrequent in low cardiac output and INTERMACS 1 or 2 patients on extracorporeal assist devices with or without dual CO₂ removal (e.g. VA-ECMO, extracorporeal cardiopulmonary resuscitation (ECPR) or CF LVAD) as well as after conventional CPR [26]. They have mostly been associated with poor neurological outcome. Prolonged intraoperative hypocapnia by itself has recently also been associated with postoperative delirium in non-cardiac surgery [29]. Our observations indicate that the effect of hypocapnia-induced cerebral vasoconstriction may even be more extensive in the presence of non-pulsatile cerebral blood flow. In contrast, the impact of hypercapnia on vasodilation was not significantly different from that seen under pulsatile flow conditions in the current study.

To guarantee propagation of pulsatility into the cerebral circulation, we compared CVR during non-pulsatile blood flow on complete CPB with the same patient’s CVR postoperatively on the ICU during pulsatile blood flow generated by the beating heart itself. In previous studies investigating induced pulsatile flow on CPB, it has never been ruled out that it was simply blunted pulsatility (i.e. with reduced systolic and pulse pressures) that was created, which lacks excess energy [30]. Moreover, even in the paediatric population, not all available pulsatile pumps generate this higher haemodynamic energy when compared with non-pulsatile pumps [31]. Mitigated pulsatility, however, which is frequently observed in ECMO patients early after implant, appears to be similar to non-pulsatile flow as it also increases muscular sympathetic nerve activity [32]. Furthermore, long duration of diastole and low levels of diastolic pressure elicited sympathetic bursts in CF LVAD as well as in VA-ECMO patients with minimal pulsatility [32]. Interestingly in this context, 2 h after exposure to VA-ECMO with a high flow rate (120–150 mL/kg/min), decreased NO production was noted in cerebral arteries of lambs indicating impaired cerebral vascular endothelial function [33]. This phenomenon likely contributes to capillary collapse and shunting when a critical closing pressure cannot be reached whereby the driving pressure is usually created by the excess energy generated through the ejection of blood from the left ventricle into the aorta [30]. O’Neil and co-workers studied sublingual microcirculation on CPB, yet without specifically looking into PaCO₂-generated vasoreactivity and found that the proportion of normally perfused microvessels decreases immediately with onset of non-pulsatile flow [2]. In a similar study in 20 patients undergoing cardiac surgery, the same researchers reported an increase in vessels with hyperdynamic flow when blood flow was non-pulsatile, which may represent capillary shunting [5]. These results could account for our observation that despite greater ΔMCAv during non-pulsatile flow, ΔrSO₂ did not change proportionally although a close relationship still persisted. Conversely, this also implies that potential benefits of pulsatile flow related to improved microcirculatory perfusion may not be traceable by higher rSO₂ values [34]. The dependence of brain tissue saturation on PCO₂ has previously been confirmed in healthy individuals and patients undergoing cardiac surgery [35]. In addition, during carotid endarterectomy, rSO₂ closely correlated with MCAv during cross clamping of the ipsilateral common carotid artery [36]. These findings are in line with our results as rSO₂ and PaCO₂ changed in the same direction. rSO₂ further changed in parallel to MCAv. However, there appears to be some uncoupling during non-pulsatile flow with enhanced reductions in cerebral blood flow during hypocapnia.

Findings from previous studies in this area can barely be compared with our results since investigations were then usually conducted with only few patients assigned to either of the two flow modes, during moderate hypothermic CPB, with different anaesthetics (e.g. N₂O), and partly at very low cerebral perfusion pressures [37]. In our trial, the 31 patients served as their own controls and only few patients were operated under mild hypothermia. In addition, several studies have shown that CVR is maintained with both anaesthetics we employed when given within normal clinical ranges [38] as was the case here, despite the fact that anaesthesia likely compromises basal vascular tone [39]. In this trial, end-tidal sevoflurane concentrations ranging between 0.3 and 1.5 MAC (mean dose 0.8 MAC) was administered intra- and low-dose propofol (median 2.5 [range 1.8–3.9] mg/kg/h) postoperatively.

