Developmental Disability in the Young and Postoperative Cognitive Dysfunction in the Elderly After Anesthesia and Surgery: Do Data Justify Changing Clinical Practice?

Authors


Abstract

The assumption that anesthesia has no serious, long-term, adverse central nervous system consequences may be true for most patients between 6 months and 60 years of age. However, for patients younger than 6 months or older than 60 years, that status quo assumption is under challenge from a growing body of evidence. Fetuses and newborns appear to be at risk because systems that would enable them to fully recover from the effects of more than 2 hours of anesthesia are still in development. In distinction, the elderly appear to be at risk because systems that once enabled them to fully recover have ever-diminishing capacity. Even for those between the age of 6 months and 60 years, full recovery may require replacing apoptosed neurons and pruning overabundant dendritic spines perhaps leaving patients not quite the same person that they were before they were anesthetized. Mt Sinai J Med 79:75–94, 2012.© 2012 Mount Sinai School of Medicine

THE YOUNG BRAIN

After 28 weeks of gestation, fetal neurons develop an acute ability to die from boredom.1 Given estimates of 40–50 billion neurons at birth,2 and evidence that at least 1 fetal protoneuron, and more likely 2, undergo apoptosis for each neuron that survives,3 a midpoint estimate is that the human brain averages about 8000 apoptotic neuronal deaths per second during the last 11 weeks in utero. Those cellular suicides are highly selective, leaving the core material and sculpting the primary architecture for subsequent CNS development.4

The trigger for that avalanche of apoptosis is a lack of synaptic feedback. Apoptosis appears to be the default program of many excitable cell types, with cell-typical activity promoting proteins like antiapoptotic Bcl-2s that prevent the default program from running its course. Put differently, the old saying “Use it or lose it” is not only for the old synaptic activity may be as crucial to the survival of late-term fetal neurons as are oxygen, adenosine triphosphate, and cerebral blood flow. So what happens to fetal neurons that would be receiving and sending signals were it not for the presence of anesthesia?

In Laboratory Animals

One of the first animal models to test the effect of anesthesia on fetuses was developed by Chalon et al. in 1981. He exposed pregnant mice to halothane and found that their offspring, and the offspring of those offspring, learned significantly more slowly than the first and second generation of control mice.5 Chalon's findings for first-generation offspring were recently substantiated and extended for in utero exposure to isoflurane,6,7 and Hogan has found first-generation effects of fetal exposure to nitrous oxide that may extend to those offsprings' offspring.8 Recent studies notwithstanding, early laboratory reports indicating a potential problem did not receive the attention that they deserved until 2003, when Jevtovic-Todorovic and colleagues published, “Early exposure to common anesthetic agents causes widespread neurodegeneration in the developing rat brain and persistent learning deficits”—a title that says it all.9 Many subsequent studies have confirmed and augmented those findings,6–8,10–32 with Istaphanous et al. reporting that “developmental neurotoxicity is a common feature of equipotent concentrations of desflurane, isoflurane, and sevoflurane in neonatal mice.”29(cf22,30)

Early laboratory reports indicating a potential problem did not receive the attention that they deserved until 2003, when Jevtovic-Todorovic and colleagues published “Early exposure to common anesthetic agents causes widespread neurodegeneration in the developing rat brain and persistent learning deficits”—a title that says it all.

Neonatal apoptosis subsequent to a clinically relevant depth and duration of general anesthesia also occurs in mammals with periods of rapid synaptogenesis more analogous to humans, including pigs23 and a nonhuman primate.24 Potentially relevant for burn victims, 24 hours of “a light surgical plane” of ketamine anesthesia also causes long-term cognitive deficits in Rhesus macaques.31 Apoptosis notwithstanding, Stratmann and colleagues found that exposing 7-day-old rats to 4 hours of isoflurane anesthesia induced a decrease in neurogenesis that contributed to a permanent deficit in hippocampal-dependent learning and memory.18,19 Neurogenesis, of course, requires neural stem cells, and Culley et al. have presented evidence that 1 minimum alveolar concentration (MAC) isoflurane reduces the production of neural stem cells by 20% in vitro 24 hours after exposure.32 Decreased neurogenesis and decreased neural stem cell proliferation notwithstanding, using 16-day-old rats, Briner and coauthors found that sevoflurane, desflurane, and isoflurane rapidly increase dendritic spine density, which “could interfere with physiologic patterns of synaptogenesis and thus might impair appropriate circuit assembly in the developing cerebral cortex.”25,26

In Humans

Since 1945, investigators have observed an association between impaired neuro-cognitive-behavioral development and postnatal exposure to surgery and anesthesia prior to 3 or 4 years of age,33–52 with Levy33 having found a statistically significant association between near-term emotional sequelae and younger age at anesthetic exposure (P < 0.0004, data not statistically analyzed in original article). However, the investigation of monozygous twins by Bartels and colleagues addressed the hypothesis that children who need to undergo surgery and anesthesia at an early age are inherently predisposed to impaired neuro-cognitive-behavioral development. They studied 110 pairs of identical twins; in each pair, one twin had been anesthetized prior to age 3 years and the other had not. They found, in a pair-by-pair analysis, that the anesthesia-exposed twins had virtually the same score as their nonexposed twins on a national measure of educational achievement administered near age 12.48 Unfortunately, the Bartels study is not known to have included children who were anesthetized prior to 6 months of age. That shortcoming, together with the finding of Kalkman and colleagues of a trend toward a greater detrimental effect of anesthesia on neurocognitive development with decreasing age at administration of anesthesia,47(see also42) suggests that if there is a period of extraordinary vulnerability in humans, it is similar to that found at analogous developmental stages in nonhumans: second trimester to 6 months postpartum.

Since 1945, investigators have observed an association between impaired neuro-cognitive behavioral development and postnatal exposure to surgery and anesthesia prior to 3 or 4 years of age, with Levy having found a statistically significant association between near-term emotional sequelae and younger age at anesthetic exposure (P < 0.0004, data not statistically analyzed in original article).

