What do recent human studies tell us about the association between anaesthesia in young children and neurodevelopmental outcomes?

Abstract

Anaesthetic and sedative drugs transiently disrupt normal neural activity to facilitate healthcare procedures in children, but they can also cause long-term brain injury in experimental animal models. The US Food and Drug Administration (FDA) has recently advised that repeated or lengthy exposures to anaesthetic and sedative drugs prior to 3 yr of age have the potential to harm the development of children’s brains and added warnings to these drug labels. Paediatric anaesthesia toxicity could represent a significant public health issue, and concern about this potential injury in children has become an important issue for families, paediatric clinicians and healthcare regulators. Since late 2015, important new data from five major clinical studies have been published. This narrative review aims to provide a brief overview of the preclinical and clinical literature, including a comprehensive review of these recent additions to the human literature. We integrate these new data with prior studies to provide further insights into how these clinical findings can be applied to children.

Anaesthetic and sedative drugs transiently disrupt normal neural activity to facilitate medical procedures in children, but they can also cause long-term brain injury in most experimental animal models of paediatric anaesthesia exposure.^1,2 Based on a review of available evidence, the US Food and Drug Administration (FDA) recently issued (December, 2016) a safety announcement advising that repeated or lengthy exposures to anaesthetic and sedative drugs prior to 3 yr of age have the potential to harm the development of children’s brains and added warnings to these drug labels.³ Children undergo almost 3 million anaesthetics in the USA alone each year,^4,5 such that anaesthesia-related neurological injury could represent a significant public health issue. Accordingly, concern about neurotoxicity has become an important issue for families, paediatric clinicians and healthcare regulators.

Several publications have previously reviewed the aggregate animal and human data, and described the complex issues involved in interpreting this literature.^1,2,6,7 Since late 2015, important new data from five major clinical studies have been published.^8–12 The purpose of this narrative review is to provide a focused overview of the preclinical and clinical literature, with a comprehensive review of the recent additions to the human literature, and to integrate these new data with prior studies to provide further insights into how these clinical findings can be applied to children.

Preclinical evidence

Most experimental models have found evidence of neurotoxicity following anaesthesia exposure in infant animals,² but a predominant mechanistic pathway has not been identified.¹ In addition to apoptosis of neurones and glia,¹³ other mechanisms implicated in the pathogenesis of paediatric anaesthesia neurotoxicity include alteration of signalling in neuroinflammatory pathways, oxygen free radical production, altered mitochondrial integrity causing acute neuronal injury^14–16 and altered neurogenesis, neurite growth, and synapse formation contributing to remodelling of neuronal circuitry and developmental dysregulation.¹⁷ Cell age is hypothesized to be a central factor for anaesthesia neurotoxicity, and those parts of the brain undergoing neurogenesis may be particularly vulnerable to the deleterious effects of anaesthesia.¹⁸ Because of regional heterogeneity of continued neurodevelopment throughout childhood,¹⁹ there is also the potential for regional heterogeneity in vulnerability to neurotoxic effects that could change with age.²⁰ This raises the possibility that the phenotype of anaesthesia-related neurotoxicity may depend on the age of exposure, and that neurotoxic effects could occur outside of periods of peak brain development in early childhood.²¹ Not unexpectedly, a range of neurological deficits after anaesthesia exposure have been described using experimental models, including cognitive deficits and delayed learning, impaired memory formation and retention, and altered motor and behavioural development.

Experimental animal models have the important advantage of studying the effects of anaesthesia in the absence of surgery. However, translation of these models to humans can be difficult due to differences in brain development trajectories, developmental age at exposure, neuronal structure, and equipotency of anaesthetic drugs administered to animals of different lifespans and species. Non-human primate models of anaesthesia neurotoxicity mitigate many of these issues, are regarded as being most translatable to humans, and have provided histological and functional evidence supporting the plausibility and potential significance of anaesthesia neurotoxicity in humans.^22–24 Two recently published non-human primate studies that examined the effects of exposure to volatile anaesthetics on behavioural development are especially notable.^25,26 Raper and colleagues²⁵ found that infant rhesus macaques exposed repeatedly to sevoflurane (three exposures of 4 h each) had increased anxiety-related behaviours at 6 months of age compared with unexposed controls. Consistent with these results, Coleman and colleagues²⁶ found that infant macaques exposed repeatedly to isoflurane (three exposures of 5 h each) had motor reflex deficits at one month of age compared with unexposed controls, and exhibited increased anxiety in response to novel social environments at 12 months of age. There was evidence of changes in some assessed parameters for a separate group of macaques receiving a single 5 h exposure to isoflurane, but these did not reach statistical significance.²⁶ Both of these studies support the concept that repeated exposure to general anaesthesia can have long-term behavioural consequences in primates, although both employed durations of anaesthesia exceeding those typically seen in most children.

