The 4th International Workshop on Perioperative Neurotoxicity in the elderly is part of a continuing series to reflect the mounting interest and progress in the area. This meeting was held in conjunction with the Euroanaesthesia 2014 Congress in Stockholm, Sweden, on May 29, 2014. Hosted by the Department of Anesthesiology and Intensive Care Medicine at Karolinska Institutet, Stockholm, and organized by Niccolò Terrando, Lars Eriksson (Karolinska Institutet, Sweden), and Roderic Eckenhoff (University of Pennsylvania, PA), it included scientists from 10 countries and topics ranging from delirium to dementia, and molecular mechanisms to epidemiology. In earlier workshops, there was an emphasis on the preclinical rationale underlying perioperative neurotoxicity. In this fourth workshop, there was a clear shift to translational and clinical research to test the hypotheses flowing from these previous laboratory studies. In this summary, we will attempt to convey the progress, challenges, and remaining questions. This is not meant to be a scholarly review of the subject, and some of the points made herein reflect opinion and unpublished data. We recognize that there is a parallel concern for the immature brain. However, this was not the focus of this workshop because it is being adequately handled by other working groups.
There is agreement that the current nomenclature is confusing and constitutes an impediment to future progress. For example, some investigators “diagnose” postoperative cognitive dysfunction (POCD) statistically at various times after surgery, whether or not the patient (or family) notices a decline. Some attendees insisted that POCD, in fact, requires the patient to be asymptomatic, whereas others asserted that outside formal studies, postoperative cognitive complaints are very clearly symptomatic. Moreover, there is considerable confusion concerning whether POCD is always, or is ever, reversible. Does POCD lead to dementia, or is dementia that occurs any time after an operation a form of POCD? Should postoperative delirium be considered an acute form of POCD? Thus, an important outcome of the discussion was the proposed creation of a consensus panel to draft a comprehensive nomenclature for what could be considered the postoperative cognitive dysfunction syndromes (POCDS). Within such a nomenclature, the importance of not implying causation was stressed. In fact, it was recognized that some, perhaps most, surgery results in postoperative cognitive improvement (POCI), especially in younger to middle-aged populations. Some reported POCI is artifactual, a result of learning and practice effects on repeated cognitive tests, but a portion may reflect real cognitive improvement due to the benefits of the surgery (e.g., pain relief, improved mobility) outweighing any detrimental effects. In summary, the group agreed that a consensus nomenclature should allow improved comparisons among studies and greater precision when discussing these cognitive disorders and in evaluating potential risk factors and mechanisms.
Of considerable recent interest has been the role of inflammation in cognitive dysfunction, in general. This is likely to be relevant to the POCDs because of the very strong preclinical support. Some of the potential mechanisms whereby cytokines contribute to sickness behavior were discussed; Colm Cunningham (Trinity College, Ireland), Daqing Ma (Imperial College, UK), and Mervyn Maze (University of California, San Francisco, CA) presented recent work showing convincingly that inflammatory stimuli such as lipopolysaccharide or peripheral surgery produce neuroinflammation and short-term behavioral dysfunction in rodents and that this dysfunction is considerably greater and more durable in animals with a preexisting vulnerability (genetic, advanced age, metabolic syndrome, neurodegeneration, etc.). This is a critical and now widespread observation in animal models, which might partially explain why clinical studies have been so inconsistent; the size and character of subgroups with vulnerabilities are rarely known and likely differ among studies. Potential mechanisms linking anesthesia, surgery, and neuroinflammation include calcium dysregulation (Huafeng Wei, University of Pennsylvania, PA), tauopathy (Robert Whittington, Columbia University, NY), and amyloidopathy (Zhongcong Xie, MGH, Harvard University, MA), but the linkage to cognitive decline remains largely hypothetical. To begin to address this, Jianbin Tong (Central South University, China) showed a loss of hippocampal dendritic spines associated with surgery-induced neuroinflammation in aged rats. Furthermore, Hugh Hemmings (Cornell University, NY) found that anesthetic exposure alone could cause a similar loss of dendritic spines in rodent neurons. These studies illustrate how surgery and anesthesia might contribute to cognitive decline through a modulation of neuronal and synaptic plasticity. Despite this intriguing preclinical evidence for a causal role of anesthesia and surgery, further preclinical studies are needed to better understand the immune-to-brain signaling after peripheral injury, to what extent this signaling is modulated by different anesthetics, how (or if) the ensuing inflammatory response leads to cognitive decline, and whether dysfunction of inflammatory resolution leads to a more prolonged cognitive decline. Finally, given that the elderly seem to be at highest risk of POCDS, it is important that these preclinical studies take place in aged animals. Of course, all these questions ultimately will need to be translated to elderly humans.
