Historical Background on Anesthetic Management of Infants
The perioperative anesthetic management of infants has greatly evolved over the past 25 years. Prior to the mid-1980s, it was assumed that infants did not perceive pain due to the relative immaturity of the developing central nervous system. Furthermore, use of volatile anesthetics was limited because of their cardiac depressant effects and the inability to precisely monitor the hemodynamic and respiratory parameters in very young infants. Thus, neonatal surgery was managed with a minimal “anesthetic” approach: nitrous oxide and muscle relaxant (Liverpool Technique). However, it was demonstrated that in premature and full-term newborns, neuroanatomic nociceptive pathways were present from the periphery to the cortex. Studies indicated that even preterm babies mount a substantial stress response to surgery under anesthesia with nitrous oxide and muscle relaxant, and that prevention of this response by fentanyl or inhalational agents suggested an improved postoperative outcome [1–4]. Furthermore, it was also demonstrated that increased neonatal metabolic (epinephrine and norepinephrine) and hormonal (insulin and glucagon) responses to cardiac surgery in neonates were extreme and associated with a high mortality rate . We now know that as early as 24 weeks’ gestation, painful stimuli are associated with physiologic, hormonal, and metabolic markers of the stress response . Our better understanding of the neurobiology of pain pathways development coupled with the evidence that pain in infants during surgery is associated with increased stress responses and increased mortality led to a significant change in perioperative anesthetic management of the youngest surgical patients. It involved not only administration of volatile anesthetics, but analgesics as well, with a goal to prevent pain sensation as well as the stress responses to surgery. Several studies demonstrated that painful stimulation in unanesthetized newborns, especially in preterm infants, leads to long-term effects. These long-term effects of early painful stimulation may involve permanent changes in pain processing and impaired brain development [7–9], including altered pain sensitivity and maladaptive behavior later in life [10,11]. Today, providing anesthesia and preventing pain and stress responses in the youngest of patients is considered a standard of anesthesia practice. Withholding anesthetics during painful procedures in infants is considered unethical.
What is Known About Anesthetic Effects on the Developing Brain?
Nationwide inpatient data in the United States indicate that 1.5 million infants, defined as those under 12 months of age, undergo surgery every year. Since millions of children and infants undergo anesthesia and/or sedation for surgery and painful procedures , significant concerns have been raised regarding the potential neurotoxic effects of sedative, analgesic, and anesthetic agents on the developing brain [13–16]. Specifically, these concerns are based on recent research findings in the field of neurobiology of anesthetic and sedative drug effects on the developing brain using different animal models. Hence as the preeminent controversy in pediatric anesthesiology: Developmental neurotoxicity after anesthetic exposure in the immature/developing brain .
During normal brain development, neurons are produced in excess; the elimination of as much as 50–70 percent of neurons and progenitor cells is critical for achieving normal brain morphology, brain size, and viability of the organism . It is well known that neurons undergo programmed cell death (apoptosis) during the brain maturation period, which can be triggered by both physiological and pathological stimuli [19–24]. Disruption of physiological apoptotic cell death during development leads to brain malformations and premature death in rodent models . Another essential process in the developing central nervous system is the formation of synaptic connections between neurons. Therefore, the critical period of neurodevelopment is also marked by a period of synaptogenesis. The pruning of neurons and synapses is activity-dependent. Gaba-aminobutyric acid (GABA), N-methyl-d-asparate (NMDA), and opioid receptors all have a direct role in neuronal migration, differentiation, and central nervous system maturation. Although the exact mechanism of general anesthesia is not entirely understood, alterations of synaptic transmission involving GABA type A receptors and/or NMDA receptors seem to play an important role.
It is plausible that drugs that either alter neuronal activity or affect the neuronal number will have some impact on final brain morphology and function. In many animal studies, apoptosis was used as a surrogate marker implicating neurotoxicity of anesthetics and sedating agents.
