MILLIONS OF CHILDREN UNDERGO SURGERY with anesthesia every year.¹ During the perioperative period they are exposed to a multitude of stressors capable of interfering with normal brain development. Pain, stress, inflammation, hypoxia, and ischemia have all previously been shown to adversely affect the immature central nervous system. However, recent findings from animal studies have indicated that sedatives and anesthetics—the very drugs used to reduce pain and stress—may themselves undesirably influence brain development by triggering structural and functional abnormalities. There is now an extensive body of work, mostly based on laboratory research, that has defined this phenomenon and explored mechanisms and protective strategies. However, translating these laboratory findings to humans in clinical settings is laden with uncertainties and questions.

Human epidemiologic studies have found mixed evidence for an association between surgery (and anesthesia) in childhood, and subsequent neurodevelopmental delay. Some studies identified more learning disabilities in children after surgery with general anesthesia early in life, whereas others have not. Although surgery or comorbidities may play a central role in these abnormalities, emerging laboratory data regarding the deleterious effects of anesthesia without surgery has forced clinicians to consider the possibility that anesthetics may play a role in this phenomenon. However, similar to the uncertainties surrounding animal studies, the interpretation of the human data is fraught with substantial limitations.

Any discussion of the long-term effects of anesthetics on the developing brain is further complicated by the fact that anesthetic drugs may be neuroprotective under certain conditions, and may indeed mitigate brain damage that is due to inflammatory responses, hypoxia-ischemia, or other insults that might occur during the perioperative period.

It is currently impossible to provide any definitive statements about the effects of anesthesia on neurodevelopment. This chapter provides an overview of the current laboratory and clinical data concerning the affect of sedatives, anesthetics, and analgesics on the immature brain.

Background

About 160 years ago, William T. G. Morton conducted the first public demonstration of a drug-induced, reversible coma, later termed anesthesia, at the Massachusetts General Hospital and instantaneously revolutionized the field of surgery. After witnessing Morton’s demonstration, Oliver Wendell Holmes coined the name for what he saw from the greek words “an” (without) and “esthesia” (sensibility). The inscription on Morton’s tombstone, “Inventor and Revealer of Inhalation Anesthesia: Before Whom, in All Time, Surgery was Agony; By Whom, Pain in Surgery was Averted and Annulled; Since Whom, Science has Control of Pain,” represents a powerful testament to the tremendously positive impact that anesthesiology has made on the field of medicine. This positive impact is now tempered by the possibility that anesthetics may impede normal brain development in the young. During the first century of their use, general anesthetics quickly became ubiquitous during surgery, although they were also regarded with serious concern because they were combustible and they depressed the hemodynamic and respiratory systems. Accordingly, up until 25 years ago, general anesthesia was rarely used in critically ill neonates because of the fear of myocardial depression and hemodynamic instability. Perioperative drug regimens were limited by some to neuromuscular blocking drugs and nitrous oxide. However, with the realization that unopposed pain exerted deleterious effects on the developing brain, that dramatic stress responses to painful stimulation can be detected even in preterm infants, and that modern anesthetics and analgesics can abolish these responses without substantial hemodynamic compromise, pediatric anesthesia, for the past two decades, has afforded critically ill neonates the benefits of amnesia, analgesia, and immobility during increasingly invasive surgery. These surgical interventions have helped to save lives and preserve the quality of life in one of the most vulnerable of patient populations. All the while, the powerful effects of general anesthetics on the brain were thought to be limited to the duration of immediate exposure, and no serious long-term adverse effects were expected following emergence from anesthesia. This notion of safety is now being seriously questioned, because apoptosis (defined as programmed cell death, see later) has been detected in neonatal animals during and immediately after anesthesia, and long-term learning defects have been reported in animals and children that were exposed to anesthesia at a young age.

Normal Brain Development

The human brain undergoes a complex and extended process of enormous growth in cell number, synapses, and connections. Combined with this expansion of cells and connections, massive regressive processes also occur during normal brain development. These expanding and regressive processes allow the brain to fully develop and eventually execute complex tasks, such as talking, walking, reading, writing, calculating, acquiring social skills, perfecting fine motor dexterity, and planning and executing long-term objectives, while at the same time maintaining all life-sustaining functions. In order to accomplish these tasks, several specialized cells populate the brain. These cells are commonly divided into two large groups: neurons and glial cells, the latter subdivided into astrocytes and oligodendrocytes. Neural development starts with proliferation from a neural stem cell, followed by migration of that cell to its final place, differentiation into a specialized neuron, outgrowth of synapses during synaptogenesis to integrate into neural networks, and, finally, myelination of axons. In humans, many of these specialized processes occur in utero, but will also continue for an extended period of time postnatally. Indeed, some brain regions exhibit lifelong neurogenesis.

In utero and during the early postnatal period, the human brain initially undergoes a rapid growth in size and cell number, and is then pared back to achieve the efficient network of about 100 billion neurons of the adult brain. At birth, the size of the immature brain is one-third that of the adult brain, doubling in size within the first year of life, and reaching 90% of its eventual size by 6 years of age.² This dramatic growth spurt coincides with a remarkable overabundance of neurons and neuronal connections. In fact, less than half the neurons generated during development survive into adulthood.^3,⁴ Superfluous neurons that lose in the competition for a limited amount of trophic factors are removed by programmed cell death. Also called apoptosis, this cellular suicide program is built into every mammalian cell. Apoptotic cells enter a well-orchestrated, stepwise, and energy-consuming destruction process that involves a cascade of enzymes called caspases and ultimately leads to the breakdown of cellular proteins and DNA.⁵ This process is heavily used during development, such as during embryonal deletion of the interdigital mesenchymal tissue to separate fingers and toes. Similar to the clotting cascade, the apoptotic cascade remains active throughout life and is held in check by antiapoptotic factors, whereas proapoptotic factors promote its execution. Neurons are protected from apoptosis by neurotrophic factors. Conversely, noxious stimuli, such as pain, hypoxia, and ischemia, can increase the level of proapoptotic factors and activate the apoptotic suicide program, leading to neuronal cell death.

In addition to an overabundance in the total number of neurons generated during early development, the mammalian brain also forms an excess of neuronal connections, or synapses, during this period. Depending on brain region, synaptic density reaches its maximum in infants and young toddlers between 3 and 15 months of age, and will undergo a progressive reduction by about half into adulthood.⁶ Connections that are active with continued electrical and chemical signals are sustained, whereas those with little or no activity are lost.

In summary, brain architecture changes rapidly and dramatically throughout life. Neuronal density is greatest during fetal life, and excess neurons are eliminated via apoptosis, predominantly in utero, during the neonatal period and throughout infancy.⁷ Rapid growth of dendrites and synaptic connections occurs during infancy and early childhood, and unneeded dendrites and synapses are trimmed back, predominantly during later childhood and adolescence.^6,⁷

Accordingly, the first several years after birth represent a critical period of development for many brain regions. Recent findings in animals suggest that exposure to anesthetics or sedatives may interfere with proper neuronal development, brain architecture, and subsequent function. Although the exact molecular mechanisms by which anesthetics afford their therapeutic properties of amnesia, analgesia, and immobility are only incompletely understood, their interaction with a wide variety of ion channels, such as sodium, calcium, and potassium channels, as well as several cell membrane proteins, including the receptors for γ-aminobutyric acid (GABA), glycine, glutamate (and N-methyl-d-aspartate [NMDA]), acetylcholine, and serotonin, make it conceivable that anesthetics could interfere with normal electrical and chemical activity in the developing brain. In fact, both GABA and NMDA play critical roles as trophic factors and in regulating neuronal maturation and programmed cell death. During brain development, GABA directs cell proliferation, neuroblast migration, and dendritic maturation.⁸ Developmental NMDA receptor stimulation fosters survival and maturation of some neurons.^9,¹⁰ It is therefore not implausible that anesthetics might interfere with these developmental processes.

Effects of Anesthetic Exposure on the Developing Brain

Concerns regarding neurologic abnormalities following general anesthesia in young children were first raised more than half a century ago, when postoperative personality changes were observed following the administration of vinyl ether, cyclopropane, or ethyl chloride for otolaryngologic surgery.¹¹ However, these abnormalities were felt to be psychological in nature because they were alleviated by the timely administration of preoperative sedative drugs.^11,¹² Approximately two decades later the focus of research shifted to examine the effects of anesthetics in animal models that represented occupational exposure in pregnant healthcare workers.¹³^–¹⁶ Delayed synaptogenesis and behavioral abnormalities were observed in neonatal rats born to dams that were chronically exposed to subanesthetic doses of halothane during their entire pregnancy. However, interest in the effects of anesthetics on children did not elicit widespread interest until publications began to appear 10 years ago. In a seminal study in neonatal rat pups, widespread neuronal degeneration was observed after repeated injections of ketamine.¹⁷ This led to numerous editorials and review articles, and more than 200 publications in which structural brain abnormalities and/or functional impairment were demonstrated after a wide variety of immature animal species (including chicks, mice, rats, guinea pigs, swine, sheep, and rhesus monkeys) were exposed to almost every sedative and anesthetic in clinical use.¹⁸^–⁶⁵ However, no discussion about the effects of drug exposure on the developing brain would be complete without examining the effects of opioid analgesics in the developing brain.^54,⁵⁵ Accordingly, this chapter will examine the various specific effects of sedatives, anesthetics, and analgesics in the immature brain.

Apoptotic Cell Death

The most widely studied deleterious consequence of exposure to sedatives or anesthetics in immature animals is apoptosis, or programmed cell death. Although neuronal apoptosis eliminates approximately 50% to 70% of neurons throughout the brain during development, at any particular time, this natural process affects only a small fraction of cells. Exposure to anesthetics or sedatives briefly, but dramatically, increases the number of apoptotic neurons (Fig. 23-1). Some studies demonstrated up to a 68-fold increase in the density of degenerating neurons after a combination of anesthetics in neonatal rats, compared with control animals, although it remains unclear what fraction of the entire neuronal population these degenerating neurons represent. Unpublished data from one of the authors’ laboratory suggest that a 6-hour exposure to a clinically relevant dose of isoflurane triggers apoptotic cell death in 2% of neurons in the superficial cortex of neonatal mice, whereas less than 0.1% of neurons undergo physiologic apoptosis in this region in unanesthetized litter mates (Loepke, personal communication). Interestingly, dying neurons are immediately adjacent to seemingly unaffected neighboring cells (Fig. 23-2), and the exact mechanism and selectivity of the cell death process remains unknown. Increased neuroapoptosis has also been observed following either in vitro or in vivo, exposure to a wide variety of sedatives and anesthetics, including chloral hydrate, clonazepam, diazepam, midazolam, nitrous oxide, desflurane, enflurane, halothane, isoflurane, sevoflurane, ketamine, pentobarbital, phenobarbital, propofol, and xenon, as well as opioid receptor agonists, in a wide variety of species, including chicks, mice, rats, guinea pigs, piglets, and rhesus monkeys (E-Table 23-1). Cell death–selective stains, such as cupric silver and Fluoro-Jade (EMD Millipore, Billerica, Mass.), have confirmed the cellular demise in neurons that positively stain for activated caspase 3, the central executioner enzyme of the apoptotic cascade.