Although changes in PaCO₂ alter the diameter of cerebral arterioles and thus determine nutritive microcirculatory blood flow through the capillary bed, the large cerebral arteries are frequently considered to be noncompliant serving as a conduit to distribute and store blood volume [27]. In young, healthy, and spontaneously breathing volunteers, however, luminal changes of up to 10% have been calculated during hypo- and hypercapnia when ΔPetCO₂ exceeded 10 mmHg [40]. It is unclear if similar changes also occur in sick, anaesthetised and ventilated patients undergoing cardiac surgery. As we did not measure the diameter of the interrogated vessel, we cannot rule out that variations in the vessel calibre impacted on the measured MCAv whereby particularly MCAv values during hypercapnia would be affected. The modest hyperoxia during non-pulsatile flow on CPB should rather decrease the slope of the curve describing the relationship between PaCO₂ and MCAv, which would be just the opposite we found [41]. As mentioned above, sevoflurane anaesthesia that we used on CPB has been demonstrated to preserve CVR (4.1 ± 0.7%/mmHg PaCO₂) as well as MCAv in healthy patients undergoing orthopaedic surgery when MAP was kept at 80 mmHg with phenylephrine [42]. MAP in our patients was generally < 80 mmHg and showed minor though statistically significant differences between pulsatile and non-pulsatile flow. In this regard, it should, however, be kept in mind that current research confirmed interactions between CVR and cerebrovascular autoregulation, primarily affecting the upper threshold [43, 44]. Nevertheless, MAPs we recorded were all within the critical PaCO₂-dependent autoregulatory thresholds that have been reported during anaesthesia [37, 45]. In our trial, cerebral perfusion pressures thus should not have had an effect on the measured MCAv. Neither during non-pulsatile nor during pulsatile flow did we observe rises in MAP accompanying periods of hypercapnia. A potential explanation may be the suppression of sympathetic activity during anaesthesia. Interestingly in this context, as CVR is also mitigated during sympathetic blockade [46, 47], increased sympathetic activity in turn, which has been confirmed during non-pulsatile blood flow [48], might augment CVR, which is exactly what we have seen. It is, therefore, probably not pulsation per se that accounts for the observed changes in the PaCO₂-dependent regulation of cerebral artery myogenic tone [44] but secondary effects induced by pulsatile blood flow like for example eNOs activity and coupling [49, 50] or modulations of the autonomic nervous system output [51]. Furthermore, a reduced myogenic tone has been observed in isolated murine cerebral arteries when subjected to static pressure, which may also account for the accentuated CVR we determined [50].

This trial still does have some limitations primarily related to the chosen protocol. Use of CPB always causes some degree of isovolaemic haemodilution that may impact blood viscosity and eventually microcirculatory perfusion. Haemodilution might thereby affect the magnitude of CVR. Data from the literature, however, is inconclusive. At haematocrits < 28%, both attenuated and enhanced CVR has been reported in patients undergoing heart surgery during hypocapnia, while no alteration was seen during hypercapnia [52, 53]. However, isovolaemic haemodilution by low-molecular-weight dextran infusion, for example, did not increase but instead decreased acetazolamide-induced cerebral vasoreactivity [54]. In addition, CBF is augmented when arterial oxygen content drops, which, however, should only have a marginal effect considering the actual haemoglobin levels in our patients during and after surgery [55]. Nevertheless, the enhanced CVR we observed under non-pulsatile flow may be related to some extent to the lower haematocrits on CPB.

