A review of postoperative cognitive dysfunction and neuroinflammation associated with cardiac surgery and anaesthesia
This article is accompanied by an Editorial. See page 216 of this issue
You can respond to this article at http://www.anaesthesiacorrespondence.com
Summary
Postoperative cognitive dysfunction is receiving increasing attention, particularly as it mainly affects the (growing) elderly population. Until recently, cognitive deficits after cardiac surgery were thought to be caused by physiological disturbances associated with the cardiopulmonary bypass technique. Although the technique of ‘off-pump’ coronary revascularisation may potentially be associated with improved outcome, long-term follow-up studies have failed to demonstrate a significant reduction in the incidence of postoperative cognitive dysfunction. The focus of research is thus shifting from cardiopulmonary bypass to other factors common to both techniques, such as surgery, anaesthesia and patient-related predisposing factors. Priming of the immune system by ageing and atherosclerosis may result in an exaggerated systemic and cerebral inflammatory response to cardiac surgery and anaesthesia, causing neuronal loss or dysfunction resulting in cognitive dysfunction. We briefly discuss the evidence for cardiopulmonary bypass-related neuronal injuries in adult cardiac surgery patients, and review the evidence that immune priming is a key factor in the pathogenesis of cognitive dysfunction after cardiac surgery.
The first reports of cognitive disturbances appeared within a few years of the introduction of open heart surgery [1]. At that time, the focus of research was on psychological aspects, such as anxiety, depression and stress related to the experimental nature of the operation and the low chance of making a full recovery or even survival. With increasing experience and improved surgical and anaesthetic techniques, peri-operative and in-hospital mortality rates have decreased markedly, to around 1–3% [2, 3]. Despite the declining incidence of other complications, the reported incidence of postoperative cognitive complications has remained largely unchanged [4–6].
The American College of Cardiology and the American Heart Association have classified the neurological complications after cardiac surgery into two categories, namely type I and type II [7]. Type-I neurological deficits include stroke and transient ischaemic attack, coma and fatal cerebral injury. These deficits are clearly defined diagnoses and can be detected by clinical neurological examination. In contrast, type-II neurological deficits are diffuse and not well-defined, and include delirium and postoperative cognitive dysfunction (POCD), involving deficits of memory, concentration and psychomotor speed [8]. At discharge from hospital, the reported incidence of POCD is 30–65%. After a few months, the reported incidences are still of the order of 20–40% [5, 6, 9].
Over the years, hypotheses about the pathogenesis of POCD have changed markedly. Until recently, POCD was assumed to be the result of physiological disturbances caused by the cardiopulmonary bypass (CPB) technique. Although intuitively appealing, the evidence underpinning these theoretical associations is becoming increasingly weaker [10]. Well-conducted randomised controlled trials, with blinding of assessors, have shown that the incidence of POCD is similar regardless of whether coronary revascularisation is performed with CPB (‘on-pump’) or on a beating heart (‘off-pump’) [11, 12], although in one recent study, patients showed better long-term cognitive performance 7.5 years after off-pump coronary revascularisation, when compared with patients operated on-pump [13]. This latter study was also randomised and blinded for assessment of cognitive function, but long-term cognitive assessments were potentially biased, since they were only performed in patients who volunteered for follow-up (about a third of the study population). Other recent evidence suggests that POCD might not be more common after cardiac surgery than after other types of major surgery [9].
Although the pathophysiology and the definition of POCD are controversial, the long-term consequences have been well established. It is associated with decreased quality of life [14], early withdrawal from the workforce and increased dependency on society [15]. It is a significant problem for patients, families and health systems, as patients are often at a vulnerable age, when a step change in cognitive function can result in a loss of independence, escalating care needs and costs [16]. Furthermore, although not apparently as life-threatening as stroke, there is class-2 evidence1 that POCD after non-cardiac surgery is associated with an increased mortality of up to 10 years (hazard ratio 1.6, 95% CI 1.1–2.4) [15].
The current theories on the aetiology of POCD now include surgery-, anaesthesia- and patient-related factors. One factor that is likely to be common to all patients with coronary artery disease undergoing surgery is an inflammatory state. The purpose of this narrative review is to examine and discuss the literature concerning POCD and cardiac surgery, with an emphasis on the evidence for a role of inflammation as a common aetiological factor in patients undergoing cardiac surgery.