FIGURE 23-1 Prolonged exposure to anesthetics causes widespread apoptotic cell death in the developing animal brain. Representative photomicrographs of brain sections from 7- to 8-day-old mice exposed to 6 hours of equipotent doses of 7.4% desflurane, 1.5% isoflurane, 2.9% sevoflurane in 30% oxygen, respectively, or fasted, unanesthetized litter mates in room air (no anesthesia). These doses were previously determined to represent 0.6 minimum alveolar concentration for the respective anesthetics. Arrows mark clusters of cells in layer II/III of neocortex that are dying from programmed cell death and are therefore labeled for the apoptotic marker, activated caspase 3 (bright green). Scale bar = 500 µm.

(From Istaphanous GK, Howard J, Nan X, et al. Comparison of the neuroapoptotic properties of equipotent anesthetic concentrations of desflurane, isoflurane, or sevoflurane in neonatal mice. Anesthesiology 2011;114:578-87.)

FIGURE 23-2 Neurons affected by apoptotic cell death following anesthetic exposure are surrounded by seemingly unaffected cells. Representative photomicrograph from a 7-day-old mouse following a 6 hour exposure to 1.5% isoflurane, showing neocortical cells stained for the apoptotic cell death marker activated caspase 3 (blue, top), the neuronal marker NeuN (red, middle), and a merged image of the two stains (bottom). The majority of dying cells are identified as postmitotic neurons, as indicated by the purple cells coexpressing activated caspase 3 and NeuN in the bottom (arrows), whereas some cells expressing caspase 3 are NeuN-negative (arrowheads). Importantly, cells affected by anesthetic-induced neuroapoptosis are surrounded by numerous, seemingly unaffected neurons. Scale bar = 10 µm.

(Image courtesy the Loepke Laboratory.)

E-TABLE 23-1 Preclinical Studies Examining Structural and Neurocognitive Effects of Anesthetic Exposure Early in Life

Apoptosis represents an inherent, energy-consuming process using a cascade of enzymes called caspases. Apoptosis is highly conserved among species and culminates in self-destruction and elimination of cells, even under physiologic conditions, when these cells are functionally redundant or potentially detrimental to the organism.²⁰⁸ Apoptosis involves an orderly breakdown of the cell that includes chromatin aggregation, nuclear and cytoplasmic condensation, and partitioning of cytoplasmic and nuclear material into apoptotic bodies for subsequent phagocytosis, without an extensive inflammatory response. That contrasts with the features observed during necrosis, which include energy failure, cellular swelling, membrane rupture, and release of cytoplasmic content into the extracellular compartment, followed by an inflammatory response.²⁰⁸ Accordingly, apoptosis has also been termed cellular suicide and is extensively used during tissue homeostasis, endocrine-dependent tissue atrophy, and normal embryogenesis (e.g., ablation of tail tissue as part of tadpole metamorphosis in amphibians). Similarly, brain cells are also produced in excess during normal brain development, and up to 50% to 70% of immature neurons are eliminated during normal brain maturation in rodents, nonhuman primates, and humans.^3,⁴ Accordingly, physiologic apoptotic cell death is critical to establish proper brain structure and function, and any disruption of this process can lead to massive brain malformations and intrauterine demise.²⁰⁹ However, in addition to this intrinsic pathway of physiologic, developmental apoptosis, cell death can also be triggered by pathologic, extrinsic factors, such as hypoxia and ischemia.²¹⁰ It currently remains unknown whether anesthesia-induced neuroapoptosis accelerates physiologic programmed cell death or whether it eliminates cells not destined to die, as in pathologic apoptosis.

Animal studies have identified a narrow window of maximum susceptibility to neuronal cell death induced by several anesthetic drugs, such as the NMDA antagonist ketamine, the GABA agonist isoflurane, or ethanol (a combined NMDA antagonist and GABA agonist). Ketamine-induced neuronal demise occurs in neonatal rodents between 5 and 7 days of age, or before 6 days of age in monkeys, but not in older animals.^17,^69,¹³⁹ Similarly, neurotoxicity was not detected after isoflurane anesthesia in 1-day-old animals or in those more than 10 days of age, whereas neuroapoptosis reached a maximum effect at 7 days of age.⁸⁷ However, preliminary data from the laboratory of one of the authors challenges the notion that anesthetic-induced neuroapoptosis is limited to this age range (Loepke and associates, unpublished data). Conversely, intrauterine exposure to clinical doses of isoflurane in prenatal rats may actually decrease physiologic apoptosis and improve subsequent memory retention,⁹⁶ whereas only supraclinical doses of isoflurane induced neuroapoptosis in this setting.¹¹²

Long-Term Brain Cellular Viability, Neurologic Function, and Behavior

In order to answer the important question whether anesthetics simply hasten physiologic apoptosis or whether they induce pathologic apoptosis, long-term neuronal density and neurologic function have to be assessed in adult animals exposed to anesthesia as neonates. If exposure to anesthetics or analgesics only temporarily accelerated physiologic apoptosis, one would expect normal cell counts and function in adult animals. Conversely, a permanent neuronal cell loss and long-term neurocognitive impairment after anesthetic exposure early in life would suggest that anesthesia-induced neuronal apoptosis may be pathologic in nature. However, this would only be true if the organism was unable to compensate for the neonatal cell loss by increasing neuronal plasticity and repair. To answer these questions, several studies measured neurologic function, assessed behavior, and/or determined neuronal density in adult animals after they were exposed to anesthesia in the neonatal period. Unfortunately, results from these studies are conflicting. Several studies reported long-term neurocognitive or behavioral abnormalities after exposure of neonates to enflurane, halothane, isoflurane, sevoflurane, propofol, or ketamine, or to a combination of isoflurane, nitrous oxide, and midazolam.* Importantly, however, many of these studies only observed abnormalities in very specific tests or subsets of neurocognitive batteries, whereas many other neurobehavioral domains remained intact. For example, a 6-hour exposure to midazolam, isoflurane, and nitrous oxide in neonatal rats led to transient impairment in a water maze learning task in young adulthood and in older animals, whereas, in the same animals, several other tests of behavior and learning, including acoustic startle response, sensorimotor tests, spontaneous behavior in an open field, and learning and memory in the radial arms maze, remained unimpaired.⁸⁵ Similarly, after a 4-hour exposure to the minimal alveolar concentration (1 MAC) of isoflurane in 7-day-old rats, long-term memory retention was abnormal at two time points, whereas performance at several other time points, as well as in other tests of learning and memory, remained intact.¹⁰⁷ Accordingly, the relevance of these limited learning deficits remains uncertain. Similar to humans, performance of rodents and primates in learning tasks depends to a great extent on maternal behavior and rearing conditions, making them strong confounders during neurocognitive testing.^105,²¹¹^–²¹³ Moreover, another obvious and important factor in neurocognitive testing is the verification of similar degrees of motivation when comparing separate groups of animals. For example, a 24-hour exposure to ketamine sedation early in life impaired subsequent performance of rhesus monkeys in learning and memory tests, in addition to decreasing their motivation to perform these tasks.¹⁵⁸

Studies of prolonged opioid administration in immature animals have also found evidence for long-term impairment in learning tasks,174–176,179–181 as well as altered pain responses in adult animals after exposure to morphine, fentanyl, heroin, or methadone early in life.^†

Conversely, several other investigations have observed no neurologic abnormalities after administration of midazolam, isoflurane, sevoflurane, or ketamine, even when using complex neurologic tests in neonatal animals.* It remains to be determined whether these differential findings are attributable to the anesthetic doses, exposure times, species, or related to the type or timing of the neurologic tests performed. Interestingly, escalating exposure times of isoflurane in neonatal rats caused neuronal apoptosis beginning at 2 hours of anesthesia, but no evidence of long-term neurologic abnormalities until 4 hours of anesthesia.¹⁰⁶ In another study, a 6-hour exposure to isoflurane caused significant apoptosis immediately after exposure in neonatal mice, but resulted in no measurable long-term deficit in performance of complex neurologic tests as adult animals.¹⁰⁵ Moreover, in brain regions significantly affected by the neonatal neuroapoptosis, adult neuronal density was not diminished compared with unanesthetized litter mates.¹⁰⁵ These findings could either suggest that isoflurane may only accelerate physiologic apoptosis or that the developing brain’s plasticity and capacity for repair could compensate for a pathologic insult early in life. Conversely, a study in similarly aged rats observed a permanent elimination of neurons, as well as neurologic abnormalities in adult animals, after exposure to isoflurane, nitrous oxide, and midazolam as neonates, suggesting that either the specific combination of anesthetic drugs (isoflurane alone vs. the combination exposure) or species differences (rats vs. mice) could affect any relationships between neonatal neuroapoptosis and long-term function and neuronal density.⁹⁷ Alternatively, these conflicting results may be explained by the dissimilar testing environment, because neurocognitive tests are easily transferable among laboratories.²¹⁴ On the other hand, neonatal apoptosis may not be causally linked to adult neurocognitive performance at all, as evidenced by substantial apoptosis immediately after carbon dioxide–induced hypercarbia in neonatal rats without long-term neurologic sequelae.¹⁰⁷