The methodology of Bartels et al. was also not able to provide an estimate of duration of anesthetic exposure. If duration of anesthesia is as important in human fetuses and neonates as it is in nonhumans,23 then a 30- to 60-minute exposure may not be sufficient to affect long-term learning capacity, even in the high-vulnerability age group. Accordingly, Hansen and colleagues' finding of no substantive impairment in children “exposed to a single, brief anesthetic procedure in infancy” leaves the question open for ≥2 hours of exposure.51 The same concerns apply to a recent report from DiMaggio and coauthors.52 They looked for developmental effects in children exposed to anesthesia and surgery prior to 3 years of age and replicated the Bartels et al. finding for twins on a smaller but more refined subsample. Nevertheless, only 6 or 7 of the 304 anesthesia-exposed children included in the DiMaggio and coauthors epidemiological investigation were aged <6 months at time of exposure, with the vast majority having been anesthetized for short procedures. Accordingly, they concluded “that a meaningful proportion of the association measured in the overall analysis of the may not be causally attributable to surgery/anesthesia”—a conclusion that we agree with in reference to children exposed to anesthesia for <2 hours after they are 6 months old.

In distinction, a recent examination of children aged <1 year (average age, 101 days) exposed to anesthesia for procedures that lasted up to about 3.75 hours found an association with decreased academic performance after correcting for related CNS complications. Thomas and colleagues analyzed achievement test scores of 7- to 17-year-old children who received general anesthesia during infancy for procedures that are not independently associated with cognitive impairments: inguinal hernia repair/orchiopexy, pyloromyotomy, and circumcision. After excluding children with any of 14 prespecified CNS problems or medical conditions associated with learning disabilities, they found that a substantial proportion of children without such risk factors scored below the fifth percentile of the normative population (P < 0.01), with increased duration of anesthesia associating with reduced performance P < 0.01; (Figure 1).49

A recent examination of children aged <1 year (average age, 101 days) exposed to anesthesia for procedures that lasted up to about 3.75 hours found an association with decreased academic performance after correcting for related CNS complications.

Figure 1.

Relationship of Iowa Test normal curve equivalent scores to durations of anesthesia for patients without central nervous system risk factors. Each child had only a single operation date during infancy. The dashed lines indicate 95% confidence intervals (from Thomas et al., 2010).49

Findings for fetuses may be stronger than those for postnatal humans. In 1986, Hollenbeck and coauthors reported decreased cognitive capacity in 4-year-olds whose mothers had been anesthetized while they were in utero.53 Several subsequent studies found analogous associations between prenatal exposure to anesthetics and developmental problems including autism,54 hydrocephalus,55 diminished general intelligence,56 impaired spatial ability,57 small head size, and mental retardation.58 Whether a component of those adversities can be attributed to anesthesia should be decipherable given studies with sufficient statistical power to correct for confounding variables without obscuring an effect of anesthesia.

Ongoing Trials

With an anticipated completion date of December 2016 and a projected sample size of 660 children, the Multi-site Randomized Controlled Trial Comparing Regional and General Anesthesia for Effects on Neurodevelopmental Outcome and Apnea in Infants (GAS) study will test for a difference in preschool IQ between children who received sevoflurane or neuraxial bupivacaine for inguinal hernia repair when they were 26–60 weeks old.59 A recent laboratory experiment does not support the GAS study, because although it found substantially reduced neuronal apoptosis in postnatal rats that received spinal injection of bupivacaine compared with rats anesthetized with isoflurane, the spinal analgesia did not last long enough (40–60 minutes) to be expected to trigger apoptosis, whereas the apoptosis-inducing sevoflurane exposure lasted for 6 hours.60 A large retrospective study by Flick and coauthors found no difference in frequency of learning disabilities between (1) a group of children whose mothers received inhaled anesthesia (16%) or did not receive inhaled analgesia (84%) for vaginal delivery, and (2) a group of children delivered vaginally whose mothers received neuraxial block, with only 3.1% also receiving inhaled anesthesia.61 That study did not test for developmental differences between the children whose mothers received inhaled analgesia without neuraxial block versus the children whose mothers received neuraxial block without receiving inhaled anesthesia (the GAS study question), but even if those relevant subsamples are large enough to withstand such an analysis, one worries that fetal anesthetic exposure during birth in the Flick study was, and during hernia repair in the GAS study will be, too brief to test the anesthesia-development question (per Hansen et al51 and DiMaggio et al49). It may also be the case that the sample size of the GAS study will be effectively diminished because too high a proportion of participants will be too old at their age of exposure (per Bartels et al45 and DiMaggio et al49).

The other major prospective study scheduled for completion in 2016, the Infant Anesthesia Exposure and Neuro-Outcome study (formerly PANDA) is aiming to enroll 1000 participants to compare “global and domain-specific neurocognitive function” between children exposed to general anesthesia prior to 3 years of age during hernia repair versus siblings of nearly the same age (within 3 years) who were not exposed to general anesthesia prior to age 3. Unfortunately, like the GAS study, this investigation will test for an effect of exposures that are probably too brief to have an effect in a study population that may be substantially composed of children who are also too old to be sufficiently susceptible.62

So Where Do We Stand?

The data in laboratory rodents are conclusive: Clinically relevant doses and durations of anesthesia during the period of rapid synaptogenesis cause neuronal apoptosis and long-term learning deficits. The same has been established in pigs and a nonhuman primate with regard to apoptosis. In distinction, the effect of anesthesia in human neonates remains a concern that is confounded by genetics, by age at anesthetic exposure, by the effects of surgery independent of anesthesia, and by the duration of anesthesia exposure. Data from human fetuses may be a cause for even more concern because they associate anesthesia with adverse outcomes that are probably less confounded by genetics (the mother's genetic predispositions would be the primary association with a need for surgery, but half of her fetus' genes are not derived from her), by age at exposure (second and third trimester have now been implicated as high-risk periods in rodents, perhaps translating back to late first trimester and beyond in humans), by the effects of surgery independent of anesthesia (although the fetus and mother are equally anesthetized, the effects of surgery on the mother are likely to be diminished in the fetus), and by duration of anesthesia (if mid third trimester is the period of peak vulnerability, with a plateau of vulnerability extending roughly equally in both temporal directions).