Because of the growing appreciation for the potential role of preexisting vulnerabilities for POCDS, and because of the need for objective measures of postoperative injury, an array of human biomarkers is beginning to be studied. The workshop started with a keynote presentation from Kaj Blennow (Gothenburg University, Sweden), a world expert in cerebrospinal fluid (CSF) and blood biomarkers of brain injury, most notably Alzheimer disease (AD), and also head trauma and surgery. Evidence is accruing for more sensitive and specific biomarkers, which provides hope that the study of POCDS will transition from the highly variable (but still necessary) cognitive assays to continuous quantitative measures of neuronal injury and pathology. Novel and extremely sensitive analytical technologies will permit early detection of neurodegeneration and/or brain injury in humans. It has been proposed in the AD field that these early biomarkers can reflect the pathologic state while still reversible, preceding the poorly reversible cognitive effects by many years. Biomarkers are thus viewed as a promising tool in evaluating progression and modulation of disease trajectory. Early prospective studies on CSF AD biomarkers (amyloid-β and tau) were presented by Miles Berger (Duke University, NC), Zhongcong Xie (MGH, Harvard University, MA), and Lis Evered (St. Vincent’s Hospital, Australia) most of which show only a small effect of uncomplicated surgery in nonselected individuals. In general, CSF amyloid-β is not acutely sensitive to surgery, but CSF tau, a marker of neuronal injury, seems to be. However, preoperative CSF amyloid-β and/or tau levels appear to predict POCDS both acutely and at 1 year postoperatively. Again, this suggests the importance of preexisting vulnerabilities and subgroup analyses in future clinical studies. Biomarker studies in patients hold the promise of both illuminating the pathophysiology of POCDS and providing objective markers of disease progression.
Another form of biomarker is brain imaging. Joseph Mathew (Duke University, NC) presented work in cardiac surgery patients, which demonstrates postoperative changes in functional magnetic resonance imaging (MRI), which correlated with POCD. Preoperative amyloid burden, measured using 18F-florbetapir positron emission tomography (PET) imaging, was not correlated with POCD. Amyloid-β burden as revealed by PET imaging changes quite slowly, so repeat PET a year later is being conducted in a small number of these subjects to assess whether cardiac surgery is associated with accelerated amyloidopathy, as has been previously proposed. Several other amyloid PET studies are in progress, and data should be available in a year or 2. It is relevant to note here that tau and microglia/monocyte PET ligands are now available and might be useful to better document more acute changes in disease trajectory after surgery. Also using imaging, Charles Brown (Johns Hopkins University, MD) found an association between number of hospitalizations and structural changes on brain MRI in the ARIC (Atherosclerosis Risk in Communities) cohort, which, while not directly surgery (yet), suggests that the stress of illness during hospitalization may be linked to worsening neuropathology and cognitive decline.
Finally, genetic biomarkers for cognitive disorders are starting to be used to stratify POCDS studies. For example, Katie Schenning (Oregon Health & Science University, OR), when analyzing the OBAS (Oregon Brain Aging Study) and ISAAC (Intelligent Systems for Assessing Aging Changes) databases, reported a significant association between surgery and subsequent dementia and that the association was stronger in the apoEε4 allele carriers. Studies on the impact of apoEε4 on cognitive outcomes after surgery have been inconsistent, but it is emerging as one of the strongest predictors of late-onset AD and thus, requires further study in the perioperative arena. No studies have emerged on the impact of surgery on cognitive outcomes in carriers of early-onset AD genetics, such as PS-1, APPswe, or trisomy-21 (i.e., Down syndrome). Such studies would be useful because these patients clearly possess the vulnerabilities that we mention earlier.