Effects on Neurogenesis and Gliogenesis

The generation of new neurons, or neurogenesis, as well as new astrocytes, or gliogenesis, is most active in the immature brain in utero or soon after birth. Accordingly, anesthetic exposure during this critical period may interfere with neuronal proliferation, or even eliminate neuronal precursor cells, thereby portending permanent reductions in neuronal density and long-term neurologic impairment. To test this relationship, several studies have investigated the effects of anesthetic exposure, primarily isoflurane, on progenitor viability and rates of neurogenesis. Although isoflurane did not cause neural progenitor cell death in vitro, 3.4% isoflurane for 4 hours did decrease the rate of neuronal proliferation and increased neuronal fate selection.¹⁰⁸ These findings were recently confirmed in vitro, where 2.8% isoflurane for 6 hours resulted in no effects on neural stem cell viability and larger doses inhibited cell proliferation.¹²¹ Similarly, morphine has been found to decrease DNA synthesis in cerebellar neuronal precursors without affecting cell survival.¹⁷⁰

Isoflurane also impairs the growth of immature astrocytes and delays their maturation after 24 hours of 3% isoflurane, although this enormous exposure had no effect on cell viability.¹¹⁹ Morphine, although increasing apoptosis in neurons and microglia, did not affect astrocytes in a study involving human fetal cell cultures.¹⁷³

Alterations in Dendritic Architecture

The immature brain accumulates an overabundance of neuronal connections in infancy, and the number of dendrites and synapses dramatically decreases after the first year of life. Several studies have examined the effects of propofol, isoflurane, sevoflurane, desflurane, midazolam, and ketamine on dendritic arborization and synaptic architecture.* A common theme in these studies is that anesthetics can affect dendritic arborization and synaptic density, and that the direction of this change, whether it is an increase or a decrease in the number of dendritic spines, depends on the age at which the animals were exposed to anesthetics, and therefore the developmental state of the brain. During the first 2 weeks of life, the evidence indicates that anesthetic exposure leads to a decrease in synaptic and dendritic spine density in small rodents, while causing an increase in the number of dendrites beyond that age.^110,¹⁹⁷ However, the permanence of these changes remains controversial, with some studies only observing a transient effect after ketamine, midazolam, or isoflurane anesthesia early in life.^120,¹⁵³

Decrease in Trophic Factors

Isoflurane- or propofol-based anesthesia in the neonatal animal has been associated with a decrease in brain-derived neurotrophic factor (BDNF),^91,^109,¹⁹⁸ a protein integral to neuronal survival, growth, and differentiation. The cellular mechanism involves a reduction in tissue plasminogen activator and plasmin, which converts proBDNF to BDNF. Accordingly, isoflurane triggered proBDNF/p75^NTR (p75 neurotrophin receptor) complex–mediated apoptosis in neonatal mice.¹⁰⁹ Moreover, prolonged exposure to opioid receptor agonists early in life has also been found to alter nerve growth factors in the immature brain.^66,¹⁶⁴

Degeneration of Mitochondria

Ultrastructural morphologic abnormalities have been reported in mitochondria of pyramidal neurons in the subiculum of 7-day-old rats after 6 hours of isoflurane, nitrous oxide, and midazolam.¹²⁶ A morphometric analysis demonstrated mitochondrial enlargement, impaired structural integrity, and decreased mitochondrial density, indicative of a protracted injury to the mitochondria after anesthetic exposure. Moreover, ultrastructural examination of pyramidal neurons of anesthetized animals by electron microscopy revealed evidence for increased autophagy, a form of cell death.

Abnormal Reentry into Cell Cycle

Ketamine induces reentry of postmitotic neurons into cell cycle in immature rats.¹⁵⁵ Neuronal progenitor cells enter cell cycle during proliferation, but mature neurons lose this ability, and if they are forced to reenter cell cycle they will follow a path to apoptotic cell death.

Destabilization of The Cytoskeleton

The integrity of the cellular cytoskeleton is critical for proper neuronal morphology and function. Actin is one of the major components of the cytoskeleton of all eukaryotic cells, and participates in important cellular processes, including cell signaling, motility, and cell division. It is also essential in the formation of dendritic spines. Isoflurane can lead to depolymerization of actin in neurons and astrocytes, initiating cytoskeletal destabilization, impairment of astrocyte morphologic differentiation and maturation, as well as neuronal apoptosis.^119,¹²³

Effects on the Developing Spinal Cord

Most animal studies have focused on the effect of general anesthetics and sedatives on the developing brain. However, it is important to also consider the developmental impact of anesthetics on the spinal cord. After 6 hours of 0.75% isoflurane with 75% nitrous oxide in 7-day-old rats, neuroapoptosis increased in the lumbar region of the spinal cord.¹⁰⁰ Similar results were reported after 6 hours of isoflurane in a similar model, although 1 hour of isoflurane or spinal bupivacaine resulted in no neuroapotosis.¹²⁴ Intrathecal ketamine causes neuroapoptosis in the developing spinal cord of 3-day-old rats, but not at 7 days of age.¹⁵⁷ Preservative-free ketamine was associated with long-term alterations in spinal cord function and gait disturbances,¹⁵⁷ whereas, in a separate study, even high-dose intrathecal morphine produced no signs of spinal cord toxicity.²¹⁵

Putative Mechanisms for Neurotoxicity

The exact mechanisms that trigger the above responses to anesthetics and sedatives in the immature brain remain unresolved. Elucidating these mechanisms is critical in order to establish the relevance of these findings for pediatric anesthesia and neonatal critical care medicine, as well as to develop mitigating interventions, if necessary. The current, overarching hypothesis is that anesthetics and sedatives interfere with normal GABA_A and NMDA receptor–mediated activity, which are the putative targets for unconsciousness, amnesia, and immobility,²¹⁶ but are also essential for mammalian central nervous system development.^9,²¹⁷ Some have suggested that administering GABA_A-receptor agonists and/or NMDA-receptor antagonists may cause abnormal neuronal inhibition during a vulnerable period in brain development, triggering apoptosis in susceptible neurons, which in turn leads to neurocognitive impairment and decreased neuronal density in adults.^22,^60,^85,⁹⁷ Other lines of evidence suggest that the NMDA receptor–blocking properties of ketamine may upregulate NMDA receptors, rendering the neurons more susceptible to excitotoxic injury caused by endogenous glutamate immediately after ketamine withdrawal.^135,¹³⁹ However, several observations violate both hypotheses; neuronal cell death has been reported during exposure to anesthetics and not only after their discontinuation. Moreover, several anesthetics with minimal NMDA-receptor interaction, such as propofol and barbiturates, have demonstrated robust neurotoxic properties, whereas the neurotoxic potency of the NMDA-antagonist xenon is limited, therefore casting doubt on receptor upregulation being the sole mechanism for anesthetic neurotoxicity. In terms of abnormal neuronal inhibition being the main trigger for apoptosis in developing neurons, GABA_A-receptor stimulation indeed decreases neuronal activity in the mature brain; however, it also causes excitation in developing neurons,²¹⁸ thereby contradicting the inhibition hypothesis. In immature neurons, intracellular chloride (Cl⁻) concentration is high, thus GABA-induced opening of Cl⁻ channels allows this anion to exit the cell, leading to membrane depolarization. On the other hand, the intracellular Cl⁻ concentration is low in mature neurons. When anesthetics open Cl⁻ channels in mature neurons, ions enter the cell, thereby hyperpolarizing the membrane. This reversal of the cellular Cl⁻ gradient occurs as a result of a switch from the immature Na⁺-K⁺-2Cl⁻ cotransporter 1 (NKCC1) to the mature brain form, K⁺-Cl⁻ cotransporter 2 (KCC2).²¹⁹ Along these lines, studies in neonatal rats demonstrated excitatory properties in the brain and episodes of epileptic seizures during sevoflurane anesthesia in neonatal rats.²⁰⁴ Isoflurane has also been shown to cause an excessive release of Ca²⁺ from the endoplasmic reticulum via over-activation of inositol 1,4,5-trisphosphate receptors (InsP3Rs) in neonatal rats in vivo and in vitro.²²⁰ A similar mechanism may be linked to the production of Alzheimer-associated increases in β-amyloid protein levels after anesthesia.²²¹ Moreover, whereas xenon and hypothermia cause neuronal inhibition, they do not appear to exacerbate isoflurane-induced neuronal cell death, as expected by cumulative neuronal inhibition, but rather significantly reduce it.^94,^101,²⁰⁷

Importantly, evidence indicates that equianesthetic concentrations of the three contemporary inhalational anesthetics cause similar degrees of neuroapoptosis, suggesting that it is the anesthetic depth and not the doses of the anesthetics that determines the cytotoxic potency.⁷¹ However, other studies have failed to link the anesthetic and the apoptotic mechanisms. Specifically, although racemic ketamine and (S)-ketamine both elicit their anesthetic effects via NMDA-receptor blockade, (S)-ketamine induces up to 80% less cell death in vitro when compared with equipotent doses of racemic ketamine.¹⁵⁴ Moreover, concomitant administration of the GABA_A-receptor antagonist gabazine did not attenuate neuroapoptosis induced by the GABA agonist isoflurane, whereas the α₂-agonist dexmedetomidine did.⁷³ Decreases in anesthetic-induced neuronal activity may therefore be less important than the disruption of the neuronal balance of excitation and inhibition, as demonstrated by a series of studies that examined anesthesia-induced dendritic morphologic changes in mice.^70,¹⁵² Whereas simultaneous blockade of excitatory and inhibitory activity with tetrodotoxin did not lead to structural changes during synaptogenesis as would have been expected from a causative relationship between neuronal inhibition and structural damage, the administration of either GABA_A-agonistic or NMDA-antagonistic compounds alone altered synaptogenesis.¹⁵²

It is not entirely clear at this time whether cytotoxicity is a direct effect of the anesthetic itself, or of anesthetic byproducts, or by the metabolic acidosis and respiratory derangements that have been observed during anesthesia in small rodents.^105,^107,²²² Hypercarbia can trigger widespread neuroapoptosis, even in unanesthetized neonatal rats exposed to increased partial pressures of carbon dioxide. Whereas apoptotic cell death was quantitatively indistinguishable from neurodegeneration in isoflurane-treated litter mates, which were also hypercarbic, adult neurocognitive impairment was only observed in the isoflurane-treated animals.¹⁰⁷

Lastly, experimental models of neurodegeneration have implicated reentry of postmitotic neurons into the cell cycle, leading to cell death. Ketamine exposure has been found to induce aberrant cell cycle reentry, leading to apoptotic cell death in the developing rat brain.¹⁵⁵

Specific Anesthetic and Sedative Agents

In order to provide a succinct overview of the available laboratory data, we briefly review the effects of each class of anesthetics separately. Although the effects of some anesthetics, such as ketamine and isoflurane, have been extensively studied on the developing brain, the effects of others, such as xenon and desflurane, have not. However, the current data suggest that all anesthetics exert deleterious effects to some degree, on the developing animal brain (see E-Table 23-1). It is important to appreciate that although the MAC values for inhalational anesthetics are fairly constant across species, equipotent dosing of intravenous (IV) medications is not. That is, the dose of most IV medications to effect sedation or anesthesia in animals is approximately 6- to 10-fold greater than that in humans. The dosing is further complicated by the different routes by which drugs are administered in neonatal animals, which include the subcutaneous and intraperitoneal (IP) routes, as opposed to the oral or IV routes in humans. The possible importance of the interspecies differences in the pharmacology of IV medications on neuroapoptosis and cognitive dysfunction has not been addressed.