Clinical evidence

Prior literature

Several observational clinical studies published prior to 2016 investigated the association between childhood exposure to general anaesthesia for surgical procedures and neurodevelopmental outcomes. These results have been summarized in several reviews.^27–30 In general, those studies that have investigated associations between select neurodevelopment or academic outcomes and multiple exposures to procedures requiring general anaesthesia (in children aged less than 2–4 yr) find significant associations.^31–34 Studies that examined single exposures or did not distinguish between single and multiple exposures are less consistent; some found impairments in a range of domains^35–40 whereas others did not find evidence of adverse outcomes.^41–45 Although these findings are often characterized as ‘conflicting,’ these retrospective observational studies use a multiplicity of study designs and outcomes, which are usually repurposed (i.e. primary data collection was not performed for the purpose of examining anaesthetic effects) and dictated by the types of available data sources. Thus, it is difficult (and perhaps unwise) to attempt evidence synthesis using this heterogeneous group of both ‘positive’ and ‘negative’ studies. For example, and also as noted by others,⁴⁶ some outcomes (e.g. academic achievement) lack sensitivity to detect or accurately describe phenotypes of anaesthesia neurotoxicity. Of these earlier studies, only the repurposed Western Australia Pregnancy (Raine) Cohort used direct neurodevelopmental assessments, finding that children who underwent general anaesthesia before 3 yr of age were more likely to have select deficits in language and cognition (abstract reasoning) compared with unexposed children.^37,46

Given the existing heterogeneity in findings, the following five new major clinical studies published since late 2015 are important additions to the literature.^8–12

GAS study

The interim results of the GAS (General Anaesthesia compared to Spinal anaesthesia) study were published in late 2015.⁸ Although primary outcome data [intelligence quotient (IQ) at 5 yr] will not be reported until 2018, this is a landmark study in the investigation of anaesthesia neurotoxicity as it represents the first, and thus far only, randomized clinical trial in the field. This multicentre equivalence randomized controlled trial conducted across 26 countries compared the effect of awake-regional vs sevoflurane anaesthesia on neurodevelopmental outcomes for 722 infants who were less than 60 weeks post-conceptual age at the time of inguinal hernia repair. Pre-specified interim neurodevelopmental outcomes were assessed at 2 yr of age using the Bayley Scales of Infant and Toddler Development III,⁴⁷ which has good psychometric properties and is frequently considered the gold standard for neurodevelopment assessment.⁴⁸ For cognitive composite score, there was no difference between groups (98.6 ± 14.2 vs 98.2 ± 14.7 in awake-regional and general anaesthesia groups, respectively) using a per-protocol analysis. While there were some instances of cross-over between groups and loss to follow up, this finding was quite robust in several sensitivity analyses, and the overall conduct and reporting of the trial were exemplary. As acknowledged by the authors, more subtle deficits may not be reliably assessed due to instability of developmental trajectories in young children and the potential for intra-individual variability when testing.⁴⁹ In addition, the Bayley-III conducted at younger ages can have weak predictive values for future disability in some populations.^50–52 These limitations notwithstanding, these interim results provide strong evidence that children exposed to relatively brief (median 54 min) sevoflurane anaesthetics in infancy do not experience detectable adverse neurodevelopmental outcomes at 2 yr of age compared with children undergoing the same procedure under awake-regional anaesthesia.