Methods
We performed a Medline-based search of all literature published up to June 2011 using the following keywords: cardiac surgery; coronary artery bypass grafting; CABG; coronary revascularisation; postoperative cognitive dysfunction; POCD; cognitive decline; cardiopulmonary bypass; inflammation; neuroinflammation; and cerebral inflammation. Relevant references from the articles identified from the literature search were also retrieved for further analysis. We primarily sought out human, randomised controlled trials, with blinding of assessment where feasible. Where no good human data were available, we searched for animal data. From the published literature, we identified the themes that form the outline of our review. To limit the number of references, we selected articles for inclusion based on a combination of the strength of the evidence and the time since publication.
In the review, the term ‘levels of evidence’ refers to the classification of levels of evidence recommended by the Oxford Centre for Evidence-Based Medicine.1 There were no articles with class-1 evidence. Most of the human studies are at level class 2b or weaker (single randomised controlled trials, or cohort studies). With regard to the potential role of CPB (and the associated physiological disturbances) in the pathogenesis of POCD, the literature is generally older, and involves observational or descriptive data (class 3 or weaker). The evidence for the links between cardiac surgery and inflammation, for a causative role of neuronal inflammation in the pathogenesis of POCD, and for a priming effect of ageing and neurodegenerative disorders on the inflammatory response, largely comes from animal data (some of which comprised randomised controlled trials), but these do not all concern cardiac surgery models.
Diagnosis of POCD
The incidence of POCD is highly dependent on the number and types of cognitive tests used, and the statistical analysis used to define a significant change in cognitive function [17]. The reported incidences of POCD vary widely, and part of the reason for this is that there are no standard, universally accepted diagnostic criteria [18]. Although eminent groups have issued consensus statements that recommend details such as core tests to be used, and the timings of these tests [19], these guidelines are often not applied [20]. To capture the spectrum of cognitive changes involved in POCD, most groups apply a battery of cognitive tests (each focused on a particular cognitive function).
Choice of tests is important because different cognitive tests differ in their susceptibility to confounders such as practice and floor and ceiling effects [17]. For a test that is susceptible to practice effects, a cognitively intact subject would be expected to improve his performance when repeating a test, and so this should be taken into account when analysing the results of repeated tests. Floor effects occur when a test is too difficult, resulting in low baseline scores, and reducing the chances of detecting a postoperative decline, particularly when a decline is defined in absolute terms. Ceiling effects are the result of tests’ being too easy, so that some subjects are able to achieve maximum scores despite cognitive decline.
Interpretation of tests may also have a critical influence on reported incidences of POCD. The outcome in a given patient or group when using a specific test battery depends strongly on the statistical methods used to define the cut-off point between POCD and normal variation in cognitive function [21]. Commonly used analytic criteria are a percentage change from baseline in a defined number of tests (usually a decline > 20% in two or more tests) or an absolute decline from baseline scores greater than a defined proportion of the standard deviation of the two or more tests (usually > 1 SD, calculated from baseline scores) [20]. These statistical methods do not relate cognitive decline with data from age-matched healthy controls, and thus fail to account for learning effects and normal variability and cognitive decline that would occur in a healthy population over the same period of time [20]. The reliable change index is an alternative method that relates the change scores to the normal test-retest variation in an age-matched control population, and is increasingly used in POCD research [9, 22].
It might be argued that POCD is a research construct based on test results and statistical criteria rather than a ‘real’ diagnosis. However, the fact that POCD is associated with adverse long-term outcomes provides a compelling reason to investigate further the pathogenesis and consequences of POCD (just as long-term outcome associations have ‘driven’ research and treatment of asymptomatic hypertension). Nonetheless, it should be acknowledged that there is debate about whether studies should continue to focus on these often sub-clinical, subtle changes, or only on clinically apparent changes reported by patients themselves [20, 23]. One of the problems of such subjective assessments of cognitive performance is that these may be influenced by depression [24]. Studies correlating subjective reports with objective measures have shown conflicting results [24, 25]. Other authors have advocated a return or greater reliance on clinical assessments [8, 18]. Unfortunately, there is no gold standard to establish the diagnosis of POCD. Thus, published incidences are strongly dependent on definitions that are mostly based on expert opinion.