Although neuroinflammation after anesthesia and surgery is consistently reported in animal models, the evidence for this in surgical patients is still scant. However, several attendees reported intriguing preliminary results. For example, Mervyn Maze (University of California San Francisco, CA) presented very preliminary data on an association between both blood and CSF interleukin-6 and postoperative delirium, while Tianlong Wang (Capital Medical University, China) found an association between several cytokines, including urine biomarkers, and POCDS, as well as a strengthened interaction in the elderly. In exploring the transition of peripheral inflammation to neuroinflammation after surgery, Sven-Erik Ricksten (Gothenburg University, Sweden) reported evidence for a disrupted blood-brain barrier using both biomarkers and MRI. This is consistent with the recent evidence of blood-brain barrier impairment and endothelial dysfunction after surgery in animals and therefore is at least 1 example of successful translation. However, despite this evidence for enhanced neuroinflammation, glucocorticoid administration (versus placebo) studies have not yet been effective at preventing POCDS. Thus, while translational studies on the role of neuroinflammation in POCDS are few and inconclusive at this point, the strength of the preclinical studies calls for more work in the area.
Several investigators are beginning to ask whether our perioperative management can influence neuroinflammation and cognitive outcomes; biomarkers will greatly facilitate such studies. Claudia Spies (Charité, Germany) discussed the use of neuromonitoring to control for the depth of anesthesia to evaluate POCDS and presented the launch of a large-scale biobanking consortium to establish valid biomarker panels and clinical outcome prediction of the POCDS among elderly patients in Europe. Miles Berger (Duke University, NC) reported no difference in postoperative CSF AD biomarkers in patients randomized to total IV anesthesia (TIVA) or inhaled general anesthesia (GA) for neurological surgery. However, he did note a lower level of the cytokine monocyte chemotactic protein 1 in CSF after TIVA compared with inhaled GA. Stacie Deiner (Icahn School of Medicine at Mt. Sinai, NY) found that plasma stress markers (catecholamines) and the incidence of delirium did not differ between patients who have received TIVA versus GA. Interestingly, high plasma catechol levels in the recovery room after surgery were a strong predictor of delirium. Jeff Silverstein (Icahn School of Medicine at Mt. Sinai, NY) in the delirium randomized controlled trial asked whether dexmedetomidine administered during surgery reduces postoperative delirium. Interestingly, preliminary analyses suggest the opposite, and on examining the associations with long-term outcomes (POCD), he reported a higher incidence of POCI than decline. Robert Sanders (University College London, UK) reported on the design of the Hip Fracture Surgery in Elderly Patients (HIPHELD) study where xenon anesthesia is being compared with sevoflurane but was unable to yet present results. Thus, the ability of any particular anesthetic or adjunct in the perioperative period to reduce the risk of POCDS remains conjectural.
Finally, David Scott (St. Vincent’s Hospital, Australia) presented results from a prospective trial demonstrating cognitive decline 1 year after surgery and found that baseline cognitive performance was a significant predictor. However, few if any centers routinely evaluate preoperative cognitive ability primarily due to the cost and time required to do so. Furthermore, it is clear that significant, preexisting cognitive disorders are routinely missed in the busy perioperative arena. Thus, Deborah Culley (Brigham & Women’s Hospital, Harvard University, MA) presented a small study using a very simple cognitive assay that requires little time and was easily administered in a routine preoperative evaluation center. Such an evaluation may reveal vulnerable and frail patients, thus allowing more informed discussions of patient risk and allocation of resources.
The above prospective clinical studies, designed to reveal a link between anesthesia, surgery, and cognitive decline, are difficult, time-consuming, and expensive. Thus, an intermediate translational step between the hypothesis-generating preclinical work and prospective clinical studies is the retrospective analysis of a growing variety of databases. In general, these databases were established for other purposes, and thus relevant and important data are sometimes missing, populations are widely variable, and surgical procedures (if documented) very different. When combined with what is likely to be a small effect size at the population level, firm conclusions regarding the relationship among anesthesia, surgery, and long-term cognitive decline have been difficult to make. For example, Juraj Sprung (Mayo School of Medicine, MN) presented recently published work arising from the analysis of Olmsted County databases. No associations between any form of anesthesia, surgery, or dementia diagnosis could be made. Conversely, as already mentioned, both Katie Schenning (Oregon Health & Science University, OR) and Charles Brown (Johns Hopkins University, MD) found significant associations between surgery and/or general hospitalizations and cognitive decline. Other data collected by Kirk Hogan (University of Wisconsin, WI), presented in absentia by Katie Schenning, also found a significant association between surgery and cognitive decline in the Wisconsin Registry for Alzheimer’s Prevention (WRAP) prospective study. Importantly, this last study is one of the few to enrich their population with a late-onset AD vulnerability; all enrollees had to have at least one parent with AD. Taken together, the epidemiology remains confusing in large part due to the well-known limitations of retrospective studies. The weight of evidence appears to favor anesthesia and surgery being associated with one or more forms of the POCDS, but definitive, prospective studies (such as a few of the above) are sorely needed to conclude this with confidence.
