Introduction

At least two major hurdles need to be overcome to seriously consider the issue of POCD mitigation. The first deals with the definition problem tackled at the beginning of this book. It is difficult to prevent or treat an undefined or variously defined syndrome. The second is an understanding of the underlying mechanisms in order to more effectively design therapies. Despite that neither of these hurdles are cleared, several attempts at POCD mitigation have been made in the hope of sparing our patient this troubling sequelae of surgery. We will summarize these strategies here as a function of whether the POCD was early or late (delayed neurocognitive recovery, or mild/major neurocognitive disorder). We will largely focus on clinical studies rather than preclinical animal models, due in part to the challenges of translation in the latter. There are a host of mitigation strategies which fall in the domain of cognitive and behavioral therapy, environmental, nutritional, emotional, and societal. While it is possible, even likely, that these approaches will be or are being more effective than most of what we discuss below, they have not yet been rigorously studied and thus will not be further addressed here.

The perioperative setting is complex and heterogeneous. The average number of drugs given approach a dozen, tissue damage and pain are induced, and the environment is unfamiliar, noisy, and stressful. Sleep disturbance is common and nutritional status often tenuous. Any of these features could underlie cognitive decline, especially in the elderly. It seems unlikely that PND has a single cause, but it is possible that there exists a dominant one. And since anesthetic drugs produce profound effects in the brain, and they are administered to most surgical patients, they are prime candidates for initial study.

Anesthesia: Regional versus General

Many surgical procedures do not require general anesthesia, but rather can be satisfactorily accomplished using regional anesthesia. Thus, a very early question was whether PND could be mitigated by avoiding general anesthesia altogether, and these, generally lower extremity surgical procedures, were an obvious choice for study of this question. Despite varying definitions of POCD and often inadequate power, most studies have concluded that there is no significant difference in the incidence of cognitive decline between general and regional anesthesia.^[1–3] But before moving on to the obvious implications of these results, it is important to point out that, at least in the United States, regional anesthesia is almost always accompanied by sedation, which in some cases approaches loss of consciousness (i.e., general anesthesia).^[4] The second confounder is that the class of general anesthetic used is often not controlled in these studies. Chapters 7 and 8 emphasize that not all general anesthetics are alike in their potential to promote cognitive impairment, at least in preclinical studies. Furthermore, as we will see below, the injectable general anesthetics may be distinctly different than their volatile cousins. Finally, general anesthetic “depth” is rarely controlled in these studies, as the technology needed for this is relatively recent. As we will see below, this might also be important. The comparison between regional and general anesthesia is therefore confounded and, as a result, differences are difficult to detect. Regardless, a lack of difference in PND incidence between regional and general anesthesia has been interpreted to indicate that anesthesia per se is not a dominant cause of PND, but the below considerations suggest that anesthetic choice and management are capable of modulation.

General Anesthetic Class

The currently used volatile general anesthetics (isoflurane, desflurane, and sevoflurane) are used more or less interchangeably in adults by anesthesiologists, with choices being made in some cases for pharmacokinetic reasons. Because these drugs are structurally and physicochemically quite similar, the mechanism of their primary effect, hypnosis, is likely to be similar. But, as most anesthesiologists know, their side effect profiles differ.^[5] Thus, it is not unreasonable to expect differences in their ability to cause or modulate PND. Adequately powered and controlled clinical studies examining differences between these volatile drugs on longer than 3 days postoperative have not been reported, but preclinical studies in rodents suggest that desflurane and possibly sevoflurane are associated with slightly lower risk of cognitive decline than isoflurane.^{[6, 7]} Although probably unrelated, it is interesting to note that the use of sevoflurane in humans has been progressively increasing over the last decade or so.^[8]

Good pharmacokinetic models, coupled with improved infusion and monitoring technology, and more specific injectable general anesthetics, have increased the popularity of TIVA (total intravenous anesthesia) for even very long surgical procedures. The most popular drug used in TIVA is propofol, an anesthetic which is frequently administered with an opioid for analgesia. Some studies have compared the effect on very early POCD (3 days) of using propofol, desflurane, or isoflurane and found inconsistent differences.^{[9, 10]} In a rare randomized clinical trial of sevoflurane versus propofol or regional for spinal surgery, there was a greater loss of cognitive abilities in the sevoflurane group than in either regional, propofol, or control groups.^[11] Furthermore, TIVA was associated with lower CSF tau in patients after surgery than sevoflurane anesthesia, a surrogate biomarker for brain injury or stress.^[12] Few other clinical studies have directly compared these drugs using longer-term cognitive outcomes, but preclinical data almost uniformly suggest that propofol anesthesia is followed by less cognitive disturbances than volatile drugs. For example, in a transgenic model of Alzheimer’s disease, desflurane anesthesia and surgery caused profound and durable learning and memory deficits, an effect that was entirely mitigated by using propofol instead of desflurane for the anesthesia.^{[13, 14]} Although good randomized clinical trials are needed, these results suggest that anesthetic choice might matter, and that TIVA using propofol could be associated with fewer postoperative cognitive symptoms than anesthesia with volatile drugs. It is important to point out, however, that TIVA is not without its own risks, such as respiratory and cardiac depression, which may not be well tolerated by the elderly or the critically ill.