Possible causes of POCD related to surgery
Altered cerebral perfusion and oxygenation
Cerebral blood flow is normally independent of perfusion pressure when the cerebral perfusion pressure is within the physiological range (approximately 60–140 mmHg). The cerebral perfusion pressure is usually regarded as the difference between the mean arterial pressure and the greater of the central venous pressure or the intracranial pressure. Cerebral blood flow may become pressure-dependent when autoregulation fails, and/or when the cerebral perfusion pressure is outside the physiological range, making the brain susceptible to harmful levels of ischaemia or hyperaemia when perfusion pressures are low or high, respectively.
During CPB, it is common practice for the arterial pressure to be maintained at around 60 mmHg. Although this may be adequate in normal people, it may be below the lower limit of autoregulation in some patients, particularly in those with a history of hypertension, which may reset the threshold for autoregulation. On the other hand, during off-pump surgery, manipulations of the heart, such as tilting it to facilitate distal anastomoses on the posterior surface, can briefly impair cardiac output, thereby reducing systemic pressure and cerebral blood flow [26]. Finally, in many patients a degree of heart failure, with peripheral and possibly cerebral oedema, may be present. In the 1990s, several magnetic resonance imaging studies demonstrated that cerebral oedema is common shortly after cardiac surgery [27, 28]. If cerebral oedema is indeed present intra-operatively or postoperatively, then this might increase intracranial pressure, resulting in a reduction in cerebral perfusion pressure to below the threshold for autoregulation, thereby causing impaired cerebral perfusion and oxygenation.
There are several reasons why autoregulation may be impaired during surgery, causing ischaemia when pressures are low, and hyperaemia and cerebral oedema when pressures are higher. One is hypothermia, which is often induced or tolerated during cardiac surgery to decrease the metabolic rate of the brain. Although this may decrease susceptibility to ischaemic and hypoxic injury, it is known that re-warming can disrupt cerebral autoregulation [29]. In particular, since such a large proportion of the cardiac output passes through the brain, fast re-warming can cause cerebral hyperthermia, which is known to be associated with adverse cognitive outcomes [30]. Another factor may be the choice of strategy to control the arterial partial pressure of CO2 (Paco2). With the pH-stat method, during periods of hypothermia the Paco2 is kept normal and as a result total body CO2 is increased. The latter has been shown to cause increased cerebral blood flow in excess of metabolic needs (i.e. loss of autoregulation) during hypothermic CPB, and to be associated with neuropsychologic impairment postoperatively [31]. It is speculated that any potential benefits of excessive cerebral blood flow during hypothermic CPB are outweighed by an increased cerebral embolic load [32]. When Paco2 is not corrected during hypothermia (alpha-stat method), cerebral autoregulation is maintained. Despite this apparent advantage, an increased incidence of jugular venous desaturation during rewarming has been shown in patients at risk for poor cerebral autoregulation receiving alpha-stat management [33]. Currently, most clinicians apply an alpha-stat strategy for adults undergoing hypothermic CPB, to maintain cerebral autoregulation and thus limit cerebral embolic load. A pH-stat approach is commonly preferred in paediatric cardiac surgery [34] and when cardiocirculatory arrest is required, although the latter is a subject of debate [35].
Overall, the evidence that impaired deranged cerebral haemodynamics is associated with neurological injury is weak and sometimes conflicting. There is level-2 evidence that suggests that maintaining perfusion pressure at more physiological levels during CPB (80–90 mmHg) is associated with less early POCD and delirium [36], whereas an intra-operative decline in mean arterial pressure of > 32 mmHg from the pre-operative baseline is associated with lower mini-mental state examination scores postoperatively [37]. However, a study using single photon positron-emission computed tomography failed to show significant correlations between changes in regional and global cerebral blood flow, and POCD [38]. The hypothesis that a pulsatile flow pattern in the CPB circuit would improve outcomes has also been opposed by several studies that found no beneficial effect of pulsatile flow compared with constant pump flow on organ perfusion [39], the inflammatory response, endothelial activation [40], and microcirculatory blood flow as seen with sublingual microvascular video recordings [41], brain tissue oxygenation [42], or cognitive outcomes two months after surgery [43]. Likewise, two cohort studies have failed to show an effect of temperature management during CPB on postoperative cerebral oedema [27, 28] and a large blinded randomised controlled trial (class-2b evidence) showed no effect on cognitive function [44].