Ketamine

Possibly the most frequently studied anesthetic is ketamine, an antagonist of the NMDA and glutamate receptor that also interacts with other cell membrane proteins, such as muscarinic and opioid receptors, as well as voltage-gated calcium channels. Ketamine’s properties, which include potent analgesia, dissociative anesthesia, and relative hemodynamic stability, have made it a popular choice for procedural sedation in children, as well as for induction of anesthesia in children with critical congenital heart disease or pulmonary hypertension.²²³^–²²⁵ However, about 10 years ago, a seminal study that examined the effects of repeated IP injections of ketamine on the brain of neonatal rats observed widespread apoptosis.¹⁷ Seven injections of 20 mg/kg of ketamine, administered over a 9 hour period in evenly divided intervals, to 7-day-old rat pups caused a 3- to 31-fold increase in degenerating neurons, depending on brain region. This led to speculation that these changes might contribute to neuropsychiatric disorders.¹⁷ These initial findings for ketamine have been confirmed in more than fifty studies in small rodents, as well as nonhuman primates, both in vitro and in vivo (see E-Table 23-1). Several of these studies have identified relationships between neurodegeneration and dose, duration of treatment, as well as animal species and age during exposure. Single doses of up to 75 mg/kg or multiple IP injections up to 17 mg/kg/hr for 6 hours were not neurotoxic to neonatal rat brains,¹²⁸ whereas single doses between 20 and 50 mg/kg were neurotoxic to neonatal mice brains.^74,¹³⁴ Six or seven repeated injections of 20 to 25 mg/kg consistently cause apoptosis in neonatal rat brains.^17,^128,^131,^149,¹⁵⁰ Although the doses of ketamine in these small-animal models appeared to be excessive and plasma concentrations in the rodents were up to 7 times greater than those in humans,¹³¹ these increased doses are consistent with the requirements for IV anesthesia in small animals (refer to later section on interspecies comparison). Coadministration of midazolam, diazepam, propofol, or thiopental compounded the neuronal injury caused by ketamine.^74,^134,¹⁴² Studies in rats, mice, and nonhuman primates suggest that the susceptibility to ketamine-induced neurotoxicity is limited to a brief period after birth, with a maximum impact between 3 and 7 days of age in small rodents and less than 35 days of age in monkeys.^17,¹³⁹ Beyond these ages, ketamine-induced neuroapoptosis subsides. In addition to apoptosis, both small rodents and nonhuman primates that were anesthetized with ketamine have exhibited impaired learning tasks later in life.^129,^130,¹⁵⁸

In summary, ketamine is the most frequently studied anesthetic, in terms to its neurotoxic effects, and has repeatedly been shown to cause widespread apoptosis (an effect that is exacerbated by the coadministration of other anesthetics), as well as neurologic impairment in adult animals exposed early in life. Importantly, long-lasting learning impairment has been demonstrated in nonhuman primates, the closest animal model to humans.¹⁵⁸ However, these animals were anesthetized with ketamine for 24 hours to produce this effect, a duration that exceeds the usual clinical scenario. Furthermore, it is unclear whether the observed learning abnormalities could not be explained by a reduction in motivation to perform the learning tasks.¹⁵⁸

Inhalational Anesthetics

The second most commonly studied class of drugs is the inhalational anesthetics (see E-Table 23-1). The anesthetics desflurane, sevoflurane, isoflurane, enflurane, and halothane exert their anesthetic properties predominantly by their agonistic effects on the GABA_A receptor, but also by differing degrees on glycine, NMDA, acetylcholine, serotonin (5-HT₃), α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA), and kainate receptors. Whereas GABA represents the main inhibitory neurotransmitter in the adult central nervous system, it has excitatory properties in the developing brain,²¹⁸ which may have implications for neurotoxicity, as discussed previously. Most studies of inhalational anesthetics examined either isoflurane alone or a combination of midazolam, isoflurane, and nitrous oxide. This combination of GABA agonists and NMDA antagonists has been repeatedly found to cause widespread increases in brain cell degeneration in neonatal animals.^85,^87,^91,⁹² In addition to the immediate deleterious effects on brain structure, long-term abnormalities in spatial learning tasks and decreased neuronal cell density in adult rats have been observed after exposure to this anesthetic combination early in life.^85,⁹⁷ One MAC of isoflurane administered as the sole anesthetic for 4 hours to neonatal rats led to neurocognitive deficits in rats when they matured to adults,^106,¹⁰⁷ whereas up to 0.6 MAC for 6 hours in neonatal mice, while causing widespread neuronal degeneration immediately after exposure, failed to cause neurocognitive deficits or decreases in neuronal density in adulthood.^105,¹¹⁵ These inconsistent findings raise the question of whether neonatal neuronal apoptotic cell death is linked directly to behavioral and learning abnormalities in adulthood. Neuronal cell death has also been observed in neonatal rhesus monkeys after 5 hours of 0.75% to 1.5% isoflurane,¹¹⁴ although long-term neurologic studies in this species have not yet been published.

Similar to isoflurane, sevoflurane has also been shown to induce neuroapoptosis in neonatal mice,^71,^115,²⁰² although the long-term effects on learning and behavior are conflicting.^72,^115,²⁰² To date, few studies have examined desflurane in this context. Desflurane causes age- and species-dependent neuronal cell death in 7-day-old mice, but not 16-day-old rats.^70,⁷¹

Few studies have compared the neurotoxicity of contemporary anesthetics. At equianesthetic concentrations of desflurane, isoflurane, and sevoflurane in neonatal mice, the degree of neuronal degeneration in the superficial neocortex, a brain region significantly affected in this model, is similar.⁷¹ These results contrast with those from a study of lower concentrations of sevoflurane and isoflurane, in which cell death was less after the former than the latter, but without differences in neurocognitive performance in adult rats.¹¹⁵ In another comparative mouse study, desflurane caused a greater degree of neurodegeneration than isoflurane and sevoflurane, and long-term neurologic impairment only occurred after desflurane.⁷² The significance of these differential findings remains unknown, but could be related to methodologic differences.

Although largely phased out from clinical pediatric anesthesia practice, halothane and enflurane have been shown to induce brain abnormalities. These anesthetics were studied in models of intermittent or chronic occupational exposure during pregnancy, and caused subsequent delayed synaptogenesis and behavioral and learning abnormalities in rats after fetal exposure.¹³^–¹⁶ ⁷⁶

In summary, inhalational anesthetics represent one of the most frequently used classes of anesthetics in pediatric anesthesia, and their neurotoxic properties have been extensively studied. Consistent findings include widespread apoptosis immediately after exposure in a wide variety of animal models, including nonhuman primates. However, neurologic impairment in adult animals exposed early in life to anesthesia has not been a consistent finding, and in fact some studies have found no neurologic impairment at all. Therefore long-lasting learning impairment has not been convincingly linked to neonatal neuroapoptosis. Primate studies on long-term neurocognitive outcomes after inhalational anesthesia early in life have not yet been published.

Nitrous Oxide

Nitrous oxide, an NMDA antagonist, is the oldest anesthetic in clinical use, although its low potency (MAC of 115% in adult humans) necessitates the coadministration of other anesthetics to provide surgical anesthesia. Anesthetic combinations with nitrous oxide often include the GABA agonist midazolam, and the mixed GABA-agonist/NMDA-antagonist isoflurane.* In rats, nitrous oxide alone did not induce neuronal apoptosis,^85,^87,⁹⁴ whereas in an in vitro study, it did cause neuronal cell death in mouse hippocampal slices.⁹⁴ When administered in combination, however, nitrous oxide has been shown to exacerbate neuronal cell death induced by isoflurane, and also to contribute to long-term neurologic abnormalities in rats when combined with isoflurane and midazolam.⁸⁵

Xenon

Due to the cost differential between other inhalational anesthetics and this rare, colorless, and odorless noble gas, xenon has not reached widespread clinical anesthesia practice, despite its NMDA-antagonistic, anesthetic properties.²²⁶ Xenon has a relatively low anesthetic potency, with a MAC measuring between 65% and 70% in adults^227,²²⁸; its low blood-gas solubility speeds emergence from anesthesia.²²⁹ Xenon’s effects on neuronal apoptosis have been examined by two groups, with slightly differing results. Although 75% xenon for 6 hours did not cause neuronal apoptosis in 7-day-old rat pups,⁹⁴ 70% xenon for 4 hours did increase neuroapoptosis in 7-day-old mice.²⁰⁷ Interestingly, both studies demonstrated that xenon decreased the neurodegeneration induced by isoflurane anesthesia,^94,²⁰⁷ which may have relevance to the phenomenon’s putative mechanism, as discussed.