What Might Be Done?

Olney and his group have proposed that anesthetic drug effects on fetal and neonatal γ-aminobutyric acid and N-methyl-D-aspartic acid receptors (NMDAR) cause translocation of the Bcl-2–associated protein to mitochondrial membranes, leading to an apoptotic cascade.63 Perhaps this problem can be alleviated by anesthetic choice in pregnant females. Maze and his group have presented evidence that “xenon mitigates isoflurane-induced neuronal apoptosis in the developing rodent brain,”64 as does dexmedetomidine,65 and xenon is currently in clinical trials for perinatal hypoxic-ischemic brain injury.66 Analogously, Laing and coauthors found that sevoflurane causes less apoptosis than isoflurane, but this difference was not manifest in behavioral tests.22(cf29,30) Several adjunct pharmaceuticals have also shown promise. L-carnitine, an l-lysine derivative that transports long-chain fatty acids into mitochondria, appears to have a beneficial effect in nitrous oxide/isoflurane-damaged neonatal rats,67 and lithium reduces damage from ketamine and propofol in neonatal mice.68 Using the early postnatal rat model, Yon and coauthors found that melatonin reduced anesthetic-induced damage in the most vulnerable brain regions: “Melatonin-induced neuroprotection was mediated, at least in part, via inhibition of the mitochondria-dependent apoptotic pathway since melatonin caused an up-regulation of the antiapoptotic protein, bcl-XL, reduction in anesthesia-induced cytochrome C release into the cytoplasm, and a decrease in anesthesia-induced activation of caspase-3 [precursor of apoptosis].”69 Melatonin has also been shown to protect against learning disorders in hypoxic-ischemic injured neonatal rats70 and excitotoxic brain-lesioned newborn mice.71

Augmentation of another endogenously generated substance with neuroprotective potential, erythropoietin (EPO),72,73 has shown promise against NMDAR-antagonist neurotoxicity in rat74 and mouse75 neonates, in human newborns with hypoxic-ischemic encephalopathy,76 and in extremely preterm human infants.77 What about hypothermia and neonatal brain damage? Creeley and Olney have reported laboratory evidence that hypothermia (30°C) attenuates anesthesia-induced apoptosis in neonatal mice,78 and human trials looking at whole-body hypothermia79 and selective head cooling80 in neonates with hypoxic-ischemic encephalopathy have found a decrease in death and/or moderate-to-severe disability. And magnesium sulfate? Systematic reviews have found a significant reduction in the rate of cerebral palsy in children born to magnesium-treated women at risk of preterm delivery,81,82 and a clinical trial found that postnatal magnesium sulfate treatment improves neurologic outcome for term neonates with severe perinatal asphyxia.83

Have the Data Already Changed Clinical Practice?

How would you answer the following question? A 27-year-old female presents with an operable, benign, slow-growing, barely symptomatic brain tumor. Her neurosurgeon has scheduled the case and estimates an operation time of 4.5 hours. She is 25 weeks pregnant. Would you:

  • A.Use state-of-the-art equipment, procedures, and drugs to proceed with the case?
  • B.Discuss with the neurosurgeon evidence that has emerged or gained renewed recognition since 2003 that 4.5 hours of anesthesia may cause neurodegeneration and persistent learning deficits in the developing brain and leave the decision in his or her hands?
  • C.Discuss the above evidence with the neurosurgeon and the parents and leave the decision in their hands?
  • D.Discuss the above evidence with the neurosurgeon and the parents and, barring development of substantive symptoms, advise postponing surgery until after the patient has given birth or undergone a caesarean section?

Our guess is that prior to Jevtovic-Todorovic and coauthors' 2003 shot-heard-round-the-anesthesia-world,9 most of us were on the A train. In the absence of survey data, our best guess is that most of us would now opt for B, C, or D.

THE OLDER BRAIN

The older brain has less cognitive reserve—less resilience to neurological challenges. Oxidative phosphorylation does not work as well. We have acquired genetic mutations that can alter outcomes. Genetic alleles that were silent when we were young manifest themselves (have phenotypic effects) as we age. And then there is free-radical buildup with reduced levels of scavengers like vitamin C, melatonin, and vitamin E.

Postoperative Cognitive Dysfunction After Noncardiac Surgery

In 1955, P.D. Bedford published “Adverse cerebral effects of anaesthesia on old people.”84 He reviewed 1193 (presumably noncardiac) patients aged >50 years who had received general anesthesia. Mental deterioration in 10% of patients appeared to be long-term or permanent—a figure that concurs with recent findings. Bedford concluded that cognitive decline is related to anesthetic agents and hypotension. He recommended that “Operations on elderly people should be confined to unequivocally necessary cases” and that “postoperative medication should not be a routine matter.” The next major study to report postoperative cognitive dysfunction (POCD) skips ahead 43 years to 1998, the first International Study of Postoperative Cognitive Dysfunction (ISPOCD).85 In noncardiac patients aged >59 years, the incidence of cognitive dysfunction 1 week after surgery was 22% higher than in age-matched controls and 7% higher 3 months after surgery (P < 0.004 for both), with 10% of patients (91 of 910) evidencing POCD (identical to Bedford's finding at a longer postoperative interval). Increasing age, duration of anesthesia, lesser education, a second operation, postoperative infection, and respiratory complications were risk factors for early POCD. However, under a circumstance of significantly reduced statistical power due to a 22% loss of follow-up at 3 months, among the risk factors that were significant in the early postoperative period, only age remained statistically significant.