At present, it is not possible to measure intra-operative cerebral blood flow or perfusion directly. It is, however, possible to perform real-time assessment of regional (frontal) cortical oxygenation using near-infrared spectroscopy. Brain tissue oxygen haemoglobin saturation reflects the balance between oxygen delivery and utilisation. Thus, low saturation levels provide an indication of a mismatch between cerebral perfusion or oxygen delivery and oxygen requirements. Two recent studies designed to investigate the influence of cerebral oxygenation monitoring and choice of anaesthetic technique on outcome have found (as secondary outcomes) an association between intra-operative cerebral oxygen desaturation and POCD, stroke and prolonged hospital stay [45, 46]. Interestingly, even pre-operative baseline cerebral oxygenation appears to be predictive of short- and long-term morbidity and mortality [47]. Thus, peri-operative near-infrared spectroscopy deserves further investigation as it has the potential to provide useful information about the incidence and time course of cerebral hypoxia, and may yield insights into the causes and possible methods of preventing and managing cerebral hypoperfusion and hypoxia during cardiac surgery.
Cerebral microemboli
The CPB circuit and the surgical field in cardiac surgery are a potential source of a variety of embolic particles such as thrombi, fat or gas bubbles. In addition, emboli can be caused by disruption of aortic atherosclerotic plaques by aortic manipulation and cannulation. It has been hypothesised that although occlusion of larger cerebral vessels causes focal neurological damage, more diffusely spread very small cerebral emboli could cause the more subtle deficits associated with POCD [48]. Diffusion-weighted magnetic resonance imaging techniques indicate that about 50% of patients undergoing on-pump coronary artery bypass grafting develop discrete lesions suggestive of microembolic infarcts [6]. Postmortem studies have confirmed the association between similar lesions and embolic infarcts [49]. It has also been shown using transcranial doppler ultrasonography that on-pump surgery is associated with larger numbers of microemboli than off-pump surgery (cohort study, level-3 evidence [50]).
Nonetheless, the correlation between embolic load and POCD is unclear. Several studies have failed to show a correlation between embolic load (transcranial doppler), or infarct load (diffusion-weighted magnetic resonance imaging) and POCD or delirium after coronary artery bypass grafting [6, 50]. In patients undergoing valvular surgery, conflicting results have been found with regard to the relationship between embolic load and POCD [51, 52]. Finally, because current imaging techniques are incapable of detecting very small lesions (diameter < 3 mm), a correlation between these very small microemboli and POCD cannot yet be excluded.
Inflammation
It has been well demonstrated that on-pump cardiac surgery induces a widespread systemic inflammatory response, that is associated with POCD [53]. Indeed, in one study, small doses of ketamine appeared to reduce the incidence of POCD following on-pump coronary artery bypass grafting, possibly via a beneficial effect on the pro- and anti-inflammatory systems [54]. This widespread inflammatory response to on-pump cardiac surgery was previously thought to be caused by activation of the immune system following contact between blood and artificial materials of the bypass circuit, and by ischaemia-reperfusion injuries of the heart, lungs and kidneys following CPB [55, 56]. In the presence of blood-brain barrier injury or disruption, it follows that systemic inflammation induced by CPB might also cause cerebral inflammation.
However, in rats, both CPB and sham surgery result in similar cerebral tumour necrosis factor-α activation, accompanied by early postoperative neurocognitive impairment [57]. The argument that CPB-induced systemic inflammation causes cognitive decline is further weakened by a failure to show significant differences in systemic inflammatory markers between patients randomised to on-pump or off-pump surgery [58]. Unfortunately, the strength of evidence from this study was undermined by the choice of sampling times, that may have missed potential early postoperative differences (inflammatory markers were only measured at baseline, after protamine administration and then after 4, 8 and 30 days). A narrative review of studies comparing leucocyte and endothelial cell activation and cytokine release in on- and off-pump surgery shows variable and conflicting results [59]. One high-quality randomised controlled animal study did show increased hippocampal expression of nuclear factor κB in rats after CPB compared with sham surgery, but this was not accompanied by neurocognitive impairment [60]. These findings suggest that there is no direct relationship between CPB, systemic and cerebral inflammation, and POCD.