PANDA study

The PANDA (Pediatric Anesthesia and NeuroDevelopment Assessment) study also employed prospective neurodevelopmental testing but in an ambidirectional cohort design.¹¹ This study assessed 105 otherwise relatively healthy children (American Society of Anesthesiologists Physical Status categories 1 or 2, >36 weeks gestational age at birth) who underwent inguinal hernia repair (median duration of 80 min) before 36 months of age and healthy siblings who had no anaesthesia or surgery prior to 36 months.¹¹ The children underwent a single detailed neuropsychological assessment between 8 and 15 yr of age, with a primary outcome of global cognitive function (IQ). Mean IQ scores between exposed and unexposed siblings were not significantly different [mean difference, −0.2 (95% CI, −2.6 to 2.9)]. There were no significant differences in scores for measures of attention, behaviour, executive function, language, memory/learning, motor/processing speed or visuospatial function. The only difference between groups was found in the parent-completed Child Behavior Checklist, with exposed children exhibiting a significantly higher frequency of abnormal internalizing scores. Limitations of this study included the possibility of a recruitment bias [e.g. mean IQ values above population norms (mean full scale scores of 111 in both groups)], a high proportion of males in the exposed cohort (90%) and that 22% of the unexposed siblings were exposed to anaesthesia after 3 yr of age. Nonetheless, this is the first study to prospectively obtain detailed neurodevelopmental testing specifically to explore the association between anaesthesia exposure and outcomes, and has significant strengths in both design and execution. The potential significance of the increased proportion of children exhibiting abnormal internalizing scores was not discussed in the research report and remains unclear as an isolated finding. In particular, it is not clear whether this is consistent with behavioural changes seen in non-human primates²⁶ or the tendency (not statistically significant) found in another observational study towards an association between single anaesthesia exposures and increased frequency of attention-deficit hyperactivity disorder.³⁴

Canadian population-based studies

Two Canadian population-based studies conducted in the provinces of Ontario and Manitoba^9,10 both used similar methodological approaches and the same assessment tool [the Early Development Instrument (EDI)]⁵³ to measure outcomes. The EDI is a teacher-completed 103-item multidimensional questionnaire that assesses children’s readiness to learn at school entry (5–6 yr) in five major domains: emotional health and maturity, general knowledge and communication skills, language and cognitive development, physical health and well-being, and social knowledge and competence. Although it is not a direct assessment of neurodevelopment and thus may lack sensitivity to detect specific neurocognitive deficits, the EDI has moderate concurrent validity with similar domains in direct measures of child development.⁵⁴ It has also been shown to discriminate the effects of other early life insults on child development.^55,56

In the study by O’Leary and colleagues,¹⁰ from a population of 188 557 children in Ontario, 28 366 were identified to have undergone surgical procedures requiring general anaesthesia before primary school age (aged 5–6 yr), with 20% undergoing more than one procedure. Exposed children were 2:1 matched [on birth year, sex, home location (urban vs rural), mother’s age and gestational age] to a control group of unexposed children. Covariates adjusted for in the analysis included aboriginal status and income quintile. Compared with control children, exposed children were at increased risk of early developmental vulnerability, which was defined as any EDI domain in the lowest 10th percentile of a population.⁵⁴ However, the magnitude of the effect size was small [odds ratio, 1.05 (95% CI, 1.01 to 1.08) in fully adjusted analysis]. Additional analyses were also performed comparing age at first exposure (<2 yr and ≥2 yr), number of surgical procedures and children with single surgical procedures with brief associated hospital stays, all finding odds ratios between 1.04 and 1.06. Because of the small effect size and relatively small numbers of children in subgroup categories, some of these results did not reach statistical significance. When individual EDI domains were analysed (according to the proportion in the lowest 10th percentile), anaesthesia exposure was significantly associated with impairment in the domains of ‘emotional health and maturity’ and ‘physical health and well-being’, but there was no effect on ‘language and cognitive development’ and indeed an improvement in ‘communication skills and general knowledge’.

In the study of Graham and colleagues,⁹ from a population of 52 175 children in Manitoba, 4470 were exposed to surgical procedures requiring general anaesthesia before 4 yr of age, with 14% undergoing more than one procedure. Exposed children were 3:1 matched using similar criteria to those employed by O’Leary and colleagues.¹⁰ Covariates included in the analysis included involvement with child and family services, gestational age and Johns Hopkins Resource Utilization Band as a measure of burden of illness.⁵⁷ In contrast to analysis by O’Leary and colleagues,¹⁰ the EDI was analysed here as a continuous variable. Compared with control children, both single and multiple exposures to anaesthesia were associated with worsened EDI score [mean difference (95% CI), −0.87 (−1.1 to −0.6) and −1.2 (−1.8 to −0.6) for single and multiple exposures, respectively]. These represent small effect sizes (approximately 0.1 standard deviation), and effects did not differ significantly between single and multiple exposures. When analysed according to age of exposure (age ranges of 0–2, 2–3 and 3–4 yr), significant differences were only observed in the older age groups. In contrast to O’Leary and colleagues,¹⁰ when individual EDI domains were analysed the largest effect size differences were observed in the domains of ‘communication and general knowledge’ and ‘language and cognitive development’.

