Hemodynamic effects of short-term hyperoxia after coronary artery bypass grafting

During and after coronary artery bypass grafting (CABG), patients are supported with mechanical ventilation and supplemental oxygen. Despite its lifesaving characteristics and key role in the treatment of vasodilatory shock, oxygen therapy may harbor considerable risks given the relationship between prolonged hyperoxia, lung injury and adverse outcome [1–5]. The effects of supplemental oxygen may be even more pertinent during cardiovascular events considering the direct effects of high arterial oxygen concentrations on the vascular tone [6, 7]. Arterial hyperoxia has the potential to alter hemodynamics and has been associated with adverse outcomes and mortality after cardiac arrest, myocardial infarction, stroke, brain injury and during mechanical ventilation [8–14], yet not during cardiopulmonary bypass [15, 16]. It is well established that high oxygen concentrations induce vasoconstriction and increase the resistance of the systemic circulation. However, the effects on vital parameters may be diverse and the venous and arterial aspects of the circulation have not been clearly distinguished in previous studies. Furthermore, the microcirculation may react differently than systemic hemodynamics [17].

Achieving hemodynamic stabilization is an important clinical prerequisite for early extubation and dismissal from the intensive care unit (ICU) after cardiothoracic surgery. Any intervention that influences hemodynamics and blood flow to the bypassed myocardial territories may impact functional recovery and requires optimal fine-tuning to achieve the best outcome. Although oxygen is generally administered in a liberal manner in the perioperative setting, the unraveling of the effects of oxygen administration on dynamic cardiovascular parameters, filling status and cerebral perfusion may provide novel insights in the pathophysiological mechanisms involved in hyperoxic exposure. Our aim was to study the acute hemodynamic and microcirculatory changes during increased oxygen supply in mechanically ventilated ICU patients after CABG surgery.

In this single-center physiological intervention study, we found that a 15-min exposure to hyperoxia affects hemodynamics in ICU patients after CABG. This was translated in several changes in central circulatory variables and in the percentage of perfused microcirculatory vessels, but showed no alterations in cardiac output and cerebral blood flow.

The circulation of blood can be described by either the CO or the venous return. As only blood returning to the heart can be pumped out in the systemic circulation, venous return should always equal CO. Major determinants of CO are preload, contractility and afterload. During hyperoxia, left ventricular afterload clearly increased. The absence of a measurable decrease in CO may be explained by a concomitant increase in preload. Indeed, we found higher CVP during hyperoxia. The alternative explanation, i.e., increased contractility, is unlikely as we found no increase in the slope of the cardiac output curve during hyperoxia.

The circulation can also be described by the venous return to the heart, which is driven by P _msf − CVP. During hyperoxia P _msf increased more than CVP. However, this did not lead to an increase of venous return due to the simultaneous increase in venous resistance.

Vasoconstriction may be the key driver of most if not all effects of hyperoxia on hemodynamics. Not only does it increase blood pressure, afterload and venous resistance, it also leads to a shift of blood from unstressed to stressed volume, as indicated by increases in P _msf and CVP. We also observed an increase in P _cc by hyperoxia. P _cc is a theoretical pressure defined by dynamic pressure flow-relations and represents the arterial pressure, below which conceptually no flow will be possible. It is a combined estimate representing all vascular circuits. Theoretically, P _cc is the sum of arterial wall tension and the pressure surrounding the blood vessel. P _cc may differ importantly between different vessels, and measured P _cc is an average value for the complete vasculature. An increase in P _cc may be especially relevant in certain disease states such as increased intracranial pressure and abdominal compartment syndrome with high pressures surrounding the vessels. In both situations, P _cc is markedly elevated. For example, intracerebral blood flow will decrease to zero when arterial blood pressure is lower than the critical closing pressure of the brain and beyond the limits of cerebral autoregulation. In such states, vasoconstriction, by either vasoactive drugs or hyperoxia, may have either beneficial effects by increasing blood pressure or harmful effects by increasing arterial wall tone and thereby P _cc.

In our study, we could not show a reduction in perfusion of the brain by hyperoxia. Similarly, in a previous report perfusion changes at all oxygen levels were relatively small [27, 28]. It should be noted that all these studies were performed in situations with normal intracranial pressure. In situations with intracranial hypertension, such as in traumatic brain injury, we cannot rule out that a further increase of P _cc by hyperoxia will decrease the pressure difference between P _a and P _cc which can lead to a lower perfusion and possible ischemia of the brain.

Comparing our findings with earlier studies on the effects of hyperoxia on hemodynamics, the cardiac output fall was less than in healthy volunteers [29] and the decrease in the percentage of perfused vessels of the microcirculation was also in a different order of magnitude than previously observed [30]. Recognizing the perfused boundary region of microcirculatory vessels as a surrogate measure for the dimension of the glycocalyx, we could not detect any hyperoxia-induced alterations. Considering the effects on the venous system, more pronounced effects are to be expected in the smaller vessels compared to larger vessels. As a limited number of arterioles are present in the sublingual mucosa, where capillaries and venules are more abundant, only slight changes were anticipated in the analyzed microcirculation when high oxygen levels are applied.