Several common surgical procedures must occur regardless of whether coronary revascularisation is performed on-pump or off-pump. These include skin incision, sternotomy, graft vein or artery harvesting, and incisions in pericardium and coronary arteries. All of these generate tissue trauma, which may precipitate a systemic inflammatory response, but interestingly, the release of cytokines and other inflammatory biomarkers, such as secretory phospholipase A2, matrix metalloproteinase-9 and tissue inhibitor of matrix metalloproteinase-1, may not be greater following on-pump than off-pump coronary revascularisation [61, 62]. Although these studies are only descriptive cohort studies (level-3 evidence), they suggest that the surgery itself may have a greater impact on the inflammatory system than CPB.
Possible causes of POCD related to anaesthesia
Temporary deactivation or depression of cerebral functions is intrinsic to the functioning of anaesthetic agents. Symptoms of impaired cognitive functioning, such as drowsiness or impaired memory, can persist for a considerable time after recovery from anaesthesia. Therefore, it is possible that hypnotic or anaesthetic agents could amplify the effects of the insults associated with cardiac surgery and cause long-term cognitive deficits or even permanent structural changes in the brain.
A number of studies report that the clinically available opioids fentanyl, sufentanil and remifentanil can be neurotoxic in rats [63, 64]. Fentanyl is associated with delirium [65], but there seems to be no clear relationship between fentanyl dosage and the incidence of POCD 3 or 12 months postoperatively [66]. There is no convincing evidence that anaesthetic agents cause inflammation resulting in POCD; indeed, control animals in recent studies that received isoflurane or neurolept anaesthesia, but no surgical procedures, showed neither cytokine activation, damage associated molecular pattern molecule elevation, nor behavioural changes associated with POCD [67, 68].
There is evidence, albeit controversial, that volatile agents may enhance the susceptibility of neurons to apoptosis, and may enhance neurodegenerative processes [69, 70]. Furthermore, burst-suppression doses of sevoflurane can cause cerebral vasodilatation, which impairs cerebral blood flow autoregulation during CPB [71]. Nevertheless, there is no evidence that volatile agents are associated with POCD. Indeed, some studies investigating the effect of depth of inhalational anaesthesia on POCD have shown the opposite. In mice, high doses of isoflurane (2%) are associated with better cognitive performance after anaesthesia than lower isoflurane doses (1%) [72], suggesting some beneficial effects of isoflurane. In humans, deeper isoflurane anaesthesia titrated to the bispectral index (BIS) also results in better postoperative information processing [73]. Furthermore, two recent studies comparing propofol with sevoflurane [46] and desflurane [74] found a higher incidence of cognitive dysfunction in patients after propofol-based anaesthesia. These results suggest that there might even be a protective effect of inhalational anaesthesia, possibly mediated by so-called anaesthetic pre- and post-ischaemic conditioning.
Further evidence against the argument that anaesthetic agents may cause POCD has come from a study that found no significant difference in the incidence of POCD at 3 months among elderly patients undergoing non-cardiac surgery under either general or regional anaesthesia [75].
Patient-related factors
As is commonly known, normal ageing is associated with structural cerebral changes, including a reduction in grey matter volume and myelinated axon length [76]. Loss of neuronal dendrites, spines and myelin, as well as alterations in synaptic transmission and receptors, possibly underlie the normal decline in cognitive function associated with ageing [77]. This decline in cognitive reserve may thus mean that for a given level of neuronal injury (for example in terms of absolute number of neurons lost), the elderly may be more likely to suffer overt cognitive consequences than younger people. It may also explain the observed association between old age and impaired baseline pre-operative cognitive function, with worse early cognitive outcomes after surgery [46].
It is also possible that the response to any cerebral insult may be exaggerated in the elderly, in terms of neuronal dysfunction or losses. Possible explanations for an exaggerated response to cerebral injury in elderly and vulnerable patients include immune priming caused by pre-existing low-grade inflammatory states associated with ageing, neurodegeneration, and other chronic inflammatory disorders (see below).
A common factor: the inflammatory response
Whether focusing on surgery- or patient-related causative factors of POCD, a recurring theme is the inflammatory response. In the relationship between inflammation and (cerebral) tissue damage, it is difficult to distinguish cause and effect. Any tissue damage is generally followed by an inflammatory reaction associated with activation of the immune system, that can result in repair and healing, but can also result in further damage.