Monk and colleagues found that 12.7% of elderly (aged >59 years) noncardiac patients had POCD 3 months after surgery86—again, within a narrow confidence interval around Bedford's 1955 report. Corroborating earlier work,87 this study also found a substantial relationship between POCD and death within 1 year of surgery.88,89 Independent risk factors for sustained POCD included greater age, less education, POCD at hospital discharge, and a history of stroke without residual damage. Consistent with many investigations, more education may indicate greater presurgical cognitive reserve, just as prior stroke may indicate presurgical reduction of cognitive reserve.89,90 Notably, Monk's 2008 study did not find duration of anesthesia to be a risk factor. However, the risk of a false-negative conclusion is high, because the sample size of elderly patients at the 3-month measurement was even smaller (308, with 39 POCD patients86) than in the ISPOCD (901, with 91 POCD patients85).

The longest follow-up study of POCD patients (median, 8.5 years) was published by the ISPOCD group in 2009: “Cognitive dysfunction after noncardiac surgery was associated with increased mortality, risk of leaving the labor market prematurely, and dependency on social transfer.”91

Postoperative Cognitive Dysfunction After Cardiac Surgery

Most of us have heard friends or relatives say something like “since he had open-heart surgery, he's not the same he can't think as well, he's not as happy.” The New York Times brought attention to this problem with an article titled “Saving the Heart Can Sometimes Mean Losing Your Memory.”92 In that article, author S. Juhar explained the basics of extracorporeal circulation and discussed reasons for memory loss, focusing on a patient who had gone back to work and found that he had difficulty with his job—a patient who could not perform functions that he had performed for many years. That article raised a great deal of concern, setting the stage for an article published a year later in the New England Journal of Medicine by Newman and colleagues.93 They found POCD in 53% of coronary artery bypass graft (CABG) patients at discharge and in 36% of patients 6 weeks later. That proportion went down to 24% at 6 months after surgery, but came back up to 42% at 5 years after surgery—a pattern of early improvement followed by subsequent decline that was predicted by POCD at discharge.94

Most of us have heard friends or relatives say something like “since he had open-heart surgery, he's not the same … he can't think as well, he's not as happy.”

Aggravating Factors

The factors that cause decline in cognitive capacity among non-CABG patients also affect CABG patients. However, some of those risk factors, like duration of exposure to anesthetics, may be masked by damage done to CABG patients from increased liability to cerebral emboli, cerebral ischemia during reperfusion, and overwarming after bypass.95

Does off-pump versus on-pump make a neurocognitive difference? Several major studies have failed to detect a neurocognitive advantage to off-pump,96,97 and although Shroyer et al. failed to find a statistically significant difference across their composite test battery, they did find a significant difference on one important test in favor of off-pump, suggesting the possibility of a false-negative conclusion.98 More recently, Puskas and colleagues found that “After a mean of 7.5 years of follow-up, patients undergoing off-pump CABG performed better than those undergoing cardiopulmonary bypass in several neuropsychological domains.”99 Less direct evidence came from a study of >16,000 patients in whom a greater incidence of delirium occurred after on-pump cardiopulmonary bypass, with duration of surgery (and so anesthesia) as a significant risk factor.100 Although these patients were not followed up for POCD, Girard and coauthors found that in “mechanically ventilated medical intensive care unit patients, duration of delirium (which is potentially modifiable) was independently associated with long-term [12-month] cognitive impairment.”101 Clearly, a relationship between depression, sedation, delirium, poor neurological outcome, and POCD should not be discounted,102–109 such that off-pump patients may be at lesser risk for POCD.

Inflammation caused by surgical trauma may also aggravate POCD and is associated with the pathogenesis of Alzheimer disease (AD) in a mouse model.110 We know about the up-regulation of interleukin-1, and this in turn can affect anesthetic receptors.111 The ensuing cascade of events ultimately affects the anesthetic γ-aminobutyric acid and NMDARs and increases production of β amyloid and we know that β amyloid, even in nondemented patients, associates with cognitive problems if there is enough of it. Genetic predispositions are another aggravating factor. For example, Matthew and coauthors have shown the contribution of P-selectin and C-reactive protein alleles in modulating susceptibility to cognitive decline caused by inflammation after cardiac surgery,112 but the role of apolipoprotein E4 remains controversial.113

Are anesthetics aggravating factors? If so, are some more toxic than others? Dong et al. found that just 2 hours of clinical anesthesia with isoflurane generates caspase-3 in adult mice,114 and Eckenhoff's group has found that a presenilin-1 mutation associated with familial AD renders PC12 cells more vulnerable to isoflurane cytotoxicity, but not to sevoflurane or desflurane cytotoxicity.115 Jevtovic-Todorovic and Carter have reported that old rat brains are equally sensitive (nitrous oxide) or more sensitive (ketamine with and without nitrous) to anesthetic neurotoxicity than are infant rat brains,116 and Culley and coauthors found that spatial memory is impaired for 2 weeks after 2 hours of 1.2% isoflurane with 70% nitrous oxide in aged rats, an effect that may have a genetic component.117 What about nitrous oxide alone? Culley et al. found that aged rats exposed to 70% nitrous oxide for 4 hours took more time to complete a maze test and made fewer correct choices before making their first error compared with control rats over the following 2 weeks.118 In a separate group of rats, they found that the same nitrous oxide exposure profoundly, but transiently, reduced the activity of cortical methionine synthase—an enzyme that is implicated in dementia and may be related to accumulation of homocysteine (a cytotoxic amino acid normally remethylated to methionine, an essential amino acid, by methionine synthase).118