These analyses have many strengths, including rich data resources to ascertain both outcome and potential confounding factors affecting this outcome (e.g. home environment, intercurrent disease, perinatal health and socioeconomic status). Their nearly simultaneous publication also provides interesting insights into the complexities of conducting and interpreting such studies. Both found no evidence for either greater effects with exposure at younger ages or a dose–response relationship between exposure and outcomes; this pattern is not consistent with findings from preclinical and other ‘positive’ observational studies and could argue for confounding by indication as an explanation (i.e. children who require surgery have underlying unmeasured factors responsible for the observed differences) rather than effects caused by anaesthesia. Of particular interest, the specific EDI domains affected differed between the two studies. Although both studies used similar approaches to study design and methodology, their method of analysing the primary outcome differed (dichotomous threshold vs continuous variable) and this might have contributed to the differences observed between them. It is also worth noting that common surgical procedures appear to differ widely among the provinces, and differences in underlying diseases (behavioural, developmental or physical) could have contributed to their findings. In the Ontario cohort, the most common procedures were myringotomy and tube placement (22%) or tonsillectomy surgery (21%); in the Manitoba cohort, dental procedures were most common (38%), especially in older children where dental procedures accounted for 57% of all procedures in children aged 2–4 yr. O’Leary and colleagues¹⁰ did undertake a sensitivity analysis examining the effect of myringotomy and tube placement on their outcomes and found no differences in effect size, but the increased tendency for healthcare utilization (and general anaesthesia for some healthcare procedures) by children with developmental and behavioural problems (who may require anaesthesia for dental procedures) could also have contributed to the observed differences in outcomes. Thus, even large studies that are conducted in a seemingly similar fashion have underlying differences that could affect results.

Swedish population-based study

Glatz and colleagues¹² identified 33 514 children who underwent a single anaesthesia exposure before 4 yr of age and no subsequent hospitalization from a cohort of over 2 million Swedish children born between 1973 and 1993. These children were 5:1 matched with unexposed children matched on sex, month and year of birth, maternal parity and county of residence. A separate cohort of children undergoing multiple procedures prior to 4 yr of age was also identified and analysed. Primary outcome measures were school grades at 16 yr of age, based on a uniform school curriculum calibrated by a national individual-level test in mathematics, Swedish and English, and IQ measured at 18 yr of age in a subset of children (males born before 1986). School grades and IQ were modestly lower in exposed children [mean difference, 0.41% (95% CI, 0.12–0.70) and 0.97% (0.15–1.78), respectively]. Effects of exposure on school grades did not differ according to age of exposure but there was a tendency towards greater effects at older ages (IQ was not analysed). Effects of exposure on school grades were approximately 3-fold greater in children with multiple exposures, although this difference did not reach statistical significance. Consistent with prior population-based studies using school grades as an outcome measure,^43,44 exposed children in this cohort were more likely to have no recorded school grades or IQ scores.

Integration of new data with prior literature

Several themes emerge from these five new additions to the literature. First, despite the significant variations in study design and outcome measures, all of these studies find either no (the smaller-sized GAS and PANDA studies) or very modest (the three large population-based studies) associations between exposure to surgical procedures requiring general anaesthesia and the outcomes examined. In the case of single exposures, these results are consistent with others from several prior investigations, including other population-based studies.^{31,34,41–45} Perhaps the greatest contrast in findings is for the PANDA and Raine studies,^11,37,46 as the latter found significant associations between even single exposures and neuropsychological assessments of language and cognition. There are several differences between these studies that could have contributed to observed differences, including a more heterogeneous group of both children and procedures (e.g. without sibling matched controls to mitigate for effects of home environment) in the Raine study. Nonetheless, these studies strengthen the preponderance of evidence suggesting that single, relatively brief exposures to surgical procedures requiring general anaesthesia are not associated with detectable deficits in most cognitive domains or academic achievement. There are some suggestions of potential effects of single exposures to anaesthesia on behaviour, which are consistent with results from some non-human primate studies, but these are of such modest effect sizes that they might not be statistically or clinically significant.