In the rat brain, both experimental subarachnoid haemorrhage [78] and global forebrain ischaemia [79] induce upregulation of inflammatory gene pathways, as represented by the transcription factor nuclear factor κB. Moreover, focal cerebral ischaemia not only results in local inflammation but also in widespread cerebral inflammation [80]. Thus, global neuro-inflammation can result from a wide variety of direct cerebral injuries.
Systemic inflammation may lead to brain tissue damage and possibly POCD. It is well known that a severe systemic inflammatory response can cause multi-organ failure, that can include the brain, leading to a range of clinical consequences including delirium and ‘septic encephalopathy’, with symptoms ranging from subtle cognitive deficit to coma with suppression of electroencephalographic activity [81]. Histological findings in animal models of sepsis show peri-microvascular oedema, swelling and rupture of astrocytic endfeet and neuronal degeneration, suggesting blood–brain barrier breakdown resulting in neuronal injury [82]. Thus, one might speculate that systemic inflammation known to be associated with cardiac surgery and to cause dysfunction of several organ systems via inflammatory effects also results in neuronal inflammation and cognitive dysfunction.
To date, there is very little direct evidence that POCD is the result of cerebral inflammation caused by neuronal injuries, systemic inflammation, or a combination of the two. Research in this field is complicated by the diagnostic difficulties of POCD, neuronal damage and inflammation. As mentioned before, diagnostic criteria vary, and cognitive tests are of varying sensitivity and specificity. Several groups have measured biomarkers of neuronal injury such as neuron specific enolase, S100B and nuclear factor κB, after cardiac surgery with CPB [83–85], and have found elevated plasma levels, but with varying correlations between these markers and cognitive function. Unfortunately, these biomarkers remain non-specific with regard to neuronal injury. In the case of S100B, the assay has been shown to cross-react with non-neuronal molecules [86]. Thus, it is currently not possible to link these findings with pathophysiological processes.
As for the serum markers of inflammation, a problem with their interpretation is the value of the relative levels of the different markers. Another critical issue when judging the significance of inflammatory marker levels is an understanding of the downstream effects of these biomarkers. Inflammation is a normal part of the physiological response to a variety of insults. It only results in harm when pro-inflammatory responses outweigh anti-inflammatory responses. An analogous issue of interpretation arises in infections and sepsis, where inflammatory processes are essential for resisting infections, but may also, when out of control, be significantly harmful. Some of the markers used as measures of the inflammatory response, such as tumour necrosis factor-α and interleukin-6, also seem critical for neuroprotection against insults (indeed interleukin-6 has both pro- and anti-inflammatory effects) [87, 88]. Thus, the relevance of inflammation can only be judged by studying its long-term consequences. Inflammation that protects against an insult and results in restoration of normal function is beneficial, whereas inflammation that results in apoptotic processes leading to neuronal loss is not.
Ideally, future studies will also identify factors that enable prior prediction of vulnerability to the harmful effects of the inflammatory response, and that may result in strategies for targeted interventions. One likely candidate is the interaction between genotype (e.g. apolipoprotein E status) and inflammation [84, 86]. Others have found a link between P-selectin and C-reactive protein genotype and susceptibility to POCD after cardiac surgery [89], and between interleukin-6 genotype and the risk of postoperative stroke [90], suggesting that targeted anti-inflammatory therapy may benefit some patients.
Immune system priming as a cause of POCD
Most patients presenting for coronary revascularisation procedures are elderly [mean (SD) age 65 (11) years], and many have co-existing peripheral vascular disease (15%) or diabetes (32%) [3]. These factors are all independently associated with inflammation. Indeed, detailed transcriptomic analysis has demonstrated a primed phenotype of peripheral mononuclear cells in some patients with vascular disease [61].
It is possible that immune priming by these factors is associated with an exaggerated inflammatory response to any significant stimulus or insult. Priming may thus induce a pro-inflammatory state, causing patients undergoing cardiac surgery, and particularly those undergoing coronary revascularisation, to suffer an exaggerated systemic and neuronal inflammatory response sufficiently severe to cause the widespread lesions responsible for POCD.
Immune priming by atherosclerosis
Chronic inflammatory conditions such as rheumatoid arthritis, psoriasis and systemic lupus erythaematosus are known to increase the risk of coronary artery disease significantly. Atherosclerosis, which is present in most patients presenting for cardiac surgery, is itself now considered to be an inflammatory disorder [91]. Several biomarkers of innate as well as adaptive immune response appear to predict the development and progression of coronary artery disease. Thus, most patients undergoing coronary revascularisation procedures will have a degree of immune system activation caused by atherosclerosis. The relationship between atherosclerosis, immune priming and POCD is at present speculative.