Evidence from Monk's 2005 study indicated that cumulative deep hypnotic time is associated with more POCD,87 with a substantial relationship between POCD and death within 1 year of surgery.88,89 In distinction, a study by Schubert's group looked at lighter anesthesia (bispectral index [BIS] 50) versus deeper anesthesia (BIS 39) and found that deeper levels of anesthesia were associated with better cognitive function 4–6 weeks postoperatively.119 That finding was subsequently replicated in adult mice120,121 and again in humans.122 Congruent with those findings, Kertai and coauthors concluded that results from the B-Unaware trial “do not support the hypothesis that limiting depth of anesthesia either by titration to a specific BIS threshold or by limiting end-tidal volatile agent concentrations will decrease postoperative mortality.”123 If the association between POCD and mortality is valid and causal,87 the Kertai conclusion123 would also imply that maintaining a high BIS does not reduce POCD. In an editorial accompanying that article, Kalkman and Peelen reminded readers that “ only adequately powered randomized trials can answer the question of whether management aimed at minimizing anesthetic exposure will improve outcomes in vulnerable patients.”124 That is true, but in the absence of such trials, we are left with conflicting evidence.

A study by Schubert's group looked at lighter anesthesia (bispectral index 50) versus deeper anesthesia (bispectral index 39) and found that deeper levels of anesthesia were associated with better cognitive function 4–6 weeks postoperatively.

An Hypothesis

As Lenz125 and others have shown, equilibrated anesthesia does not mean equal anesthesia. Using glucose utilization as a measure of metabolic rate, we can see (Table 1) that anesthetics affect different brain subregions to a greater or lesser extent—with most areas showing a reduction in metabolism, some areas showing no change in metabolism, and a few areas in which metabolism actually increases during anesthesia.

Table 1. Local Cerebral Glucose Use During 1 and 2 MAC of Isoflurane and Sevoflurane Anesthesia in Subregions of Rat Mesencephalon, Diencephalon, and Telencephalon.
  1 MAC1 MAC2 MAC2 MAC
Brain RegionControlIsofluraneSevofluraneIsofluraneSevoflurane
  1. Abbreviations: MAC, minimum alveolar concentration.

  2. Data are expressed as µmol/100 g/min−1 ± SE, N = 6. Adapted from Lenz et al.125

Substantia nigra (compact part)58 ± 3.349 ± 258 ± 3.360 ± 760 ± 4.1
Interpeduncular nucleus83 ± 4.162 ± 13100 ± 7.874 ± 16119 ± 4.5
Medial habenula64 ± 2.559 ± 2.362 ± 3.397 ± 7.384 ± 4.1
Hippocampus CA350 ± 3.347 ± 4.356 ± 4.955 ± 4.368 ± 4.9

If synaptic feedback is key to preventing the apoptotic cascade and nonanesthetized neurons have the energy to apoptose whereas anesthetized neurons do not, nonanesthetized neurons that do not receive sufficient input from connecting (anesthetized) neurons to trigger depolarization might be at greater risk than anesthetized neurons—and there may be fewer nonanesthetized neurons at the interface between anesthetized and nonanesthetized brain regions during deeper anesthesia, because deeper anesthesia may penetrate more deeply into the nonanesthetized subregion, thereby reducing the size (surface area) of the interface (Figure 2).

Figure 2.

Neurons in nonanesthetized cortical subregions (red and black) may be in a sort of reverse penumbra—a penumbra where the cluster in the center is viable because each neuron is sufficiently connected to nonanesthetized neurons to maintain a local-talk network. But farther away from the core, at the interface with anesthetized subregions (green), there may be nonanesthetized neurons (black, with dendrites connected to anesthetized neurons) that do not receive sufficient input to generate an action potential in response. Neurons with dendrites that connect to anesthetized neurons would be in a state of nonanesthetized inactivity, which may be physiologically worse than being anesthetized during a period of inactivity because, unlike anesthetized neurons, nonanesthetized neurons retain the capacity to transcribe and translate enzymes needed to undergo apoptosis (apoptosis is an active process, requiring substantial metabolic activity). Indeed, a state of prolonged nonanesthetized inactivity at the interface between an anesthetized subregion and a nonanesthetized subregion may be analogous to the state of a neuron in a developing fetus that fails to make sufficient functional dendritic and axonal connections, and so undergoes apoptosis. The conjecture here is that there would be tens of thousands to tens of millions of such neurons at the interface between anesthetized and nonanesthetized cortical subregions, depending on the size of the nonanesthetized subregion and the depth of anesthesia.

In this conjectured cortical subregion scenario, a nonanesthetized neuron whose dendrites synapse with anesthetized axons is in solitary confinement. Eventually, that nonanesthetized neuron may kill itself. The nonfunctional fetal neuron and the nonanesthetized adult neuron may undergo apoptosis for the same reason—insufficient functioning connections—but in the case of the fetal neuron, that is because the fetal neuron has not made enough functional connections, whereas for the nonanesthetized neuron, too many of its anesthetized neighbors have become nonfunctional. Whether a neuron finds itself in solitary confinement because it fails to make enough connections (fetal neuron), or because it is stuck in a neighborhood from which most of its neighbors appear to have left (nonanesthetized neuron with dendrites in an anesthetized brain region), the effect may be the same. Much is at stake here, and there is much to learn. As put by Farag et al, “Our observations highlight the need for further studies to better understand the contribution of perioperative management to POCD.”119

The nonfunctional fetal neuron and the nonanesthetized adult neuron may undergo apoptosis for the same reason—insufficient functioning connections—but in the case of the fetal neuron, that is because the fetal neuron has not made enough functional connections, whereas for the nonanesthetized neuron, too many of its anesthetized neighbors have become nonfunctional.