Many general anesthetics act at least in part through allosteric enhancement of the GABA_A receptor. In general, these effects contribute to anxiolysis, amnesia, and sedation, and are not thought to directly contribute to POCD. However, recent work has suggested that GABA_A receptors containing the α₅ subunit may be enhanced long after the drug has been eliminated.^[15] This receptor has also been linked to amnesia, and therefore may contribute to cognitive disorders in the weeks to months following anesthesia and surgery. Inverse agonists of this receptor subtype are now available, and may prove to be useful if given in the early postoperative period. Preclinical research in this area is ongoing.

General Anesthetic Depth

The technology for monitoring anesthetic depth using processed EEG signals has become widely available and is being more commonly used in all sorts of surgical procedures. An hypothesis has been advanced that PND may be associated with “excessively” deep anesthesia. The neurophysiological rationale for this is not entirely clear, except for the general concept that precisely reconstructing preoperative cognitive states might be less likely from more disconnected states (isoelectric or burst-suppression EEG patterns). The evidence for this is still weak, but two studies suggest merit. First, postoperative delirium was significantly more likely in elderly patients who were more deeply sedated for repair of their hip fracture than a lightly sedated group, using a targeted EEG strategy.^[4] In a small prospective randomized study of monitored general anesthetic depth, patients receiving monitor-targeted anesthetic management had significant less cognitive decline at all time points up to a year after surgery.^[16] In contrast, a recent paper showed that “deeper” anesthesia was associated with better cognitive recovery in older patients,^[17] suggesting neuroprotection. Thus, while the evidence should still be considered preliminary, it is possible that a focus on avoidance of excessively deep anesthesia may be a mitigating factor for the PND syndromes. The risk of this approach is the increased likelihood of intraoperative awareness, which itself might be an independent risk factor of PND.

Sedatives and Psychotropics

It is well known that specific psychotropic, anxiolytic, and sedative medications are poorly tolerated by the elderly – the group at highest risk for the PND syndromes. For example, the benzodiazepines are generally thought to enhance confusion and delirium in the elderly, but have not yet been linked to PND specifically. For example, benzodiazepine blood levels a week after surgery were not correlated with the degree of cognitive decline in a small group of geriatric patients.^[18] Haloperidol is frequently administered to patients experiencing delirium, yet some evidence exists that it actually prolongs delirium and increases mortality.^{[19, 20]} Thus, although the data are inconsistent at this point, an easy mitigating factor, already deployed by many centers providing operative care of elders, is to limit or eliminate the use of specific medications when not otherwise indicated as listed in the American Geriatrics Society 2015 Beers Criteria.^[21] In addition to the specific sedatives and psychotropic medications mentioned above, the use of anticholinergics, antidepressants, and certain pain medications (e.g., Demerol) in elders should be viewed with caution.