The landmark research report revealed potential neurotoxic effects of anesthetics on the developing rat brain . Blockade of NMDA receptors with the antagonist MK801 (an agent similar to ketamine) resulted in widespread neuronal apoptosis in the brains of week-old rats (corresponding to the peak synaptogenesis period and the most vulnerable period in this species). Subsequently, a combination of midazolam, nitrous oxide, and isoflurane for six hours in the same animal model of seven-day-old rat pups caused widespread apoptosis, changes in hippocampal synaptic function, and was associated with long-term persistent memory and learning impairments . These reports fueled laboratory investigations of other anesthetic and sedative drugs used in clinical practice. In fact, all commonly used anesthetics, such as benzodiazepines, ketamine, propofol, nitrous oxide, and isoflurane, have all been shown to exacerbate neuronal cell death. For a list of studies, please refer to Loepke and Soriano’s recent review report . Of those, a number of recent studies were conducted in nonhuman primates. These reports demonstrated enhanced apoptosis by isoflurane , but also confirmed differences in effects depending on the age of exposure  and duration of exposure to injectable anesthetic/analgesic (e.g., ketamine) .
Common themes are: (1) prolonged and repeated exposure to anesthetics/sedating agents, and (2) the use of combinations of different types of anesthetics (drugs with different mechanism of action), may contribute to an increased apoptotic effect in different animal models. Despite the abundance of valid research reports, the pattern of detrimental neurobehavioral effects is inconsistent and has not been shown in all experiments. This variability of the long-term effects was attributed to sensitivity of different brain regions to different anesthetic/sedating agents [27,32–34].
Difficulties in Extrapolating Data from Animal Models to Human
The empirical evidence supporting the neurotoxicity of anesthetic and sedative drugs in experimental animal models serves as the basis for our concern with potential harm imposed by anesthetic and sedative drugs on humans. Based on shared neurobiology, potential developmental neurotoxicity of anesthetics and analgesics depend on several factors: (1) dose of the drug; (2) developmental age of the patient; (3) duration of exposure; and (4) presence or absence of pain . However, there are several factors to be taken into account that limit simple translation of data obtained in animal models to neurobiology of human development and to current clinical practice.
Species differences are well documented with respect to appropriate dose of certain drugs. Specifically, anesthetic requirement for injectable anesthetics are much higher in rodents than in humans, by a factor of 10 for ketamine and a factor of 100 for both propofol and morphine. The dose of inhalational agents, such as isoflurane, is about 2–3 times higher in rats than humans. Furthermore, even with respect to the animal studies, results are conflicting depending on the dose of drug administered (e.g., ketamine). Neurocognitive dysfunctions were reported after injections of high or repeated doses [26,35–38], but not in those cases with a smaller dose resulting in lower plasma concentrations [30,35,36].
Critical Period and Duration of Drug Exposure
The window of vulnerability to drugs coincides with the developmental period of synaptogenesis, also known as the brain growth spurt period. The association of rat and human developmental stages depends upon several endpoints such as the number of brain cells, degree of myelination, brain growth rate, synaptogenesis, as well as measures related to more contemporary neuroinformatics [39–41]. In rodents, this critical period of neuronal differentiation and synaptic development is limited to a time window up to the fourth postnatal week [42–44]. The most vulnerable period for anesthesia-induced neurodegeneration appears to be very brief in animals, occurring during the first postnatal week in small rodents. In humans, the brain growth spurt, characterized by synaptogenesis and accompanied by dendritic and axonal growth, as well as myelination of the subcortical white matter, extends from the last trimester of pregnancy up to the first few years of postnatal life . In fact, the newborn rat model at one week of life has been extensively used in relation to early (premature, neonatal, and infant) development in humans [46–48]. Thus, it is difficult to simply extrapolate comparative developmental trajectory of rats (days to weeks) to humans (years) [49,50]. A five-hour anesthesia exposure in a rat covers a significant proportion of the animal’s period of development and may not have the same effect as five-hour exposure in a human. As of now, the critical period of human brain development is still undetermined, although some experts argue it lasts until 2–5 years of life.
Translating the Neurobehavioral Effects
Neurobehavioral tests in rodents and even in monkeys tend to be very crude (e.g., fear conditioning, spatial reference memory, water/maze-based memory consolidation tests, short-term memory, early long-term memory, etc.). It is possible that anesthesia exposure may have subtle effects in humans that cannot be detected in animal models.
As described earlier, untreated pain and stress can have detrimental effects on the developing brain. Untreated pain in animal models was also shown to be associated with enhanced apoptosis . Unfortunately, the majority of animal studies were conducted without surgery or painful stimulation. Only one limited study reported no effects of painful stimulus (e.g., tail clamping) on degree of apoptosis . In contrast, other studies found a decrease in apoptosis after ketamine administration in a rodent model of chronic inflammatory pain [51,53]. In animal models, nociceptive stimulation causes less apoptosis than general anesthesia.