Regional Versus General Anesthesia

The above perspective may account for the frequently observed lack of difference in POCD, or the weakness of the difference in POCD, between patients who receive general anesthesia and patients who receive regional anesthesia with sedation.126,127 Ancelin found that “Adding sedation to peridural anaesthesia led to a decline in verbal secondary memory,”128 and Sieber et al. found that lighter sedation during spine surgery led to less delirium.129 Again, there are empirical and neuropathological reasons to suspect a link between delirium, deep sedation, poor neurological outcome, and POCD101–109 and there is evidence that patients with the apolipoprotein epsilon4 allele experience postoperative delirium at more than twice the rate of patients without apolipoprotein epsilon, adding a genetic link to account for some of the variability.130 If the association between POCD and deep sedation had been discovered before the association between POCD and general anesthesia, perhaps we would have come more readily to the hypothesis that lighter anesthesia119–122 and deeper sedation128,129 occupy an equivalent middle ground (approximately BIS 45–55) when it comes to increasing the risk of POCD relative to both lighter sedation (approximately BIS ≥60) and deeper anesthesia (approximately BIS 35–45).131

If the association between POCD and deep sedation had been discovered before the association between POCD and general anesthesia, perhaps we would have come more readily to the hypothesis that lighter anesthesia and deeper sedation occupy an equivalent middle ground (approximately BIS 45–55) when it comes to increasing the risk of POCD relative to both lighter sedation (approximately BIS ≥60) and deeper anesthesia (approximately BIS 35–45).

Anesthesia and Neurodegenerative Diseases

What about the effects of anesthetics on the neurodegenerative diseases? Hydrophobic cavities keep sticky proteins from becoming irreversibly glued together. Unfortunately, molecules of inhalational anesthetics can fill those cavities and reduce the amount of energy required to maintain protein assembly.132 This anesthesia-facilitated disinhibition of protein binding helps monomers aggregate into oligomers, and if those monomers are amyloid β (Aβ), the resulting oligomerization can lead to protofibrils that are small enough to diffuse into neurons and large enough to be neurotoxic (Figure 3). Amyloid β oligomers appear to contribute to the neurodegeneration characterized by AD in the early 20th century. By the middle of the 21st century, 13 million Americans are projected to have AD. Many of them will need to be anesthetized, and many of them will have been anesthetized before they became demented.

Figure 3.

Possible mechanisms by which inhaled anesthetics, through increasing intracellular amyloid-β (Aβ) level, and Aβ and τ aggregation, and/or disruption of intracellular calcium homeostasis, could induce synaptic dysfunction and neuronal apoptosis and ultimately produce cognitive decline in the aged brain. (A) In the amyloid and τ pathway, β-site amyloid precursor protein (APP)-cleaving enzyme (BACE) generates C-terminal fragments (CTFβ) from membrane-bound APP. Cleavage of CTFβ by γ-secretase releases Aβ monomers into the cytosolic and extracellular space. Anesthetic exposure increases the levels of BACE and γ-secretase, thereby increasing the levels of intracellular and decreasing CTFβ levels. Inhaled anesthetics also interact with the Aβ monomers to promote the formation of small soluble oligomers. These oligomers further associate to form fibrils and extracellular plaque, which have been found to be increased in mice with transgenic Alzheimer disease after exposure to halothane. These effects activate caspase, initiating apoptosis, and cleaving the adaptor protein GGA3, which is required for BACE lysosomal degradation. This results in increased BACE levels, further enhancing the production of Aβ, and introducing a vicious cycle that ensures apoptosis. Microtubule (MT)-bound τ becomes hyperphosphorylated and detached by anesthetics and hypothermia, resulting in τ aggregates and decreased MT stability. (B) On the right side of the figure, anesthetics increase cytosolic calcium via several mechanisms. For example, inhaled anesthetics activate the endoplasmic reticulum (ER) membrane inositol 1,4,5-trisphosphate receptors (IP3R) and ryanodine receptors (RyR), increasing cytosolic calcium and depleting ER calcium. These drugs also activate the sarcoplasmic/ER calcium adenosine triphosphatase (ATPase) (SERCA1), further enhancing the activity of ER calcium release pathways. Further increases in cytosolic calcium levels might be caused by activation of N-methyl-D-aspartate receptors, and inhibition of calcium clearance via plasma membrane calcium ATPase. Increased cytosolic calcium loads the mitochondria with calcium, releasing cytochrome c, further contributing to apoptosis. Finally, ER calcium depletion via the above mechanisms can induce apoptosis directly. Both the Aβ/τ and calcium pathways contribute to synaptic dysfunction and apoptotic responses (from Tang et al., 2010).132Abbreviations: Aβ, amyloid beta; APP, amyloid precursor protein; BACE, β-site APP-cleaving enzyme; Ca, calcium; CTF, C-terminal fragments; ER, endoplasmic reticulum; GGA3, Golgi-localized, gamma adaptin ear-containing, ARF-binding protein 3; IP3R, inositol 1,4,5-trisphosphate receptors; Mito, mitochondria; MT, microtubule; NMDAR, N-methyl-D-aspartate receptors; RyR, ryanodine receptors; SERCA, sarcoplasmic/ER calcium adenosine triphosphatase.

The role of inhalational anesthetics in the above scenario has been verified in vitro by a decade of work from Eckenhoff and coauthors,132 now supported in vivo by mouse models.114,133 In addition to the Aβ-anesthesia connection, Xie's group has utilized human neuroglioma cell cultures to add anesthesia-induced apoptosis as a factor contributing to AD.114,134–136 Do the rodent and cell-culture findings apply to humans? Eckenhoff's group recently reported that the total-τ/amyloid-β (1–42) ratio in cerebrospinal fluid, the only biomarker validated for use in the diagnosis of AD by the Alzheimer's Disease Neuroimaging Initiative (ADNI), elevates during surgery and anesthesia in healthy patients and rises above ADNI's threshold for mild cognitive impairment within 48 hours.137 In an article titled “Coronary artery bypass surgery provokes Alzheimer's disease–like changes in the cerebrospinal fluid,” Palotas and colleagues found an increased τ/amyloid-β ratio in patients 6 months after surgery.138

Results from retrospective studies remain inconclusive, but are unsettling. Examining records of 9170 veterans, Lee et al. compared the risk of developing AD within 5–6 years of CABG surgery under inhalational anesthesia versus the risk of developing AD within 5–6 years of percutaneous transluminal coronary angioplasty, the latter seldom requiring general anesthesia.139 After adjustment for age, length of hospital stay, comorbidity, and number of procedures, the CABG patients developed AD at nearly twice the rate of percutaneous transluminal coronary angioplasty patients (hazard ratio: 1.71, P < 0.04). Yes, CABG patients faced more predisposing factors than percutaneous transluminal coronary angioplasty patients, including embolic ischemia, but given in vitro evidence supporting mechanisms for a causal link between anesthesia and AD, it would be reckless to dismiss prolonged inhalational anesthesia as an independent contributing factor to Lee and coauthors' finding.