Ageing as a cause of priming
It is well known that ageing is associated with susceptibility to POCD [55]; age is a predicting factor for POCD, irrespective of the preceding type of surgery [9]. In humans, ageing without comorbidity is associated with low-grade systemic inflammatory activity – plasma tumour necrosis factor-α and interleukin-6 levels are significantly higher in older than younger humans [92]. Elderly patients are more susceptible to sepsis, and when they develop sepsis, the morbidity and mortality are significantly greater than in younger patients [93]. It is thus highly likely that in humans the microvascular endothelium is primed by ageing, making the elderly more susceptible to the harmful effects of inflammation [94]. This susceptibility to the harmful consequences of inflammation in the elderly probably applies to the brain, as well as to organs such as the kidneys and heart.
In animals, a number of studies have assessed the relationship between inflammation and cognitive performance in aged and younger adult mice, after an immune response evoked by peripheral infusion of lipopolysaccharide. After lipopolysaccharide administration, aged mice show an increased number of active microglial cells, and increased numbers of interleukin-1b-positive microglial cells in the hippocampus, and a corresponding greater decline in working memory [95]. Findings in other studies that compared aged with younger adult mice include an increase in both pro-inflammatory and anti-inflammatory cytokines [96], a prolonged and exaggerated elevation in the brain inflammatory cytokines and oxidative stress, with subsequent more pronounced sickness behaviour [97], and prolonged depressive-like behaviour linked to a higher turnover rate of brain serotonin [98]. Even after minor non-cardiac surgery, aged mice show increased levels of inflammatory markers in the hippocampus, associated with mild cognitive impairment [99]. Taken together, these studies provide growing evidence (level 2b–3) that ageing itself causes a low-level inflammatory state, which primes the immune system and results in a more pro-inflammatory state in response to multiple stimuli.
Neurodegeneration as a cause of priming
Another factor known to be associated with susceptibility to POCD is pre-existing cognitive impairment [55]. Recent evidence suggests that neurodegenerative disorders such as Alzheimer’s and Parkinson’s diseases (and interestingly also schizophrenia), are also associated with chronic neuroinflammation [100, 101]. Chronic neurodegeneration is associated with morphological and cell surface marker evidence of selective activation of microglial cells, but only minimal pro-inflammatory cytokine synthesis [102]. However, in rat brains in which microglial activation is primed by chronic inflammation with prion disease, a subsequent systemic challenge with endotoxin causes further morphological changes in hippocampal microglia associated with a profound pro-inflammatory profile, that can lead to neuronal death [103]. Interestingly, rats exposed to priming followed by a further inflammatory insult not only show an exaggerated inflammatory response, they also show evidence of exaggerated behavioural and cognitive changes and accelerated neurodegenerative disease [104, 105].
It is unclear how well these experimental findings with endotoxin extrapolate to less severe precipitants of inflammation. Nonetheless, this growing body of work suggests a possible (albeit speculative) explanation for the increased susceptibility of patients with pre-existing poor cognitive function to POCD.
The future
It is important to improve the identification and quantification of pre-operative risk factors for POCD, as this will assist clinicians and patients in the decision-making process (and informed consent) when invasive procedures are planned. To do this, the focus of research should shift from direct procedure-related factors such as CPB, to patient-related predisposing factors. To identify causative factors in a condition with such an apparently multifactorial aetiology, reliable measures of outcome are required. Future research has to implement validated cognitive test batteries in combination with standardised analysis with the reliable change index.
Accurate identification of pre- and postoperative risk factors for POCD will of course generate important hypotheses for potential targeted strategies for prevention of POCD. Routine intra-operative monitoring techniques focus on systemic haemodynamics and oxygenation, but provide little information about cerebral perfusion and oxygenation. Near-infrared spectroscopy offers a non-invasive method of assessing cerebral haemoglobin oxygen saturation, thereby giving an indication of the overall balance between oxygen delivery, requirements and uptake. It thus has the potential to demonstrate untoward intra-operative consequences of off-pump surgery. Early work also suggests that the technique can be used to assess cerebral autoregulation [106]. The role of inflammation in the pathophysiology of POCD requires further investigation, with special attention to possible priming factors such as atherosclerosis, diabetes mellitus and other chronic diseases. To facilitate these efforts, better methods of identifying and quantifying neuronal inflammation and neuronal injury are needed.