Another issue that needs to be addressed is the different types of anaesthetics that were employed here as their impact on the BIS level, CBF, and CMRO₂ may not be the same. Sevoflurane was administered intraoperatively due to its ascribed cardioprotective properties but could not be continued into the postoperative period as we lack the technology that allows its administration in the ICU. Data from eight young and healthy volunteers indicate that 1.5 vol% sevoflurane reduces regional CBF less than propofol (propofol plasma concentration ranged between 2.6 and 4.6 μg/mL) whereby CMRO₂ was equally decreased and MAP and BIS depressed to a similar degree [56]. In relation, the propofol dose our patients received was much less due to the concomitant administration of remifentanil, which may explain the greater neuronal activity in the frontal cortex evidenced by higher BIS levels postoperatively. Therefore, one should not expect to find a significant propofol-induced depression of global CBF although we cannot comment on the eventual regional distribution of CBF. In this context, alterations in the cortical CBF/CMRO₂ ratio have only been observed at higher propofol concentrations (i.e. 7.8 ± 2.1 mg/kg/h) that were also accompanied by a BIS level as low as 20, which makes it unlikely that propofol in this study affected postoperative rSO₂ [57]. As a matter of fact, mean MCAv under 6 mg/kg/h propofol anaesthesia (a higher dose than we employed) observed in ASA 1 patients was by trend somewhat higher but comparable with MCAv of our patients in the ICU (41–49–58 cm/s for hypo-, normo-, and hypercapnia). In addition, relative CVR was preserved in this investigation [58]. In comparison, equivalent MCAv measured again in young and healthy ASA 1 patients, however, under 1.2 and 1.5 MAC sevoflurane anaesthesia were 25/33–47/55–76/78 cm/s [59, 60], which are strikingly similar to those obtained in a spontaneously breathing non-anaesthetized young volunteer (30–50–70 cm/s) [41]. Obviously, mean absolute MCAv can vary to some extent. Hence, relative CVR seems to be a more appropriate indicator of the effect of PaCO₂ alterations on CBF as absolute measures.

Despite the mentioned drawbacks of our protocol, it does have advantages, too. Because we did not compare different groups of patients subjected to different flow modes, we could reduce differences in ΔMCAv resulting from other potential confounders (e.g. diabetes mellitus, arterial hypertension, age, atherosclerosis of the cerebral vessels, previously unrecognised anatomical abnormalities). Furthermore, it seems improbable that CVR was affected by mild hypothermia with α-stat pH management that was instituted in few patients on CPB [52]. Moreover, none of the patients had developed a fever when the second measurement was carried out in the ICU. Additionally, catecholamines as well as levosimendan were administered at low dosages that should not affect CVR under anaesthesia to a meaningful extent [61, 62]. They were frequently already started intraoperatively and continued after surgery.

In conclusion, in older anaesthetized cardiac surgery patients, non-pulsatile blood flow caused enhanced CVR. If, however, unphysiological PaCO₂ values in conjunction with this flow mode actually increase the odds of undesired cerebral complications, as it was the case after traumatic brain injury, requires further investigations [63]. Especially with a leftward-shifted oxygen-haemoglobin dissociation curve and a heterogenous microvascular perfusion, rSO₂ recordings may additionally overestimate true brain tissue oxygenation during hypocapnia when CBF becomes non-pulsatile. This would further diminish the prognostic value of rSO₂ post ECPR regarding neurological outcome [64].

As even short-term non-pulsatile flow conditions rapidly affect the cerebral circulation, these findings could be especially relevant in patients requiring extracorporeal mechanical assist devices that do generate either no or merely blunted pulsatile brain perfusion. Patients with low left ventricular ejection fraction receiving emergent VA-ECMO and depicting low-pulsatile systemic perfusion could potentially be particularly vulnerable regarding cerebral insults during hypothermic and hypocapnic states [3, 65, 66]. The currently poor neurological outcome after cardiac arrest with or without ECPR is often clinically initiated by convulsions and myoclonus. In this context, hypocapnia-induced vasoconstriction leading to tissue alkalosis and hypoxaemia might also play a certain role in the elicitation of seizures in susceptible patients immediately after global brain ischaemia [67‐69].