Bohnen and coauthors performed a case-controlled retrospective study of 252 AD patients.140 Unfortunately, 199 of the 252 controls (non-AD patients) had prior exposure to general anesthesia, which greatly dilutes their statistical power to evidence an effect of anesthesia relative to AD patients. Nevertheless, Bohnen et al. found non–statistically significant effects in the direction of a link between AD and general anesthesia on each of 3 independent variables: cumulative exposure to anesthesia, exposure to ≥6 episodes of general anesthesia (odds ratio: 1.44), and cumulative exposure to ≥600 minutes of general anesthesia (odds ratio: 1.63). Gasparini and coauthors also performed a retrospective case-controlled study of AD patients. In their study, the controls were Parkinson disease patients and patients with other neurological diseases141—but a link between general anesthesia and other neurological diseases has been hypothesized,132 such that lack of a difference in anesthetic exposure between AD patients and patients with Parkinson or other neurological diseases does not imply a lack of a deleterious effect of anesthetic exposure.

Although a connection between anesthetics and AD has received more attention than a possible relationship between anesthesia and Parkinson or Huntington disease, 2 investigations suggest that further research is warranted. Peretz and coauthors have found evidence that supports an elevated risk of Parkinson disease among anesthesiologists as compared with internists,142 and Wang et al. have found in vitro laboratory evidence that isoflurane may exacerbate Huntington disease.143

So there is evidence that anesthetics are a particular problem for older patients. But before we make recommendations more firm than Bedford's admonition from 1953 that “Operations on elderly people should be confined to unequivocally necessary cases,”84 we need to know more about genetic profiles in order to know which older patients are most at risk.

POTENTIAL ALLEVIATING FACTORS

How might we reduce the risk of POCD in older patients? Are some anesthetics less deleterious than others? Crosby's group has presented data indicating that “In aged rats, propofol anesthesia is devoid of the persistent memory effects observed with other general anesthetic agents in this model. Thus, it appears that general anesthesia–induced memory impairment may be a function of the agent rather than the anesthetic state itself.”144 In a hippocampal slice model, desflurane was more protective than propofol,145 and in a rat cardiopulmonary bypass model, isoflurane with 60% xenon has been shown to prevent the decrement in neurocognitive function caused by bypass under isoflurane alone.146 The protective potential of xenon is being evaluated in humans by Maze's group, and preliminary results are encouraging.147 Complementary in vitro work by Wei and Xie also suggests that sevoflurane and desflurane are less-potent triggers of apoptosis than isoflurane.148

One of our department's laboratories has investigated the effects of lidocaine during global ischemia in rats, finding that neuron death in the hippocampus is substantially reduced in animals that have received clinically relevant doses of lidocaine. Function was also better retained after global ischemia in animals that received lidocaine.149,150 Looking at CABG patients, Wang and coauthors found that lidocaine (1.5 mg/kg bolus followed by a 4-mg/minute infusion during operation and 4 mg/kg in the priming solution of cardiopulmonary bypass) reduced POCD measured 9 days after surgery.151 Looking at a larger number of CABG patients, Mitchell and colleagues also found reduced POCD in patients who received lidocaine—from 75% to 40% at 10 days (P < 0.025), from 75% to 46% at 10 weeks (P < 0.05), and then from 48% to 28% at 6 months (not significant).152,153 Most recently, Newman's group at Duke reported a significant reduction in POCD at 6 weeks and 1 year among nondiabetic cardiac patients.

Newman's group at Duke reported a significant reduction in POCD at 6 weeks and 1 year among nondiabetic cardiac patients.

This effect was most pronounced in nondiabetic patients who received <43 mg/kg lidocaine (total dose), whereas lidocaine appears to have had a deleterious effect in diabetic patients and in patients who received higher total doses.154

What about EPO, melatonin, memantine, and statins in the elderly? Xiong et al. report that delayed administration of EPO reduced hippocampal cell loss, enhanced angiogenesis and neurogenesis, and improved functional outcome after traumatic brain injury in rats.155 Genc has reviewed reasons why EPO may be therapeutic in Parkinson patients,156 and Lauretani and colleagues found that EPO levels are lower in patients aged 60–98 years with impaired peripheral nerve function and/or a clinical diagnosis of polyneuropathy.157 Haljan et al. found a trend toward improved neurocognitive recovery with EPO use in CABG patients,158 and in post-hoc analyses Tseng et al. found EPO to be protective in subarachnoid hemorrhage patients who are younger, nonseptic, and on statin therapy.159 Cheng and colleagues' review of the beneficial effects of melatonin in experimental models of AD is encouraging,160 and a clinical study by Furio et al. found that melatonin improved cognitive function in elderly outpatients who suffered from mild cognitive impairment.161 The jury is still out on the neuroprotective potential of memantine in Alzheimer patients,162 but laboratory evidence remains encouraging.163 The jury has also looked hard for evidence that statin therapy prevents or ameliorates AD, but a definitive verdict is still pending.164 Nevertheless, because there are other important reasons for the elderly to take statins, we can expect to need to anesthetize many AD patients who are on statins. If the evidence from stroke165 and CABG166 patients applies, statin therapy should not be withdrawn without specific indication for withdrawal.