In summary, translation of data from animal studies to the neurotoxic effects of anesthetics used in pediatric practice has yet to be established and can only be made if the mechanisms of both anesthesia and neurotoxicity are the same (for a review, see ).
As described in recent review articles addressing the topic of anesthesia and neurotoxicity of the developing brain [28,55], no studies report structural brain abnormalities in children after anesthesia. It is impossible to histologically examine the brains of infants who were exposed to anesthetic drugs (the way animal brains were analyzed). Furthermore, there are no imaging techniques that are sensitive enough to detect neuroapoptosis. Certainly, even if the imaging technology were available, data interpretation would be complicated by the fact that the infants would have to undergo sedation/anesthesia to obtain the images . Neurodevelopmental outcome studies are possible in children, but must be carefully designed to minimize bias and confounding factors. Specifically, neurodevelopmental outcome could be defined as the presence of neurodevelopmental disorders, such as autism, mental retardation, language delay, learning disability, or attention deficit hyperactivity disorder [56–58]. Several studies report association between surgery and an increased risk of poor neurobehavioral outcome [59–61]. According to one recent report, even infants with a relatively minor procedure, such as pyloric stenosis repair, have an increased risk of poor developmental outcome . However, due to many confounding factors (e.g., age, severity of disease, surgery), the precise role of anesthesia, analgesia, and/or sedation is very difficult to determine . There is only a paucity of studies that evaluated the neurocognitive outcomes of infants who have received prolonged or repetitive exposure to general anesthesia or sedation. In the NOPAIN trial, Anand et al.  noted that there was a poorer neurologic outcome as well as increased mortality in premature infants sedated for prolonged periods of time with midazolam compared to placebo or morphine.
Since 2007, several studies have attempted to examine the association between surgery and anesthesia in the clinical setting. Each of these studies relies on retrospective examinations of existing databases. The first of these studies published by Wilder and colleagues from the Mayo Clinic examined an existing cohort of children born between 1976 and 1982 whose complete medical and school records were retrospectively reviewed for evidence of any type of learning disability . These studies were conducted to determine whether those who were exposed to anesthesia before four years of age were more likely to have any type of learning disability than those not exposed. The authors found that those with a single exposure were not at increased risk, although those with two or more exposures were at significantly increased risk, and that risk increased with duration of exposure. Subsequently, the same group re-examined the cohort with a goal to reduce the potential for confounding factors (comorbid conditions), providing reassurance that the findings were not related to an excess of disease in the exposed group . The findings of this study were very similar to those of the previous study; the children exposed two or more times before their third birthday were at nearly twice the risk for a learning disability as those not exposed. A third study using this same cohort examined the risk of attention deficit hyperactivity disorder, and found similar results while controlling for comorbidity . Another study in older children also showed long-term differences in language and cognitive function after exposure to anesthesia . Another group by Di Maggio et al. examined a birth cohort from 1999 through 2002 enrolled in the New York Medicaid Program [68,69]. After correcting for age, sex, and complicating birth-related conditions, children who had hernia repair were more than twice as likely to be diagnosed with developmental or behavioral disorder. A study from Denmark examined the risk of poor neurodevelopmental outcome and hernia repair . After correcting for sex, birth weight, parental and maternal age, and education, no difference was reported in academic performance between those who underwent a hernia repair and those who did not. Another study by Iowa researchers reported that the duration of anesthesia and surgery correlated negatively with school performance scores between children undergoing surgery in infancy . The impact of anesthesia and surgery on educational assessments have been examined in large databases from national registries in Canada and Sweden. In contrast to the previous reports on smaller populations, these “large data” assessments reveal that children less than three years old were equivalent to the non-exposed cohort, while older children had statistically significant decrements in their educational assessments [72,73].
Interpretation of these cohort studies has several limitations . Although retrospective chart reviews may give some information about gross neurological and learning deficits, in most cases the information is incomplete and, thus, it is difficult to draw accurate conclusions. The biggest concern is the presence of confounding factors. Adjustment for these factors can only partly remove this effect. This is because adjustment for the many, likely unknown, confounding factors is impossible. An important next step is to conduct well-designed prospective clinical studies. Prospective measurement of outcome following a research protocol ensures that data collection will be consistent for all children.