As indicated by all of the above, no anesthetic, perioperative approach, or adjunct can currently be recommended as an effective deterrent to all POCDS, although some interventions have been reported to reduce the incidence of postoperative delirium and early postoperative cognitive dysfunction. Nonetheless, no anesthetic has been clearly implicated or exonerated. Thus, in the absence of clear evidence of causation, considerable discussion centered around what to tell surgical patients about the risk of POCDS and whether this should be part of the consent process. Because anesthesiologists currently inform patients of risks at least as low as the estimates for POCDS, it was recommended that the risk of POCDS (perhaps excluding dementia for which there is the least evidence) be included and documented in preoperative discussions, just as the risk of cardiovascular events and mortality currently are. There was a clear consensus that some form of standardized brain function testing, like that of any other key organ system, should eventually be included in the pre- and postoperative period. Risk factors (vulnerabilities), once clearly identified, need to be documented to appropriately stratify and discuss risk with patients and their families and implement therapies designed at least to mitigate early POCDS. Finally, the lack of firm answers to many of the questions posed in this workshop should make it abundantly clear that more research, primarily in the translational and clinical domain, is urgently needed. Should causal relationships emerge, mechanistic studies should ultimately reveal approaches for mitigation.
Miles Berger: Duke University, NC; Charles Brown: Johns Hopkins University, MD; Kaj Blennow: Gothenburg University, Sweden; Gregory Crosby: Brigham & Women’s Hospital, Harvard University, MA; Deborah Culley: Brigham & Women’s Hospital, Harvard University, MA; Colm Cunningham: Trinity College Dublin, Ireland; Stacie Deiner: Icahn School of Medicine at Mount Sinai, NY; Maryellen Eckenhoff: University of Pennsylvania, PA; Roderic G. Eckenhoff: University of Pennsylvania, PA; Lars I. Eriksson: Karolinska Institutet, Sweden; Lis Evered: St. Vincent’s Hospital, Australia; Hugh C. Hemmings: Cornell University, NY; Daqing Ma: Imperial College London, UK; Joseph Mathew: Duke University, NC; Mervyn Maze: University of California San Francisco, CA; Sven-Erik Ricksten: Gothenburg University, Sweden; Robert Sanders: University College London, UK; David Scott: St. Vincent’s Hospital, Australia; Jeff Silverstein: Icahn School of Medicine at Mount Sinai, NY; Claudia Spies: Charité, Germany; Juraj Sprung: College of Medicine, Mayo Clinic, MN; Francois Stapelberg: University of Auckland, New Zealand; Katie Schenning: Oregon Health & Science University, OR; Niccolò Terrando: Karolinska Institutet, Sweden; Jianbin Tong: Central South University, China; Tianlong Wang: Capital Medical University, China; Huafeng Wei: University of Pennsylvania, PA; Robert Whittington: Columbia University, NY; Ting Yang: Karolinska Institute, Sweden; Zhongcong Xie: Massachusetts General Hospital, Harvard Medical School, MA.
Name: Niccolò Terrando, PhD.
Contribution: This author helped prepare the manuscript.
Attestation: Niccolò Terrando approved the final manuscript.
Name: Lars I. Eriksson, MD, PhD.
Contribution: This author helped prepare the manuscript.
Attestation: Lars I. Eriksson approved the final manuscript.
Name: Roderic G. Eckenhoff, MD.
Contribution: This author helped prepare the manuscript.
Attestation: Roderic G. Eckenhoff approved the final manuscript.
This manuscript was handled by: Gregory J. Crosby, MD.