Study differences may be largely explained by the use of anesthesia and mechanical ventilation as both affect hemodynamics and anesthesia also induces a considerable decrease in stressed volume. Furthermore, even in the presence of healthy lungs, both mechanical ventilation and bypass surgery may inflict an inflammatory response which can modify the effects of hyperoxic ventilation on the circulation in comparison with healthy subjects. Remaining differences may be clarified by the short exposure time in our procedures, although no further increase in PaO₂ was to be expected from a longer exposure and therefore a steady state in hemodynamics was assumed.

The increase in stressed volume and P _msf by hyperoxia mimics the effects of administering a fluid bolus, yet without increasing the R _vr. It is well known that the effects of extra fluids on CO are most pronounced in situations with underfilling of the vasculature explaining the relative conservation of cardiac output during hyperoxia in our postoperative, sedated patients, compared to healthy subjects. The effects of hyperoxia closely resemble the effects of norepinephrine and are in contrast to the effects of propofol [31, 32]. We earlier showed that intravenous administration of norepinephrine resulted in increases in R _sys, R _vr and P _msf. Interestingly, CO increased in some but not all patients after norepinephrine [32]. An increase in CO was associated with a higher SVV. Thus, it appears likely that the effects of a shift from unstressed to stressed volume by vasoconstriction, with an increase in P _msf, is mostly found in patients with vasoplegia and/or a decreased circulating volume. Hence, the effects of hyperoxia on CO are determined by the balance between volume recruitment (P _msf) and change in R _vr and baseline heart function, as observed before [32]. Although our results clearly indicate that hyperoxia increases venous resistance by venous vasoconstriction and that left ventricular output resistance (R _vo) did not change, we must realize that our description of the circulation is not complete. We cannot describe the part between the site where P _cc exists and the site where P _msf exists. Therefore, there is a missing part of the circulatory circuit, i.e., the distal arterial compartment, where control of the peripheral circulation is performed by the pre-capillary sphincters.

A recent study with an alternative cardiac output monitor of the arterial pressure wave showed a poor correlation with the thermodilution obtained CO values while changing norepinephrine doses [33]. However, measurements carried out by our group suggest that the Modelflow technique is capable of measuring the effects induced by vasoconstriction in an accurate manner [32], suggesting that vasoactive agents may not importantly affect the precision of your technology. This was also underlined by the CO values measured by the LiDCOplus monitor that showed a similar pattern compared to the Modelflow technique in our study. Furthermore, the determination of P _msf is not dependent on the accuracy of the Modelflow technique. Indeed, extrapolation of the venous return curve to flow zero is independent of absolute cardiac output. The ability to follow changes in cardiac output within a patient has been clearly demonstrated before [18]. We also showed that beat-to-beat changes in Modelflow cardiac output follows cardiac output by beat-to-beat analysis of electromagnetic probe flow signals [19].

Acknowledging that our findings are to be reproduced in a larger cohort and different clinical settings, the following study aspects should be considered. First, the small sample size and the specific subgroup of patients do not warrant a broad generalizability for the observed effects. Hemodynamics in the current patient group may be affected by the effects of the recent bypass and the potential mediators of ischemia reperfusion and inflammation. Other subsets of critically ill patients may respond differently than patients in our cohort who were in a relatively stable condition before starting study procedures. Two patients received small doses of norepinephrine during the experiment to keep the blood pressure in the same order of magnitude as the other included patients but showed stable hemodynamics and the dose was not changed throughout the experiment. Also, excluding these patients from our analyses showed virtually no change in our results.

There may be a time effect in which recovery and stabilization of patients in the ICU after surgery may influence hemodynamics. However, assuming that the effect of hyperoxia was transient and respecting a 15-min time gap between the two exposures, the carryover effect was minimized and each case served as its own control (self-matched) [34]. Adjusted changes in estimates were based on within-subject comparisons of exposure to hyperoxia with exposure to normoxia. Sampling bias was minimized by continuously measuring central circulatory variables, which provide a highly accurate representation of the parameters over the time periods. Cerebral blood flow, microcirculation and parameters assessed from the inspiratory hold procedures were measured intermittently, yet at representative sampling moments during the sequential time points and averaged as appropriate.

Since we could not detect large differences between the outcomes of univariate and multivariate statistical models accounting for repeated measurements, the observed effects may be predominantly attributed to hyperoxic ventilation, rather than to concomitant changes in other parameters. Other covariates that were considered, such as PaCO₂, are therefore not a likely explanation for the hemodynamic changes as seen during the experiment. While a short period of supraphysiological arterial oxygenation may disturb the hemodynamic balance, the effects of long-term exposure to hyperoxia are still uncertain but may be essential regarding patient-centered outcomes.

Short-term hyperoxia after cardiac surgery induces significant alterations in systemic circulation mainly by vasoconstriction of both the venous and arterial circulation and an increase of mean systemic filling pressure. The increase in stressed volume and systemic filling pressure by hyperoxia resembles the effects of administering a fluid bolus or norepinephrine. This may have clinically important consequences in critically ill patients when hemodynamic and microcirculatory changes are vital, but the effects were not clearly linked to relevant changes in cardiac output and cerebral blood flow.