CSF cytokine findings
We demonstrated increases in both pro- and anti-inflammatory CSF cytokines after vascular surgery on the thoracic aorta. The most marked increases occurred in the pro-inflammatory cytokines, IL-6 and IL-8. The increases in CSF IL-6 confirm previous findings in a similar cohort [16], whereas, to our knowledge, we are the first group to show similar changes in CSF IL-8. Interleukin-8 is a member of the CXC chemokine family, implicated in a wide variety of inflammatory diseases [23]. CSF IL-8 has been shown to be increased in AD, Parkinson’s disease [24] and following traumatic brain injury [25]. In patients undergoing other forms of surgery, findings of marked increases in CSF IL-8 have also been demonstrated [12, 14].
The increase in CSF cytokines in this study could have resulted from a dysregulated inflammatory response to the peripheral stimulus of surgery driving neuroinflammation within the brain. Alternatively, these findings may have been secondary to silent cerebral infarcts, which have been shown to be increased following thoracic aortic endovascular procedures [26]. Ischaemic strokes have been shown to drive an increase in proinflammatory cytokines [27]. It is also possible that the CSF cytokine changes may have occurred due to other post-operative complications such as endoleaks or infection.
Intercorrelation findings
Levels of cytokines in the CSF, while not exhibiting an association with each other prior to surgery, strongly correlate on day one after surgery. Figure 2 highlights the complex balance between pro- and anti-inflammatory cytokines which may drive neuroinflammation. The lack of an association between post-operative CSF and serum samples is consistent with other studies in this area [12, 14], and suggests that changes in cytokine levels in blood cannot be used reliably as surrogate markers of CSF cytokine changes.
Proposed mechanisms
The results of this study need to be interpreted alongside the current postulated mechanisms of brain dysfunction. These include excessive neuroinflammation [28], the production of reactive oxygen species (ROS) [29], and dysregulated neurotransmission [30].
These mechanisms are all mediated through activation of microglia, which when stimulated can release ROS [29] and cytokines [31]. Microglia are also key drivers of the kynurenine inflammatory pathway, which, in turn, can drive glutamatergic neurotransmission [32]. In the healthy brain, microglia are fundamental in maintaining tissue homeostasis by removing accumulated debris [29]. However, their overactivity may be harmful in disease states [33].
The postulated mechanisms of brain dysfunction are not mutually exclusive, with multiple mechanisms likely to be acting together [30]. Indeed, pro-inflammatory cytokines have been shown to activate the kynurenine pathway, which in turn leads to the generation of ROS through quinolinic acid production [34]. The cytokine changes demonstrated in this study may therefore have implications for several mechanisms of brain dysfunction.
Therapeutic targets
In contrast to neurodegenerative processes, peri-operative brain dysfunction occurs at a predictable time point, giving a potential opportunity for prevention [35]. Currently, no effective treatment for PND exists [36]. The suggestion that CSF IL-6 and IL-8 hold a key role in the post-operative neuroinflammatory pathway, raises the question of whether direct cytokine inhibition could attenuate these effects. However, as we have demonstrated in this paper, multiple cytokines, both pro- and anti-inflammatory, increase after an operation and thus blocking the action of only one of these cytokines may not necessarily inhibit neuroinflammation. Furthermore, as has been suggested in Alzheimer’s disease, a degree of cytokine-driven neuroinflammation may be neuroprotective [3,4].
Limitations
Similar to many studies in this area, this study had a small patient cohort, which may have limited our ability to demonstrate the true magnitude of peri-operative cytokine changes. Future studies should involve larger sample sizes across multiple settings to address this problem. Samples were only taken at two time points, with one patient’s CSF sample taken after surgery on day 0 rather than day 1. This patient was not excluded due to an already small cohort. Within the cohort, the surgical approach was not homogenous, with some patients undergoing open surgery and others undergoing endovascular surgery, which could represent a further complicating factor. The small numbers of patients in different surgical groups limits meaningful comparisons. A further confounder was that seven out of ten patients had undergone previous vascular surgery, often with serious post-operative morbidity. Ideally, this study would have corrected for underlying co-morbidity and baseline inflammatory status. In future studies, neuroimaging, to look for radiological evidence of stroke, should be included to investigate how much of the neuroinflammatory burden may be driven by ischaemic strokes. Finally, we cannot exclude the possibility that the inflammatory response was driven by the insertion of the spinal catheter, rather than surgery or anaesthesia, but this is felt to be unlikely due to the magnitude of cytokine changes [14].
This study looked solely at cytokine changes, which is only part of the neuroinflammatory process after surgery [35]. Measurement of the Q-albumin to determine the integrity of the BBB would also have been useful [7]. Future studies would ideally also examine the CSF cell count and immunoglobulin subtypes [37], and other markers of neuronal injury [8] to more fully understand pathophysiological processes. This was not possible within the scope of this study.
A final key limitation of this study was that patients did not undergo peri-operative cognitive testing. Formal cognitive testing for delirium, using screening tools such as the 4AT [38] and neuropsychological testing, would allow for the more direct investigation of correlations between observed CSF cytokine changes and the magnitude of cognitive dysfunction in PND.