Second, all previous human studies that have distinguished between single and multiple exposures to procedures requiring general anaesthesia show an association between multiple exposures and deficits in the outcomes studied,^31–34 and the new large population-based studies add to this evidence albeit with very modest effect sizes. The significance of multiple exposures is consistent in the preclinical literature, including the recent primate studies.^25,26 However, the evidence for a dose–response relationship between number of exposures and outcomes, found in most^31,32,34 but not all³⁷ prior studies, is apparent only in the Swedish¹² but not the Canadian^9,10 studies. The heterogeneity of outcomes examined among these studies may have contributed to the heterogeneity of results. For example, those studies showing a dose-response relationship also tend to show larger effect sizes (hazard ratios of approximately 2^31–34) in comparison to the small effect sizes in most studies not showing such a relationship (e.g., odds ratio of 1.05 in the Ontario study¹⁰).

Third, the recent large population-based studies provide no evidence that the association between exposure and outcome is greater in younger children (as widely hypothesized and supported by most experimental animal data). This could be a function of a longer than previously suspected period of vulnerability in children, regional differences in brain maturation rates and hence vulnerability affecting phenotype expression, or structural or functional plasticity mitigating acute brain injury from anaesthetic drugs in younger children. It could also argue that the very modest associations found between anaesthesia exposure and outcomes for older children in these studies could also have resulted from confounding by indication or a lack of sensitivity for the outcome measures used to detect or accurately describe phenotypes of anaesthesia neurotoxicity.⁴⁶ Regardless, there is currently no evidence from human studies that age of exposure affects outcome, save an analysis of the Raine cohort where anaesthesia exposure at ages less than 3 yr was associated with deficits in language and abstract reasoning, and later exposure between 3 and 10 yr was associated with decreased motor function.⁵⁸

Finally, the recent population-based studies found that anaesthesia exposure explains only a small fraction of the variability in outcomes measured, much less than other risk factors, such as home environment, perinatal and socio-economic covariates. For example, Glatz and colleagues¹² found that differences in school grades associated with a single anaesthesia exposure were much less than differences associated with sex, maternal educational level and month of birth. A single anaesthesia exposure in this cohort was associated with a mean difference of 0.4% (95% CI, 0.1–0.7%) lower school grades but maternal education level (10–12 vs >12 yr) was associated with a mean difference of 9.9% (95% CI, 9.6–10.2%).¹² Graham and colleagues⁹ also found that socioeconomic factors (e.g. income assistance or the presence of child and family services involvement) were associated with decreases in EDI. In addition to these studies, using a cohort of 3441 children born in the Netherlands, de Heer and colleagues⁵⁹ found that cognition was reduced for 415 children exposed to anaesthesia before 5 yr of age but other covariates such as maternal (e.g. smoking while pregnant) and perinatal (e.g. prematurity) factors were again associated with greater reductions in cognition. Quantifying the relative contribution of anaesthetic drugs to adverse neurodevelopment is one of the most important goals for informing clinicians and families. Even if the association between anaesthesia exposure and neurodevelopment is modest compared with other factors, exposure is potentially modifiable, compared with many of these other factors. Thus, quantifying the relative contribution of anaesthetic drugs to any adverse neurodevelopmental outcomes remains an important goal to inform clinicians and families. However, multiple factors moderating neurodevelopment will continue to complicate attempts to link anaesthesia exposure with neurodevelopmental outcomes.

Conclusions

The question of whether the neurotoxicity observed in experimental animal models after exposure to anaesthetic drugs is clinically relevant in children is challenging, from both scientific and clinical perspectives. Using heterogenous sources and types of evidence to answer this question and inform clinical decision-making is necessary considering the limited potential for further randomized trials in this field. There are however many successful examples of causally linking exposures to outcomes in the absence of randomized trials (e.g. smoking and cancer), but this approach will require much patient work and careful nuanced interpretation. For example, given the fact that paediatric neurodevelopment itself is a very complex process governed by multiple factors that are incompletely understood, it is very likely that exposure to anaesthesia affects some domains but not others and that other factors may moderate these effects. The recent additions to this literature, all with significant methodological strengths, represent important contributions to the available evidence, and provide additional insights into the potential relationship between exposure to anaesthesia in young children and neurodevelopmental outcomes.

Authors’ contributions

Writing and revising paper: J.D.O., D.O.W.

Declaration of interest

None declared.

Funding

None.

Author notes

*

Handling editor: Hugh C Hemmings Jr