Benzodiazepines

Benzodiazepines, such as clonazepam, diazepam, and midazolam, have been investigated regarding their effects on the immature brain, either alone or in combination with other drugs. These GABA agonists are frequently used in toddlers and older children for preoperative anxiolysis, but infrequently in neonates and infants. Studies have shown that they increase neuronal degeneration in small-animal models, depending on the dose, region of the brain, species, and age of the animal. Although single doses of 5 mg/kg of diazepam or 9 mg/kg of midazolam IP did not increase neuronal cell death in neonatal rats,^68,⁸⁵ 5 mg/kg subcutaneous diazepam did cause neuronal cell death in some brain regions in mice, though this was not associated with learning deficits in adults.⁷⁴ In these studies, the neuroapoptosis associated with diazepam was significantly augmented by the coadministration of other sedatives, such as ketamine.⁷⁴ Studies have consistently reported increased neuroapoptosis in neonatal rats after diazepam in doses of 10 mg/kg or greater,^68,⁶⁹ an effect that was prevented by the coadministration of the benzodiazepine-antagonist flumazenil in one study.⁶⁸ Two studies reported no neurocognitive learning disabilities in adult mice after they were sedated with diazepam or midazolam as neonates.^74,¹⁶⁶

Chloral Hydrate

The sedative chloral hydrate, a chlorination product of ethanol that is a GABA agonist as well as NMDA antagonist, has been largely supplanted by barbiturates and benzodiazepines in pediatric clinical practice. However, it is still being used in doses of up to 120 mg/kg for sedation for radiology studies,²³⁰ and its neurotoxic properties have been investigated. Preliminary results indicate that it causes neuroapoptosis in the cerebral cortex and the caudate-putamen complex in immature mouse pups in doses of 100 mg/kg or greater.⁶⁷ The neurofunctional outcome in adult mice, however, has not as yet been investigated.

Barbiturates

Barbiturates act largely via the GABA_A receptor, but also exert effects via nicotinic acetylcholine, AMPA, and kainate receptors. Thiopental alone, in doses of 25 mg/kg subcutaneously in neonatal mice, did not induce apoptosis, although when doses of 5 mg/kg were administered in conjunction with 25 mg/kg of ketamine subcutaneously, neuronal degeneration occurred and was associated with long-term impairment of learning and memory.¹⁴² Pentobarbital and phenobarbital have been shown to induce neurodegeneration in mouse and rat pups. Furthermore, after receiving these sedatives as neonates, long-term alterations in brain protein expression and learning and memory have been observed,^103,^184,¹⁸⁶ although in one study these long-term alternations may be attributed in part to hypoxia and hypercarbia during the neonatal sedation.¹⁸⁴ Interestingly, estradiol has been shown to attenuate phenobarbital-induced neuroapoptosis.^68,¹⁸⁵

Propofol

In recent years, propofol has supplanted the barbiturates as the IV induction agent of choice in children. Propofol predominantly acts via GABA and glycine receptor–agonistic properties, but also weakly on nicotinic, AMPA, and NMDA receptors, and has been repeatedly studied regarding its neurotoxic profile, both in vitro and in vivo. The overall consensus of this body of literature is that propofol, in a dose- and exposure time–dependent fashion, has dramatic effects on the developing brain in animals. Propofol has consistently caused neuroapoptosis after doses exceeding 50 mg/kg (subcutaneous or IP) or repeated doses exceeding 20 mg/kg/hr for 4 to 5 hours in neonatal small rodents.^142,^148,^191,¹⁹⁶ Interestingly, lithium protected from propofol-induced neuroapoptosis in neonatal mice.¹⁴⁸ However, after 24 hours of IV anesthesia with propofol (6 mg/kg/hr) and fentanyl (10 µg/kg/hr), there was no evidence of apoptosis.¹⁹² Apart from overt neuronal cell death, propofol decreases the GABAergic enzyme glutamic acid decarboxylase,¹⁸⁸ decreases nerve growth factors,^194,¹⁹⁸ and causes neurite growth cone collapse in tissue culture.¹⁹⁰ In addition, propofol alters dendritic spine architecture in developing rats, depending on the age at the time of anesthetic exposure.¹⁹⁷ For example, dendritic spine density decreased if the rat was anesthetized during week 1 of life, but it increased spine development if exposure took place during week 3 of life; the mechanism of these differing responses remains elusive.¹⁹⁷

Dexmedetomidine

Unlike other anesthetics and sedatives, dexmedetomidine is a sedative and analgesic that does not interact with GABA, NMDA, or opioid receptors, but rather acts by stimulating presynaptic α₂-adrenergic receptors. Given this different mechanism of action, it is interesting that dexmedetomidine, administered in three IP injections of 1 to 25 µg/kg at 0, 2, and 4 hours to 7-day-old rat pups, did not cause neuronal apoptosis, but attenuated isoflurane-induced neuroapoptosis. Even a dose of 75 µg/kg dexmedetomidine (75-fold greater than the median effective dose needed for sedation), did not cause apoptosis in this model. Furthermore, dexmedetomidine prevented the long-term cognitive impairment from isoflurane.⁷³

Opioid Analgesics

To date, only one study investigated neurotoxic properties of opioids in relation to an inhalational anesthetic regimen. Mechanically ventilated, neonatal pigs, sedated with an IV bolus of 30 µg/kg fentanyl followed by 15 µg/kg/hr for 4 hours, exhibited significantly less neuroapoptosis in several regions of the brain, compared with an intramuscular injection of 1 mg/kg of midazolam, followed by 4 hours of 0.55% isoflurane and 75% nitrous oxide.⁷⁹ These initial findings are interesting, although future neurotoxicity studies that include opioid infusions need to also include adjuvants that produce amnesia, as is customary in pediatric anesthesia practice. Moreover, when comparing different anesthetic regimens regarding their neurotoxic potential, equianesthetic potency should be established during the study.

Importantly, however, the long-term consequences of opioid administration to the immature brain needs to be further elucidated before recommending such a regimen as an alternative strategy. Similar to GABA and NMDA receptors, opioid receptors are also intimately involved in early brain development and synaptogenesis,^231,²³² which would make it plausible that opioids could similarly affect brain development during the critical period of synaptogenesis. Increased neuronal cell death and decreased neuronal density after perinatal exposure to opioid receptor agonists, such as morphine and heroin, have been observed in developing animals.^82,^83,²³³ Chronic buprenorphine and methadone treatment early in life have been found to diminish concentrations of nerve growth factors in the immature brain.^66,¹⁶⁴ Moreover, immediate and permanent reductions in µ-opioid receptor density have been observed after perinatal exposure to morphine,^168,¹⁶⁹ and may be associated with long-term impairment of memory and cognitive function in small animals,^171,^{174–176,179–181} as well as exaggerated nociceptive responses to a pain challenge later in life.^177,¹⁸² Stimulation of the κ-opioid receptor may amplify neuronal cell death induced by proapoptotic agents.²³⁴ High-dose fentanyl has been shown to significantly exacerbate white-matter brain lesions induced by glutamatergic overstimulation in mice.⁷⁸ All these animal data suggest that prolonged perinatal exposure to opioids may have immediate and long-term effects on the developing brain and may amplify proapoptotic stimuli, although the effects of a brief or single dose exposure to opioids is not well understood. These data suggest that further studies into the interactions of opioid analgesics and anesthetics in the developing brain are warranted.

Exposure Time, Dose, and Anesthetic Combinations

Neuronal injury depends on the anesthetic dose and/or duration of anesthesia for inhalational anesthetics, and the dose, route of administration, and the number of doses for injectable anesthetics. Moreover, combinations of several anesthetics and sedatives have, in general, caused more neuroapoptosis and long-term cognitive changes than have single drugs. For example, combinations of midazolam, nitrous oxide, and isoflurane cause a much greater degree of neuroapoptosis in neonatal rats than isoflurane alone, even when the latter is administered at a greater inspired concentration.⁸⁵ In the case of ketamine, its neurodegenerative potency is amplified when coadministered with thiopental or propofol.¹⁴² Anesthetic combinations of mixed GABA agonists and NMDA antagonists demonstrate exaggerated effects. This evidence supports the notion that a deeper level of anesthesia increases the neuroapoptotic injury. However, two anesthetics, xenon and dexmedetomidine, which actually attenuate the neurotoxic effects of isoflurane anesthesia,^73,^94,²⁰⁷ do not lend support to this notion.

Deleterious Effects of Untreated Pain and Stress

Given the evidence for deleterious effects of anesthetics and analgesics on the developing brain and the paucity of complete anesthetics devoid of cytotoxicity, would children fare better by withholding potentially toxic drugs and allowing them to experience the pain and stress of surgery? Of course such a practice would be unethical, and there is a plethora of evidence of the deleterious effects caused by the stressful conditions of exposing the immature nervous system to recurrent painful stimulation. Noxious stimulation in the neonate has been associated with subsequent hyperalgesia or hypoalgesia, depending on the type and severity of injury.²³⁵ Repetitive painful skin lacerations for procedures as minor as blood draws in neonates have been found to lead to long-term, local sensory hyperinnervation.²³⁶ In addition to these local responses, repetitive, inflammatory pain early in life has been found to result in hyperalgesia and lasting changes in nociceptive circuitry of the adult dorsal horn.²³⁷ Repeated painful injections into the paws of rat pups resulted in a generalized thermal hypoalgesia.²³⁵ In addition to altered pain processing and sensory perception, repetitive or persistent pain in the neonate alters behavior and cognitive function in adulthood, decreases pain thresholds, and increases vulnerability to stress and anxiety disorders or chronic pain syndromes later in life.^141,²³⁸^–²⁴¹

In addition to painful stimulation, early adverse emotional experiences can also induce long-lasting abnormalities, such as imbalances of the inhibitory nervous system,²⁴² impairment of normal development of the nociceptive system, long-term behavioral changes,²⁴³ and persistent learning impairment.¹⁷⁹ Therefore fetuses, neonates, and infants subjected to pain and stresses associated with invasive procedures without adequate anesthesia and analgesia may be at risk for long-term adverse outcomes.