The inability to reduce PND by using regional anesthesia as summarized above suggests that something about the surgery might contribute to the various forms of PND. Certainly it is well known that surgery produces considerable systemic inflammation, which correlates roughly with the magnitude of tissue damage or physiological trespass. That this peripheral inflammation can be transduced to the CNS is also well appreciated. Transduction can be mediated at least two ways. First, cytokines, chemokines, and other humoral mediators and messengers can cross the blood-brain barrier (BBB), either through active transport processes, or through a BBB rendered leaky by advanced age, neurodegeneration, or vascular disease.^[22] Once there, these factors can act directly on resident inflammatory cells (microglia or astroglia), or on neurons themselves. The second means of inflammation transduction into the CNS is via the migration of inflammatory cells, such as monocytes and macrophages.^[23] Like cytokines, once in the CNS, these cells can initiate or amplify neuroinflammation. Finally, peripheral inflammation is sensed by cholinergic neurons in the gut, and directly conveyed to the CNS via vagal afferents.^[24] The CNS, in turn, sends signals back to inflammatory cells in the periphery via vagal efferents in a form of feed-forward amplification. The ultimate effect of these various inflammatory pathways and mechanisms is to create a soporific effect which is often termed sickness behavior. It is likely that this inflammation-triggered sickness behavior is responsible for, or actually constitutes the most common form of PND, delayed neurocognitive recovery (dNCR) – such as in the first week to month after surgery. This carefully orchestrated system is designed to facilitate inactivity and healing, and tampering with it is likely to cause undesired effects. For example, it was found that induced-dysfunction of either IL-6 or TNF-α, while capable of mitigating the cognitive consequences of orthopedic surgery in a murine model, significantly altered innate immunity and tissue healing.^{[25, 26]} As opposed to tampering with the pro-inflammatory side, work has begun to evaluate the anti-inflammatory side, at least in preclinical models. The inflammatory resolution system is no less well orchestrated. A combination of anti-inflammatory cytokines (IL-4, lL-10) and lipid-derived resolvins appear to be important in resolving inflammation both peripherally and in the CNS.^{[27, 28]} We will discuss other forms of mitigation of the inflammatory response below.

Steroids, NSAIDs, and Cholinergics

Should peripheral inflammation be the primary culprit in initiating the CNS events resulting in PND, then limiting it using conventional anti-inflammatory medications is a plausible intervention. For example, in a mouse study, meloxicam was found to be effective at reducing early cognitive defects associated with reduced microglial activation.^[29] In a clinical trial, perioperative parecoxib administration reduced both delirium and POCD in patients up to 6 months (mild NCR [postoperative]) after arthroplasty surgery.^[30] Two trials of dexamethasone have recently been published, and both consisted of single administrations of the steroid at the beginning of surgery. One of these trials used high-dose dexamethasone (1 mg/kg) before cardiac surgery, with cognitive evaluations out to 12 months postoperatively. No significant effect of dexamethasone was noted.^[31] Similar negative results were reported for lower-dose dexamethasone given during noncardiac surgery.^[32] Interestingly in both studies, a trend existed for worsening cognitive decline in the higher-dose range of dexamethasone, perhaps an example of the careful inflammatory balance necessary for optimal cognitive recovery from a major stress. A high dose of dexamethasone increased POCD.^[33]

A wide variety of other drugs with anti-inflammatory properties (among other effects) have been tried, largely in preclinical trials with no consistent results. Lidocaine, for example, is primarily used as a local anesthetic for blocks and regional anesthesia, or as an antiarrhythmic, but is also recognized to have anti-inflammatory properties. In preclinical models, systemic lidocaine reduced early postoperative cognitive decline.^[34] Of interest, however, are the general anesthetic drugs already mentioned above. It is clear that some of the injectable anesthetics have significant anti-inflammatory properties, either as scavengers of reactive oxygen species (ROS), or by interfering with the inflammatory cascade. For example, propofol has been noted to be anti-inflammatory,^{[35, 36]} perhaps through multiple mechanisms, and thus has received considerable attention, because as a general anesthetic, it would be a relatively easy-to-justify alternative to inhalational anesthetics. A meta-analysis of all RCTs comparing propofol to inhalational anesthesia concluded that propofol was associated with a lower incidence of dNCR (“early POCD”) in the elderly.^[37] Likewise xenon has been touted as a “neuroprotective” anesthetic, although not necessarily anti-inflammatory. Although results with xenon have been inconsistent, at least one study, found that the incidence of PND after xenon or propofol anesthesia were not different.^[38]

It is now clear that cholinergic signaling in the CNS and peripherally is compromised in the elderly, and may contribute to age-related cognitive disturbances, such as Alzheimer disease and PND. This is the basis for the currently widespread and marginally successful anticholinesterase therapy in patients with mild cognitive impairment (MCI) and dementia. Whether anticholinesterase therapy is largely a symptomatic treatment as opposed to therapeutic (disease modifying) remains unclear, but most research suggests the former. Anesthesia and surgery acutely disrupt cholinergic signaling due to the various drugs used for hypnosis, muscle relaxation and reversal, and in the inflammatory pathways. As a result, several groups have searched for correlations between cholinergic signaling and PND. Although perhaps underpowered, these studies have been unable to find a significant association between cognitive outcomes and anticholinesterase use or anticholinergic activity.^{[39, 40]} A few small prospective trials of donepezil, given immediately prior to surgery have been uniformly negative.^[41–45] Like every other pharmacological option, anticholinesterase therapy has potential adverse effects, including bradycardia and hypotension.