Two high-profile multicenter clinical trials, the GAS (General Anesthesia compared to Spinal anesthesia) and PANDA (Pediatric Anesthesia and NeuroDevelopment Assessment) trials have recently published their findings. The GAS trial is a randomized controlled trial that provided strong evidence that an hour of exposure in infancy does not result in neurologic deficit as measurable at two years of age . However, an assessment at two years of age does not rule out an effect on higher executive function, cognition, and memory. The five-year follow-up evaluation is underway and will provide more data on cognitive outcomes. The PANDA study prospectively examined the impact of inguinal hernia surgery in infants under 36 months of age on an extensive battery of neurocognitive tests . When compared to a sibling cohort naïve to surgery and general anesthesia, no significant differences in these neurocognitive domains were detected. Both of these negative studies only examined the impact of short exposures to general anesthesia and surgery. This does not rule out an effect with longer exposures. These findings are consistent with the lack of toxicity and neurobehavioral deficits after short exposures to anesthetics in laboratory animals.
Clinical Practice in the Light of Basic Science and Epidemiological Studies
Caution should be applied in simply translating and extrapolating available preclinical data to current anesthesia perioperative management. Despite the evidence for widespread neuronal cell death in newborn animals, a clinical marker of anesthesia-induced neurotoxicity has yet to be identified in children. Being that we still lack the overt clinical evidence for impairment in neurodevelopment of children following anesthetic and sedating drug administration, there is no reason to easily dismiss the animal data and reports from limited clinical studies. On the other side, pediatric anesthesiologists do not have any alternative to current anesthesia practice for care of premature and term neonates and infants.
This brings us to a difficult question: Should we change our clinical practice ?
These concerns were addressed at the March 29, 2007 public hearing of the Anesthesia and Life Support Drugs Advisory Committee of the US Food and Drug Administration (transcript is available at ). After reviewing the preclinical data on anesthetic-induced neurotoxicity, the US Food and Drug Administration Advisory Committee issued the following statement: “[although] there are no adequate data to extrapolate the animal finding to humans” the well-understood risks of anesthesia (respiratory and hemodynamic morbidity) continue to be the overwhelming consideration in designing an anesthetic, and the understood risks of delaying surgery are the primary reasons to determine the timing. Despite these parallel risks associated with general anesthesia, the FDA published a cautionary communication on the use of anesthetic and sedative drugs in patients aged three years and under . Certainly, young children usually do not undergo surgery unless the procedure is vital to their health. In some instances, however, postponing a necessary procedure may itself lead to significant health problems and may not be an option for the majority of children [79,80]. Therefore, pediatric anesthesiologists should use the currently available knowledge to guide their practice.
How to Address Concerned Family Members
All families vary in regard to what information is desired when it comes to anesthetic care of their children. Parental desire for anesthetic information has been studied and virtually all studies report that the vast majority of parents want to know about the risks of anesthesia, including severe and rare risks such as death . Furthermore, it was reported that both parents  and children  most often want detailed information on well-defined, readily apparent, and largely short-term concerns such as: pain, nausea, anesthetic induction, and emergence. Since these studies addressed risks that are accepted and short-term, rather than those that are hypothetical and long-term, it is difficult to extend these data to the issue of anesthetic neurotoxicity.
It should be clinicians’ practice to take an individualized approach by discussing anesthetic risks in general terms and inviting parents and older children to ask for additional, more specific information. No mention of anesthetic-related neurologic injury is typically made unless parents or children specifically ask . When asked, emphasis should be placed on the following: (1) unknown risk of neurologic injury in the context of all risk associated with anesthesia and surgery; (2) the low rate of harm associated with anesthesia; and (3) the lack of compelling data clearly implicating anesthetic drugs in subsequent cognitive deficits of children.
Summary and Conclusions
It is important to understand that physiologic pruning of redundant neurons is an integral part of normal brain development. A plethora of research reports published over the last two decades demonstrating enhanced apoptotic cell death and long-term developmental dysfunctions in animal models implicate potential detrimental effect of anesthetics on brain development. However, the clinical phenotype of anesthesia-induced neurocognitive impairment still remains elusive. Considering that infants and children mostly undergo life-saving procedures, postponing the surgery is not usually a reasonable option. Certainly, in all such instances, providing anesthesia and preventing pain and stress responses in the youngest of patients is considered a standard of anesthesia practice.