Accordingly, preemptive administration of analgesics and sedatives, such as morphine or ketamine, has been found to ameliorate the deleterious effects of neonatal pain in some of the animal studies.^141,^179,²⁴¹ However, either the presence of painful stimulation or 5 days of morphine administered to neonatal mice independently impaired adult rewarded behavior, although the combination did not.¹⁷⁹

Clinical reports have also demonstrated that human neonates and infants can mount a metabolic and endocrine response to perioperative stress and painful stimulation, which include surges in catecholamine, cortisol, β-endorphin, insulin, glucagon, and growth hormone levels.²⁴⁴^–²⁴⁶ Some of these markers, such as cortisol, remain increased for more than a year after the insult, possibly as a result of cumulative stress related to multiple painful procedures early in life.²⁴⁷ Inhalational anesthetics, opioid analgesics, as well as regional anesthesia, inhibit intraoperative stress and improve postoperative outcomes.^245,^248,²⁴⁹ Moreover, adequate perioperative anesthesia reduced the incidence of other complications, such as the incidence of sepsis and disseminated intravascular coagulation, that led to a decrease in overall mortality.²⁵⁰ Even less invasive procedures, such as circumcisions performed without analgesia in young boys, can exaggerate responses to painful challenges (e.g., immunizations) later in life.²⁵¹ Conversely, topical or regional anesthesia for circumcision blunts not only the immediate humoral stress response during the procedure,²⁵² but also pain-induced, long-term hyperalgesia.²⁵¹ In preterm neonates, painful stimulations early in life have also been associated with subsequent diminished cognition and motor function.²⁵³ In a retrospective study of children greater than 1 year of age who were born at less than 32 weeks gestational age without significant neonatal brain injury or major sensorineural impairment, the more skin-breaking procedures from birth to term (including heel sticks, intramuscular injections, chest tube placements, and central line insertions) predicted poorer subsequent cognitive and motor development (as assessed using the Bayley Scales of Infant Development II) when compared with term controls. Importantly, after controlling for severity of illness, duration of morphine administration, and exposure to postnatal dexamethasone, gestational age at birth was not significantly associated with cognitive or motor outcome. These findings suggest that repetitive pain-related stressful experiences and not prematurity per se was responsible for the poor neurodevelopmental outcome.²⁵³ Although this study did not examine the effects of anesthetic or analgesic administration during painful stimulation on subsequent outcome, a small, retrospective study suggested an improvement in outcome after administering anesthesia during painful stimulation.²⁵⁴ Painful stimulation during reduction of herniated bowel without anesthesia in infants suffering from gastroschisis tended to more frequently lead to serious adverse events, such as bowel ischemia, need for total parenteral nutrition, and unplanned reoperation, than in infants undergoing the same procedure with general anesthesia.²⁵⁴ However, despite persistently large numbers of painful and stressful procedures performed in vulnerable neonates, data indicate that the majority of these procedures are still not accompanied by adequate analgesia or anesthesia.²⁵⁵

Together, the data from animals and humans suggest that pain-related stress experienced early in life is deleterious to the developing nervous system, that pain in children remains undertreated, and that analgesics and/or anesthetics may alleviate many of the degenerative effects of unopposed pain and improve outcome. If anesthetics protect from the deleterious effects of painful stimulation, but confer their own cytotoxic adverse effects, the major question remains whether other compounds can alleviate these adverse effects.

Potential Alleviating Strategies

Whereas the translational relevance of anesthetic-induced neurotoxicity in neonatal animals to humans remains unresolved, several laboratory studies have tested strategies to alleviate some of the deleterious effects of anesthetics and sedatives in animals. These have been directed at both the immediate effects after anesthesia, such as neuronal cell death, as well as long-term abnormalities, such as neurocognitive impairment. Importantly, many studies administered adjuvants immediately before or during anesthesia, thereby maintaining an adequate level of anesthesia and analgesia to avoid the deleterious effects of unopposed pain.

The sedative dexmedetomidine and the anesthetic xenon possess very limited neurotoxic potencies themselves, but dramatically reduce isoflurane-induced neuroapoptosis.^73,^94,²⁰⁷ In addition, the coadministration of dexmedetomidine prevents long-term memory impairment after 6-hour isoflurane exposure in rats.⁷³ Interestingly, neuroprotective preconditioning has been demonstrated in pheochromocytoma PC12 cells in which pretreatment with a small concentration of isoflurane prevented the neurotoxic response to a subsequent greater dose.⁹⁵ In in vitro studies, the inositol triphosphate–receptor antagonist xestospongin C, tissue plasminogen activator, plasmin, inhibition of the neurotrophic receptor p75^NTR and the RhoA receptor, or prevention of cytoskeletal depolymerization with either jasplakinolide or TAT-Pep5 significantly attenuated isoflurane-mediated neuroapoptosis.^104,^109,¹²³ In addition, l-carnitine attenuated neuronal apoptosis after 6 hours of isoflurane and nitrous oxide in 7-day-old rat pups.¹⁰² Supplementation with the naturally occurring hormones β-estradiol or melatonin prevented the deleterious effects on neuronal survival of a prolonged exposure to midazolam, isoflurane, and nitrous oxide.^87,⁹¹ Similarly, coadministration of β-estradiol significantly reduced phenobarbital-induced neuroapoptosis.^68,¹⁸⁵ The only potentially protective agent that has been unsuccessfully tested in published studies is the GABA antagonist gabazine.⁷³ On the other hand, pilocarpine reduces neuroapoptosis induced by the GABA agonists isoflurane and midazolam, while augmenting the damage after administration of the NMDA antagonist phencyclidine, based on preliminary results from neonatal mice.⁹³ Preliminary data from the same laboratory also suggest that whole-body hypothermia with a targeted brain temperature of less than 30° C may protect from the neuroapoptotic ramifications of a 4-hour exposure to 0.75% isoflurane or to 40 mg/kg of IP ketamine in neonatal mice.^41,¹⁰¹ Another therapy that has been successfully tested in mouse pups that received 40 mg/kg of ketamine or 100 mg/kg of propofol was lithium, 3 to 6 mEq/kg subcutaneously, which abolished the anesthetic-induced neuroapoptosis in cortex and the caudate-putamen complex.¹⁴⁸

Because the applicability and extent of anesthetic neurotoxicity has not been established in humans, it is obviously premature to recommend any of these protective strategies for children. Moreover, the safety of many of these drugs and interventions has not been established in human neonates and infants. Tissue plasminogen activator and plasmin promote fibrinolysis and may not be first-line treatments during invasive surgical procedures. The sex hormone β-estradiol may not be a feasible adjuvant in prepubescent boys. The safety of pilocarpine in young children may be hampered by its proconvulsant activity observed in animal studies.^256,²⁵⁷ Lithium has been labeled harmful to the human fetus and may cause neurocognitive impairment in young children.²⁵⁸^–²⁶⁰ Whole body hypothermia below 30° C may not be a clinically feasible modality because even mild perioperative hypothermia, at least in adults, has been causally linked to numerous complications, including increased blood loss and transfusion requirements, morbid myocardial outcomes, prolonged postanesthetic recovery and hospitalization, thermal discomfort, as well as an increased risk of surgical wound infections.²⁶¹ Therefore hypothermia to treat anesthesia-induced neurotoxicity is unlikely to play a substantive role during routine pediatric anesthesia, but may have a role in infants undergoing hypothermic cardiopulmonary bypass for heart surgery to treat congenital defects. Unfortunately, the latter population often presents with neurocognitive abnormalities before anesthesia.²⁶² Although xenon’s scarcity renders it a very expensive anesthetic, dexmedetomidine’s wider availability and increased usage in pediatric anesthesia makes it a more attractive option for further research into protective strategies.^263,²⁶⁴

Anesthetic Neuroprotection

Even relatively brief periods of inadequate oxygen or blood flow to the brain may lead to neuronal injury and long-term neurologic impairment, because of the limited tolerance of the brain to ischemia. Importantly, animal studies have repeatedly confirmed the protective properties of anesthetics when administered during episodes of brain hypoxia-ischemia. However, most of these studies have been conducted in adult animals.²⁶⁵ In immature animal models, anesthetics have also been found to reduce neurologic injury and improve functional outcome after brain ischemia. Desflurane alleviates neuronal cell death and early neurologic dysfunction in a model of the neonatal pig during hypothermic cardiopulmonary bypass and deep hypothermic circulatory arrest.^266,²⁶⁷ Isoflurane treatment before hypoxia-ischemia protects the brain and improves survival in neonatal rats and mice.²⁶⁸^–²⁷⁰ Both xenon and sevoflurane protect the immature brain during simulated hypoxia-ischemia in an in vitro model.²⁷¹ Furthermore, sevoflurane may be protective in a neonatal mouse in vivo model.²⁷² These findings in immature animals suggest that critically ill human neonates could potentially benefit from these protective properties during clinical scenarios of neurologic injury, such as cardiopulmonary bypass, neurologic surgery, or perioperative cardiocirculatory arrest; these potential benefits need to be weighed against the potential neurotoxic properties of the current anesthetics.

Critical Evaluation of Animal Studies and Interspecies Comparisons

To determine whether the findings from animal studies should guide clinical practice, it is important to rationally evaluate how well animal studies represent the human clinical scenario.*

To date, no animal model has been developed that covers all aspects of human physiology and pathophysiology during surgical procedures. The task of modeling potential vulnerabilities of the developing human brain in animals is complicated by the fact that human CNS development is much more protracted than that of any of the model species. However, human neurocognitive performance is also much more complicated compared with lower animals, and therefore a potential injury might have a greater impact in such a complex system.

Duration of Exposure

In order to elicit toxic effects, the designs of many animal studies include durations of anesthesia from 1 to 24 hours. Accordingly, some of these times may extend well beyond the average time of routine pediatric anesthesia. However, expressing the duration of anesthesia as a fraction of a subject’s life span, thereby equating a 6-hour anesthetic in mice to a 2-week or greater anesthetic in humans, is probably an oversimplification. The duration of neurodevelopment may not be immediately relevant when considering the likelihood of injury at a cellular level. Nonetheless, the extent of the injury and its functional implications may potentially be related to the duration of the entire brain developmental period, even though the mechanism of anesthetic-induced neurodegeneration is unknown. Because human brain development occurs at a much slower pace than in any of the other species, similar exposure times could have different effects on potential susceptibility and the ability for postexposure repair among species. For example, the brain reaches adult size at 20 days of age in rats, 3 years of age in rhesus monkeys, 7 years of age in chimpanzees, and not until 15 years of age in humans.^2,^279,²⁸⁰ Exposure to anesthesia during a larger proportion of the period of development could be expected to have a greater impact on total development and maturation. Similarly, given the plasticity of the developing brain, it seems conceivable that brains with slower growth rates possess more time for repair.^281,²⁸² Conversely, in complex organisms, such as humans, relatively minor degrees of injury in crucial areas and at crucial times during development could have relatively profound long-term effects.

