There are many observational studies investigating the relationship between anesthesia exposure in childhood and later neurocognitive and behavioral outcomes. In this issue of Anesthesiology, Silber et al.  report another such observational study. This study, however, is different, and the data can be interpreted in two different ways, both of which provide important new insights into this research. To appreciate the work at its true value, it is important to remember that evaluating the specific impact of early-life exposure to general anesthetics on subsequent neurocognitive and behavioral development in children is hampered by the inherent complexity of the perioperative period. The concomitant presence of surgery is generally considered as the most important confounding in this regard. Likewise, it is widely agreed that, in contrast to preclinical science, human investigations on developmental anesthesia neurotoxicity are focused on the combined effects of anesthesia and surgery rather than on anesthesia exposure alone.  However, this assumption may still be considered as reductionist because it does not take into account the contribution of other potential factors, such as the effects of hospitalization itself on outcome. Silber et al.  provide us with interesting human epidemiologic big data from otherwise healthy children on how hospitalization for nonsurgical diseases may compare to anesthesia and surgery in terms of subsequent neurobehavioral outcome.  Using the Medicaid database, the authors identified 134,388 healthy children undergoing appendectomy and compared the incidence of subsequent codes for neurobehavioral diagnoses in this population to 671,940 matched healthy controls. Importantly, they have also performed a similar comparative analysis between 154,887 otherwise healthy children hospitalized for pneumonia, cellulitis, or gastroenteritis and 774,435 matched healthy controls. Their analysis reveals that both surgical and medical patients have an increased likelihood of subsequent neurobehavioral diagnoses when compared to healthy controls and that this probability is higher in medical patients when compared to appendectomy patients.

The investigations from Silber et al. are important first because they provide us with new perspectives on the possible underlying mechanisms of any causal relationship between anesthesia and surgery and later neurocognitive and behavioral diagnoses. It has been known for decades that children may develop lasting behavioral problems after hospitalization.  The majority of research in this domain has focused on sick children necessitating intensive care and has repeatedly shown strong associations between hospital stay and adverse neurocognitive and neurobehavioral outcome.  The association described by Silber et al. between hospital stay in otherwise healthy children for common pediatric diseases and the subsequent diagnosis of neurobehavioral disturbances is new but not fully unexpected. Disease-related inflammation may be one of the major culprits. Indeed, both biologic rational and epidemiologic data support a link between early-life inflammation and subsequent psychiatric pathologies.  The role of hospitalization-related psychologic factors may also be important. Whatever the relative contribution of each of these, and probably many other mechanisms underlying the association between hospitalization and subsequent neurobehavioral diagnostic codes, these observations present a strong argument in favor of considering hospitalization as a major confounder when studying developmental anesthesia neurotoxicity.

The second important insight from the work of Silber et al. is that they provide evidence that an increased probability of neurobehavioral diagnosis subsequent to hospitalization may be solely due to increased healthcare utilization and a more thorough posthospitalization medical follow-up of these children and hence the higher odds for diagnosing neurobehavioral disturbances. The association described by the authors between hospitalization and the consequently increased use of dental and ophthalmological services brings some arguments in favor of this possibility.  This type of analysis is known as a control outcome or placebo outcome and is used to check for bias. The principle is that there is no plausible biologic link between exposure (appendicectomy or medical admission) and outcome (dental or eye examination) and so any found must be due to an inherent and unrecognized bias. This same research group has performed a similar control analysis before in another Medicaid data set in which they were seeking to determine whether there was an association between exposure to anesthesia and attention-deficit/hyperactivity disorder medication use.  As a control they looked at, and found, an association between medication use for other conditions and anesthesia exposure when there was no plausible biologic link, implying the presence of an unrecognized bias here too.  Together, these findings provide good evidence that large data sets using administrative data of healthcare use are probably inappropriate for examining the relationship between anesthesia and behavioral outcome, and this undermines several of the previously published studies reporting such a link. It is notable that the majority of studies suggesting a link with attention-deficit/hyperactivity disorder and anesthesia used such data sets, and as such, Silber et al.’s elegant control analysis  significantly undermines any idea that there is any evidence that anesthesia might cause attention-deficit/hyperactivity disorder. The risk of bias in healthcare utilization analyses is mitigated by using research cohorts or data linkage studies not reliant on healthcare utilization, and the findings from these, along with trials, should be given more weight than the healthcare utilization studies.

Putting aside the bias outlined above, one major question stemming from the observations of Silber et al.  is related to the appropriate control group to be used when studying the effects of early-life anesthesia and surgery on neurodevelopmental outcome. So far, the majority of clinical studies on developmental anesthesia neurotoxicity compared anesthetized children to healthy controls. In light of the current results, this may not be the best approach if one will specifically want to study the effects of anesthesia and surgery on outcome. Defined primary outcomes of three major prospective studies did not reveal a difference in general intelligence between anesthetized and nonanesthetized children, whereas a meta-analysis of the secondary outcomes of these investigations suggests that some behavioral changes in children may be associated with having undergone previous anesthesia and surgery.  The fact that hospitalization in the absence of surgery is also associated with comparable neurobehavioral changes suggests that these phenotypes may not be specific to the effects of anesthesia and surgery. This caveat is important to integrate into the design of future epidemiologic studies aimed to address developmental anesthesia neurotoxicity.

How should the work of Silber et al. influence future research on developmental anesthesia neurotoxicity in humans? Should we change the concept and focus on perioperative neurotoxicity rather than on anesthesia or anesthetics-induced neurotoxicity? There are many arguments in favor of such a move. Decades of research in this field has taught us that, in contrast to laboratory science, it is virtually impossible to separately study the impact of general anesthetics on human neurodevelopment. Anesthesia management of systemic and brain homeostasis, surgery-induced inflammation, stress, and other psychologic factors associated with hospitalization are all inseparable components of the complex perioperative period. We do not yet understand how these components interact with each other, nor do we know to what extent anesthetic drugs influence the individual or combined impacts of these factors. It is currently uncertain whether there are any specific neurobehavioral patterns causally linked to early-life perioperative exposure. However, if there are indeed such phenotypes, that may still be considered of potential public health relevance even if they are not causally related to anesthetic drugs. Therefore, embracing a more holistic perioperative view when designing future studies in the field is necessary. At the end of the day, it is not the drug we administer that matters most but the brain health of children we care about.