Preconditioning

Although fetuses and the elderly are particularly sensitive to ischemia, hypoperfusion, and hypoxia, “Nietzsche's Toxicology: whatever doesn't kill you might make you stronger”167 could lead to improved clinical management of patients with fragile brains. In 1964, Dahl and Balfour published evidence of “prolonged anoxic survival due to anoxia pre-exposure.”168 This phenomenon was eventually replicated in a model of cerebral ischemia,169 and induction of endogenous proteins of repair and the genes that code for them are now well documented. Our laboratory has added sevoflurane as a potential preconditioner,170 and Maze's group reported that in comparison with sevoflurane,171 nitrous oxide, and hypoxia,172 xenon preconditions in a manner that “might mimic the intrinsic mechanism of ischemic preconditioning most closely.” But if a limited dose of anesthesia triggers the same protective mechanisms as a limited bout of hypoxia, how much anesthesia can we give before what would have been a protective effect becomes a deleterious effect on balance?148

Although fetuses and the elderly are particularly sensitive to ischemia, hypo-perfusion, and hypoxia, “Nietzsche's Toxicology: whatever doesn't kill you might make you stronger” could lead to improved clinical management of patients with fragile brains.

Clinically acceptable means of accomplishing cerebral preconditioning are being sought. Volatile anesthetics notwithstanding, pharmacological cerebral preconditioning may be eclipsed by mechanical remote ischemic preconditioning (RIPC). Several clinical studies have demonstrated that three 5-minute inflations of a blood pressure cuff to 200 mm Hg around a patient's upper arm, followed by 5-minute intervals of reperfusion, improves outcome after cardiovascular procedures,173–176 and evidence from laboratory investigations indicates that the same technique initiated prior to neurosurgery may improve outcome.177–180 Clinical studies of RIPC in neurosurgical patients are underway or have recently been completed,181 and a published study by Hu and colleagues has reported reduced biochemical markers of neuronal ischemia and improved rate of recovery after cervical decompression in patients who received RIPC.182

Clinical studies of remote ischemic preconditioning in neurosurgical patients are underway or have recently been completed, and a published study by Hu and colleagues has reported reduced biochemical markers of neuronal ischemia and improved rate of recovery after cervical decompression in patients who received remote ischemic preconditioning.

Neurogenesis

The old adage that neurogenesis is only for the young was shown to be wrong for rodents in 1965,183 is known to be wrong for nonhuman primates,184 and is almost certainly wrong for humans.185 This raises the possibility that negative effects of surgery and anesthesia on the elderly, as well as the very young, can be compensated for by therapies that strengthen the neurogenic response. Results in rats encourage the conclusion that “neural precursors resident in the brain initiate a compensatory response that results in the production of new neurons. Moreover, administration of growth factors can enhance this compensatory response [and] we may eventually be able to manipulate these precursors to improve recovery of function.”186,187 In addition to ischemic preconditioning,188 granulocyte-colony stimulating factor189 and EPO176,190,191 appear to be such manipulators, and neurogenesis may be the mechanism of electroconvulsive therapy in patients with depression.192,193

The old adage that neurogenesis is only for the young was shown to be wrong for rodents in 1965, is known to be wrong for nonhuman primates, and is almost certainly wrong for humans.

Unfortunately, however, as noted above, Culley and colleagues have presented evidence that 1 MAC isoflurane reduces the production of neural stem cells by 20% in vitro 24 hours after exposure.32 Noting that “a 15%–20% decrease in neurogenesis in vivo is sufficient to impair hippocampal-dependent memory in rodents,194 and even more striking, recall of remote spatial memory in adult animals depends on recruiting as few as 1%–4% newly born neurons into hippocampal circuits195 even a transient adverse effect of isoflurane on self-renewal of NSCs (neural stem cells) might have far-reaching consequences for brain development and function across the life-span.” If so, this raises concern about the effect of all anesthetics on neural stem cell proliferation “at critical periods of brain development as well as in adulthood.”32

CONCLUSION

Reports of possible adverse cognitive effects of anesthetics on young patients appeared in our literature back in the 1940s, on elderly patients in the 1950s, and on fetuses in the 1980s. These problems and some of their potential solutions are not new, but our awareness of them has experienced a renaissance over the past decade. Fortunately, the vast majority of anesthetics delivered to infants last <1.5 hours or are administered to children aged >6 months. For that majority, we agree that the “evidence is most consistent with the premise that ‘anesthesia per se,’ given to an otherwise healthy child who needs only a ‘routine’ surgical procedure, is not neurotoxic”196 or is not toxic enough to cause a currently measurable adverse effect. However, for children aged <6 months, fetuses of any age (given Hogan's recent finding8), and patients aged> 60 years, until and unless we are able to classify substantive anesthetic neurotoxicity as a rare complication, the conservative first-do-no-harm approach should: (1) count anesthesia on the cost side of the cost/benefit equation when making decisions about whether and when to proceed with surgery; (2) avoid nitrous oxide, isoflurane, and ketamine, because multiple laboratory studies indicate that they are particularly toxic (with one study indicating that desflurane may join that group30); (3) limit the duration of continuous anesthesia to <2 hours whenever possible; and (4) consider the possibility that deep sedation may trigger as much neuronal apoptosis as light anesthesia.

At the very least, newfound concerns generated by available data should inspire a great deal of translational research. If that research is funded, our guess is that we will soon have anesthetic, sedative, and adjuvant drugs ranked according to their deleterious effects and cerebral preconditioning with augmentation of endogenous processes of regeneration will deliver brain protection and recovery to the very young, the old, and everyone in between—before the younger among us are too far gone to benefit!

Acknowledgements

Some of the content of this article was presented as a Refresher Course Lecture at the Annual Meeting of the American Society of Anesthesiologists, Chicago, IL, on October 15, 2011.

DISCLOSURES

Potential conflict of interest: Nothing to report.

Ancillary