Continuing efforts are being made to identify biochemical markers that are specific to neuronal damage, and to relate the levels of these markers to the severity of the damage and the prognosis. Possible candidates include glial fibrillary-associated protein and brain- and heart-type fatty acid binding protein. Glial fibrillary-associated protein levels are related to severity of injury and adverse outcomes after traumatic brain injury [107], and to neurological outcome after stroke [108] and subarachnoid haemorrhage [109]. Brain- and heart-type fatty acid binding protein show higher sensitivities and specificities than protein S100B and neuron specific enolase in the rapid detection of brain injury in stroke, trauma and neurodegenerative diseases [110].
Progress is also being made with newer imaging modalities that will assist this research. Newer magnetic resonance imaging sequences, such as diffusion-weighted imaging, used in combination with increasingly powerful magnets (offering higher resolution), offer the potential to detect smaller microembolic infarcts. Other potential uses of magnetic resonance imaging include tractography (assessment of white matter tract integrity) [111], spectroscopy [112] (e.g. for assay of gamma-amino-butyric acid levels), and assessment of pre- and postoperative functional connectivity [113]. Functional magnetic resonance imaging has shown a reduction in pre-frontal activation in patients after on-pump cardiac surgery that was correlated with intra-operative cerebral embolic load, but not with a decrease in cognitive function [114]. This shows that functional magnetic resonance imaging can be capable of detecting subclinical functional impairments, but the clinical relevance of these findings remains to be established.
Several positron-emission tomography based techniques are also very promising. One such technique has been the development of a peripheral benzodiazepine receptor ligand, 11C-PK11195. Activation of microglia is associated with expression of peripheral benzodiazepine receptor on mitochondria. This receptor is selectively expressed in the mitochondrial walls of activated microglia [100, 101], and thus positron-emission tomography scanning with 11C-PK11195 can provide a quantitative assay of cerebral inflammation. Other positron-emission tomography modalities that may provide important information include functional methods for assessing cerebral glucose metabolism (e.g. fluoro-deoxyglucose labelled positron-emission tomography) [115], and positron-emission tomography tracers useful for assessing amyloid plaque and tau tangle (a deposition of defective microtubule stabilising proteins) concentrations [116].
Further studies into the pathophysiology of POCD are necessary, and may eventually inform strategies and therapies to prevent or attenuate POCD. Such studies must of course include further investigations to define better the influence of choice of hypnotic, and dose of opioid, on the inflammatory response to surgery and on the incidence of POCD. Alongside these studies, further work is required to determine the benefit, if any, of immune system modulation, not only by agents whose chief effect is an anti-inflammatory one, such as the corticosteroid dexamethasone, but also by other agents that may exert beneficial effects on the balance between pro- and anti-inflammatory mediators, such as interleukin-6 or tumour necrosis factor-α, and interleukin-4 or interleukin-10, respectively.
Conclusion
In the search for causative factors of POCD after cardiac surgery, the focus of research is moving from surgery-related factors (such as those related to CPB) to patient-related risk factors. The systemic inflammatory reaction that occurs after cardiac surgery may be associated with POCD. This inflammatory reaction was previously attributed to the CPB circuit, but increasing evidence shows inflammation to be of equal severity in off-pump surgery. A low-grade baseline state of systemic and neuronal inflammation is likely in groups of patients at risk of POCD, such as the elderly and those affected by neurodegenerative disease or atherosclerosis. These known risk factors may contribute to the development of POCD by priming the neuronal inflammatory system, and therefore lead to an exaggerated inflammatory response to CPB and cardiac surgery. The collection of evidence underpinning this theory is complicated by the great variety in methods for cognitive testing and measurement of inflammation and neuronal damage. The development of validated cognitive test batteries, standardised statistical methods and newer and more specific markers of neuronal damage will, one hopes, provide more insight in the complicated associations between risk factors, chronic low-grade inflammation, the inflammatory response and POCD in cardiac surgery. These insights might in turn have consequences for our future management of providing anaesthesia for cardiac surgery.
Footnotes
Competing interests
No external funding and no competing interests declared.