Anesthetic Doses

In all animal studies, doses needed for injectable anesthetics to cause neuronal degeneration were greater, sometimes by orders of magnitude, when compared with weight-based doses commonly used in human clinical practice. However, to produce immobility, small animals also required significantly greater doses of IV anesthetics than larger animals. Due to their smaller size, greater metabolic rate, and shorter physiologic time, drug doses in animals significantly differ from those used in humans.²⁸³ Using a process called allometric scaling, which takes these species differences into account, animal drug doses comparable to human doses have been estimated to be approximately three-, six-, or twelvefold greater, respectively, for monkeys, rats, or mice (see also Chapter 6).^283,²⁸⁴ Whereas drug doses used in anesthetic neurotoxicity studies frequently still exceed these doses calculated using allometric scaling, they do not include a significant safety margin (Fig. 23-3). Moreover, using allometrically scaled doses, plasma concentrations for ketamine, for example, were approximately 3 to 10 times greater in small rodents and monkeys than those observed during clinical human practice.^131,¹³⁹ This suggests that the neurotoxic properties observed with large doses of injectable anesthetics, such as ketamine, would only have direct applicability to humans if the anesthetic and the neurotoxic effects were based on the same molecular mechanism, which has yet to be verified. Otherwise, animal studies would expose subjects to much greater plasma concentrations of a neurotoxicant than those used during anesthesia in humans, thereby leading to an overestimation of the neurotoxic effects of IV anesthetics in laboratory studies.

FIGURE 23-3 Neurodegenerative effects of injectable anesthetics in developing animals are dose-dependent. Blue bars represent doses of the respective anesthetics not causing neurodegeneration in rats and mice; pink bars represent doses causing neurodegeneration in animals. Figure includes only studies using single injections of anesthetics and doses used in animals were scaled to doses for children by using an allometric scaling technique, based on calculations outlined in reference 284. Neurotoxic data are based on experiments described in references 17, 68, 69, 74, 85, 128, 131, 134, 139, 142, 151, 194.

Doses for inhaled anesthetics generating immobility in animals, on the other hand, are much closer to clinically used doses. Moreover, similar to human anesthesia, potency of inhaled anesthetics increases with subject age, necessitating larger doses in younger animals,^71,^222,²⁸⁵ which could suggest closer applicability of laboratory data to humans.

Experimental Versus Clinical Conditions

Another important distinction between laboratory studies in animals and clinical studies in humans is the presence of significant comorbidities, such as the presence of stress and pain associated with the underlying disease that require the administration of anesthetics and analgesics. Very few of the laboratory studies to date have included the presence of surgical stress that may increase or decrease the degree of anesthesia-induced neurotoxicity. Tail clamping or injection of caustic substances are used as a model for surgical stress. Results from these studies varied: one study reported no influence of the painful stimulus on anesthesia-induced apoptosis,²⁰⁶ whereas a second reported that painful stimulation increased anesthetic-induced neuroapoptosis.¹²⁷ In contrast, therapeutic effects have been demonstrated with concomitant administration of small doses of analgesics or sedatives, compared with no analgesia or sedative when pain was induced.^141,¹⁷⁹

When studies measured metabolic and respiratory effects of anesthetic exposure in small rodent species, several demonstrated significant differences from pediatric anesthesia practice, such as extensive hypercarbia, metabolic acidosis, and hypoglycemia observed in the small rodents.^105,^107,²²² Tracheal intubation and mechanical ventilation do not seem to completely obviate these abnormalities.²²² In stark contrast to anesthesia in children, administering clinical doses of anesthetics for as little as 2 to 4 hours can be lethal for more than 20% of small rodents,¹⁰⁵ even when intermittent painful stimuli were applied.¹⁰⁷ Moreover, rearing conditions after anesthesia have a profound impact on the brain’s repair mechanisms after injury. Environmental enrichment and exercise dramatically increase neurogenesis in rodents and therefore may facilitate plasticity and repair after anesthesia, compared with regular cage housing conditions.^286,²⁸⁷ Children face daily cognitive challenges in their “enriched” environment, different from laboratory animal housing,^79,^114,^139,¹⁵⁸ that could attenuate the postulated neurocognitive effects of anesthesia. This is highlighted in a recent study in which environmental enrichment reversed the deleterious effects of anesthesia on subsequent neurologic performance in rats, resulting in performance similar to environmentally enriched, unanesthetized animals and superior to unenriched control and anesthesia-exposed animals.²⁰⁶ This study suggests that neurobehavioral outcome depends on multiple factors and that anesthesia may be only one minor insult compared with the many other more significant events in childhood.

Comparative Brain Development

A major obstacle in translating animal data to humans centers on the difficulties of matching up brain maturational stages in model animals with the equivalent stages of the immature human brain. The ongoing discussion regarding these comparisons is somewhat reminiscent of the cliché of 1 “dog year” being equivalent to 7 “human years.” Because animal studies have suggested that anesthesia-induced neuroapoptosis may be limited to very defined, early stages of development, such as from 3 to 10 days of age in small rodents,^17,⁸⁷ it becomes imperative to identify the equivalent period during human brain development, in order to assess human applicability of the animal data and to adequately plan clinical studies.

Brain architecture and development, however, are not easily compared among mammalian species. Small rodents, such as mice and rats, have a smooth (or lissencephalic) brain surface, whereas humans and monkeys exhibit the typical fissured, gyrencephalic brain surface of gyri and sulci. Overall brain size and number of neurons are vastly different between humans and animal species. Moreover, brain development varies dramatically in terms of timing and duration. Whereas considerable brain development takes place postnatally in rodents, most critical steps in humans occur in utero.²⁷³ It has generally been accepted that mice and rats are born at relatively earlier stages of brain maturity compared with humans, but that neurodevelopment in small rodents rapidly catches up with humans, mostly during the first 2 to 3 weeks of the rodent’s life. Older data based on simple estimations of brain cell numbers and degree of myelination have been interpreted to mean that the first week of life in small rodents, when the peak of vulnerability to anesthetic neurotoxicity occurs, equates to an extended time span in humans, from the third gestational trimester all the way to the third year of life.^2,^22,²⁷⁴ However, more contemporary approaches, including computational models, have approximated the 7-day-old rat to be closer in brain maturity to human fetuses at 20 to 22 weeks gestation, and the immature rhesus monkey to approximate the immature human brain closer to term (Fig. 23-4).²⁷⁵^–²⁷⁷ ²⁸⁸ According to these models, stages of brain maturity equivalent to term human neonates are not reached until after postnatal day 14 in rats or mice (online calculator available at http://www.translatingtime.net, accessed January 23, 2012). Therefore it remains questionable whether rodent models hold relevance to routine pediatric anesthesia in term neonates and infants, or whether they more closely correspond to brain maturational states during fetal surgery in midgestation. Nonhuman primates, on the other hand, exhibit brain maturation more comparable to human brain maturation at birth and could be more applicable to neonatal anesthesia. Interestingly, a recent study in rhesus macaques corroborates the greater fetal susceptibility by demonstrating a higher degree of neuroapoptosis after ketamine anesthesia in utero compared with postnatally.¹⁶⁰ However, major differences in brain development still exist between humans and nonhuman primates. In general, brain development progresses at a much slower pace in humans and developmental stages are up to 50% longer. Even on a cellular level, remarkable differences exist between humans and animals; cell cycle duration during cortical neurogenesis is approximately 17 hours in mice, 28 hours in macaque monkeys, and 36 hours in humans.²⁸² All these differences indicate that it is not sensible to directly apply observations of anesthetic effects in the developing animal brain of any species to human clinical anesthesia practice.

FIGURE 23-4 The degree of neuronal apoptotic cell death following anesthetic exposure is highly dependent on the age of the animal during exposure. Graphs demonstrate the percent increase of apoptotic cell death compared with natural apoptosis following an exposure to an isoflurane/nitrous oxide–based anesthetic in immature rats (dark gray) or very young macaque monkeys (dark brown) or prolonged exposure to ketamine in immature rats (light gray) or macaque monkeys (tan). Solid lines connect the available data points; interrupted lines represent extrapolations of the available data. Neurotoxicity data were derived from references, 17, 91, 114, 139. To infer the potential age of anesthetic vulnerability in humans, respective brain maturity in the animal species during exposure were equated to the corresponding state of the developing human brain, and relative human ages for each data point were plotted accordingly, using the mathematical model outlined in references, 276, 277, 288 and available in the online calculator at http://www.translatingtime.net.

Assessing Neurobehavioral or Cognitive Outcomes

Translating neurodevelopmental outcomes from animals to humans is difficult. Human cognitive performance includes the vast capacity for learning, the ability for abstract thinking, the aptitude for solving complex mathematical equations, and even the capability of inventing and operating complex machinery. Cognition is a complex process that includes such diverse processes as perception, attention, working memory, long-term memory, executive function, language, and social cognition. These brain functions are all difficult to model in animals.²⁷⁸ It is therefore imperative to critically evaluate any animal models attempting to replicate human cognitive performance. Moreover, it is important to assess the validity of these models in the context of the critical period for human brain development that they are trying to represent. Current assessment of neurocognitive performance after developmental exposure largely relies on hippocampal-dependent tests administered to adult animals following anesthetic exposure early in life.^85,¹⁰⁵^–¹⁰⁷ ^,158 However, it remains unclear whether these are the same domains that may potentially be targeted in children during anesthesia. Brain regions maximally affected by anesthesia may change with the age of the animal and developmental state of the brain.^79,¹⁶⁰ Accordingly, subsequent neurobehavioral abnormalities may vary depending on the age at which anesthesia was administered.

This discussion, however, does not entirely discredit the results from small animal studies, but rather seriously limits the generalizability of their findings to clinical pediatric anesthesia practice. More closely resembling clinical pediatric anesthesia practice, several large-animal models have used tracheal intubation and mechanical ventilation.^79,^114,¹⁶⁰ However, none of these large-animal studies included surgical stimulation during the anesthesia.^79,^114,^139,¹⁵⁸

Long-Term Outcome in Children Exposed to Anesthesia and Surgery

The difficulties in translating animal data to clinical anesthesia in human infants increases the importance of identifying potential clinical evidence for or against neurologic abnormalities in children after otherwise uneventful anesthesia early in life. Unfortunately, this question is not easily answered.

There is evidence for an association between surgery with anesthesia in early childhood and subsequent altered neurodevelopmental outcome.³⁷ Some human cohort studies have demonstrated an association between major surgery in the neonatal period and poor neurodevelopmental outcome.²⁸⁹ Children born with esophageal atresia had a lower IQ and more frequently suffered from depression, emotional, and behavioral problems compared with the general population.²⁹⁰ Children with congenital diaphragmatic hernia repair also have a high rate of neurologic sequelae.²⁹¹ Extremely premature, low–birth-weight neonates who underwent laparotomy had poorer neurodevelopmental outcomes compared with matched controls.²⁹² A cohort of infants who underwent major surgery did not perform as well in school as a matched control group of healthy infants or with infants who had major nonsurgical medical conditions.²⁹³ In a randomized trial of indomethacin treatment in 426 infants less than 1000 g at birth, neurologic impairment was present in significantly more of the 110 children who had undergone surgery (53%) compared with that in the 316 children who had received medical therapy (34%).²⁹⁴ In a study of extremely preterm infants, the IQ of those who had undergone surgery was lower at 5 years of age and exhibited more sensorineural disability than those who had not undergone surgery.²⁹⁵

These studies all included infants with major confounding factors, such as congenital malformations or very preterm birth, which might increase the risk of poor neurodevelopmental outcome independent of surgery and anesthesia. Moreover, surgery itself can lead to poor outcome as a result of the perioperative neurohumoral and inflammatory response, or the hemodynamic instability associated with major surgery. Severity of illness may have also been greater in surgical patients compared with their respective control groups. Thus, whereas these studies suggest that some infants undergoing major surgery are at increased risk of poor neurodevelopmental outcome, none of the studies provide conclusive evidence that surgery or even anesthesia is the cause of the increased risk of poor neurodevelopmental outcome.

Several recent cohort studies have focused primarily on the effects of anesthesia. In an established population–based, retrospective birth cohort, Wilder and colleagues studied the association between anesthetic exposure before 4 years of age and the subsequent development of learning disabilities.²⁹⁶ Regression was used to calculate hazard ratios for anesthetic exposure as a predictor of learning disability, with adjustment for gestational age at birth, sex, and birth weight. Of 5357 children in the cohort, 593 had been exposed to general anesthesia before 4 years of age. Compared with those not exposed to anesthesia, a single exposure was not associated with an increased risk of learning disability (hazard ratio = 1.0; 95% confidence interval [CI] of 0.79 to 1.27). However, children who underwent two separate episodes of anesthesia had an increased risk of a learning disability (hazard ratio = 1.59; 95% CI of 1.06 to 2.37) and those who underwent 3 or more separate episodes of anesthesia had an even greater risk (hazard ratio = 2.60; 95% CI of 1.60 to 4.24). The association between learning disability and multiple episodes of anesthesia remained after adjusting for American Society of Anesthesiologist physical status. The risk for a learning disability also increased according to the cumulative duration of anesthesia. However, this study suffered from several deficiencies. Because the study reported anesthetics administered between 1976 and 1982, the most common anesthetic treatment was halothane and nitrous oxide, and none of the children were monitored with pulse oximetry or capnography. It is not possible to determine in how many of these children excessive hyperventilation or unrecognized desaturation had occurred. Furthermore, the maternal birth histories were not described (e.g., magnesium may cause neuroapoptosis or be neuroprotective). Three different learning disabilities were considered with equipoise in the final analysis, and these disabilities were not tested in all children, but only when a teacher or parent requested testing. These questions limit the external validity of these data.

To reduce the impact of confounding factors, using the same population-based, retrospective birth cohort, the same group conducted further studies with a matched cohort design.²⁹⁷ The researchers matched 350 children exposed to anesthesia before the age of 2 to 700 children not exposed to anesthesia. The matching was based on several known risk factors for learning disabilities: gender, mother’s education, birth weight, and gestational age at birth. Outcomes of interest were: learning disability, need for individualized education program for an emotional or behavioral disorder, and group-administered achievement tests. In the analysis, an adjustment was also performed for burden of illness. The primary finding was that children exposed to two or more occurrences of anesthesia (but not a single occurrence) were at increased risk for having a learning disability (hazard ratio = 2.12 with 95% CI of 1.26 to 3.54), and an amplified need for individualized education programs for speech and language impairment. However, there was not an increased need for a program for emotional or behavioral disorders. The authors also detected an association between multiple exposures to anesthesia and lower mathematical scores. The same criticisms apply to this study as to the earlier study from this institution.²⁹⁶

To investigate a potential association between perinatal exposure to general anesthesia during cesarean delivery and subsequent diagnosis of learning disability,²⁹⁸ the risk of a learning disability was compared in the 193 children delivered via cesarean section under general anesthesia, the 304 delivered via cesarean section under regional anesthesia, and 4823 delivered vaginally without any anesthesia. The association between mode of delivery and learning disability was adjusted for sex, birth weight, gestational age at birth, exposure to anesthesia before 4 years of age, and maternal education. The risk of disability was similar between children delivered by vagina with no anesthesia, and cesarean delivery under general anesthesia, but risk of disability was less in children delivered via cesarean under regional anesthesia than by vagina with no anesthesia (hazard ratio = 0.64, 95% CI of 0.44 to 0.92; P = 0.017). The results implied that brief exposure to general anesthesia during delivery was not associated with subsequent learning disability, but the reason why the risk was less with regional anesthesia compared with no anesthesia is unclear, although this may suggest the possibility of substantial confounding influences. To explore the possibility that the regional blockade was protective, the authors subsequently compared those born without general anesthesia by vaginal delivery, with and without regional analgesia, and found no difference in risk of learning disability.²⁹⁹

Using the same epidemiologic cohort, the same research group detected an increased prevalence of attention deficit hyperactivity disorder (ADHD) in children who had undergone repeated surgical procedures with anesthesia, compared with none or only one exposure.^299a Kalkman and associates performed a small pilot study to test the feasibility of using a cohort of children who had urologic surgery, in order to test the association between age at surgery and neurobehavioral outcome (measured with the Child Behavior Check List).³⁰⁰ In their sample of 314 children, they found no evidence for an association between timing of surgery and neurobehavioral outcome.

DiMaggio and colleagues performed a retrospective cohort analysis using the New York State Medicaid records.³⁰¹ The researchers matched 383 children who underwent hernia repair before 3 years of age to 5050 children who did not undergo inguinal hernia repair. After adjustment for sex, age, and complicating birth conditions (such as low birth weight), children who had hernia repair were more than twice as likely to have a subsequent diagnosis of a developmental or behavioral disorder (hazard ratio = 2.3, 95% CI of 1.3 to 4.1).

To reduce environmental confounding effects, the authors then proceeded to use the data from the New York State Medicaid program to construct a retrospective sibling birth cohort.³⁰² Once again, they assessed the association between exposure to anesthesia in children less than 3 years of age and the subsequent risk of diagnosis of developmental or behavioral disorders. A total of 10,450 siblings were identified; 304 of whom underwent surgery before 3 years of age, without any history of behavioral or developmental disorders before the surgery, and 10,146 children who did not undergo surgery. The association of exposure to anesthesia with subsequent developmental or behavioral disorders was assessed with both proportional hazards modeling, and pair-matched analysis. As with their previous study, they found evidence for an association between surgery and poor neurodevelopmental outcome. The incidence of developmental or behavioral disorders was 128.2 diagnoses per 1000 person-years for those that had surgery and 56.3 diagnoses per 1000 person-years for those that did not. This association persisted when adjusted for sex, history of birth-related medical complications, and clustering by sibling status; the estimated hazard ratio of developmental or behavioral disorders associated with any exposure to anesthesia was 1.6 (95% CI of 1.4 to 1.8). The risk increased from 1.1 (95% CI of 0.8 to 1.4) for one operation to 2.9 (94% CI of 2.5 to 3.1) for two operations and 4.0 (95% CI of 3.5 to 4.5) for three or more operations. The number of siblings available for a matched analysis was relatively small. There were only 138 sibling pairs. In the pair-matched analysis there was no evidence for an association between surgery and poor outcome, with a relative risk of 0.9 (95% CI of 0.6 to 1.4); however, the small numbers may limit the power of this analysis.

Even better than sibling studies, investigations involving identical twins may further reduce confounding environmental and genetic influences. Performing such a study, Bartels and associates examined the possible association between anesthesia exposure before 3 years of age and school performance in 1143 monozygotic twin pairs.³⁰³ In the identical twins who were discordant for exposure to anesthesia (one twin was exposed to anesthesia and the other was not), their school performance was identical. This would suggest that surgery with anesthesia may not be the cause of poor school performance. Interestingly, the school performance in both the discordant pairs and the pairs where both twins underwent surgery was poorer than that for concordant twin pairs where neither twin was exposed to anesthesia. This finding could imply that there may exist an unknown genetic factor that increases the risk of both the need for surgery and poor school performance.

The Western Canadian Complex Pediatric Therapies Follow-up Group recently published a prospective follow-up study of 95 infants at 2 years of age who underwent surgical correction for congenital heart disease at the Alberta Children’s Hospital from April 2003 to December 2006.³⁰⁴ A multiple logistic regression analysis of variables associated with developmental delay found that only more days of ventilatory support postoperatively and older age at surgery were associated with significant developmental delay. There was also no evidence for an association between the sedation variables and significant motor delays. However, the use of two neurocognitive assessment tools during the study period precluded an analysis of the cognitive and motor scores as continuous variables, which may have been a more sensitive way to detect subtle impairment.

Hansen and colleagues used a very large Danish birth cohort to compare academic performance in 2689 children that had inguinal hernia repair in infancy with 14,575 children who were randomly selected as an age-matched sample from the general population.³⁰⁵ Those who had hernia repair performed worse at school, but there was no evidence for any association between surgery and school performance after adjustment for likely confounding variables. In contrast to the studies of DiMaggio and Wilder, this study detected no evidence for an association between surgery and poor neurobehavioral outcome.

Lastly, in a recent study using the Western Australian Pregnancy Cohort (Raine), Ing and colleagues found an association between language and abstract reasoning deficits in 10-year-old children who were exposed to one or more anesthetics prior to age 3, compared with previously unexposed children. After adjustment for confounders, any anesthetic before age 3 increased the risk ratio of disability in receptive, expressive, or total language as measured by the Clinical Evaluation of Language Fundamentals (CELF) test, as well as abstract reasoning (Raven’s Colored Progressive Matrices) to between 1.7 and 2.1 of unexposed controls, whereas other tests of vocabulary, behavior, and motor function were unaffected by exposure.^305a