“…what do we do when…animal studies do not translate to or may even diverge from clinical findings?”
FOR decades, the question of neonatal anesthetic toxicity has variably met with passionate concern, perplexity, or indifference among the anesthesia practitioner and investigator communities. What began as a laboratory observation and academic curiosity of unknown clinical relevance, leading to clinical research and clinical concern, was elevated to a real clinical predicament by an unexpected 2016 U.S. Food and Drug Administration (FDA) Safety Announcement1 declaring that “repeated or lengthy use of general anesthetic and sedation drugs during surgeries or procedures in children younger than 3 years or in pregnant women during their third trimester may affect the development of children’s brains,” with admonitions to healthcare professionals, parents, pregnant women, and caregivers. This was followed in 2017 by FDA–approved formalized changes to several drug labels to memorialize this warning.2 The aftermath has seen heightened consternation and confusion, with variable response among parents, practitioners, regulators, anesthesiology societies, healthcare institutions, and their risk managers, as well as changes (or not) in informed consent, and several position statements and commentaries.3–5 Having allowed this initial flurry to subside, Anesthesiology this month features two comprehensive review articles6,7 and accompanying editorials8,9 on anesthetic developmental neurotoxicity in animals and in humans.
When a safety “signal” from animal research is detected, whether about therapeutics in clinical trials or (particularly) about drugs or interventions already in clinical practice, that signal must be acknowledged and interrogated. Additional confirmatory and exploratory in vitro and animal investigations may be informative and valuable for understanding the safety, or the lack thereof, of the specific drug. They may enhance our understanding of the immediate signal itself, the mechanisms underlying the signal, or drug safety more generally. Safety signals should also provoke and encourage subsequent rigorous observational and mechanistic clinical investigations. Ultimately, it is only clinical studies that can inform on safety in humans. This must be considered a requisite necessity to enlighten clinicians and patients and to maximize therapeutic benefit and patient safety.
The synergistic coupling of mechanistic animal studies with clinical research, to afford an understanding of why drugs do or do not have therapeutic or toxic effects, can strengthen clinical observation. But what do we do when these animal studies do not translate to or may even diverge from clinical findings?
Considerations about the fidelity of animal research translation to clinical medicine raise the question of whether the relationship represents continuity or dichotomy. What is “translation”? Subsequent to the conceptual formalization of translation,10 the success of that translation has been challenged.11,12 From a therapeutic perspective, examples of promising pharmacologic interventions in animals that did not succeed in humans are plentiful, including experimental approaches to acute stroke, neuroprotection, sepsis, and corticosteroids for head injury (and other indications). From a safety perspective, preclinical findings of toxicity may also not be relevant to humans or even replicated in animals. In contrast to unsuccessful therapeutic translation, the inability to corroborate preclinical toxicity in humans is considered fortuitous. However, the lack of translation in general can be disappointing, frustrating, expensive, and concerning to research and medical communities and to the public and policy makers.
The reasons for lack of translation can be broadly divided into problems with preclinical (models and methods) and clinical research (design and implementation) and overenthusiastic interpretation and messaging of both. Animal research models may be convenient, precise, and reproducible but may not accurately reflect the complexity of human physiology and disease. Animal research methods may have the same problems as those underlying the well-described nonreproducibility of some research in general.13–16 Clinical trials may be premised on fragile preclinical data and models having little relevance to patients. One such example are clinical trials in aged hypertensive stroke patients where outcome is measured 3 months after treatment, which were based on stroke studies in juvenile healthy rats with short-term recovery intervals. Those who conduct therapeutic clinical trials have a responsibility to assure that the foundational preclinical portfolio is sufficiently robust.
This issue’s review articles and editorial views, offering evidence for both anesthetic developmental neurotoxicity in animals and minimal effects of anesthesia on the developing human brain, can be read in the context of this translational framework and the FDA safety warnings. The comprehensive review by Dr. Vesna Jevtovic-Todorovic6 summarizes the evidence for developmental neurotoxicity in mice, rats, guinea pigs, and more recently monkeys caused by a diverse group of drugs, including volatile anesthetics, propofol, benzodiazepines, ketamine, and nitrous oxide. In animals, single and repeated neonatal exposures cause morphologic changes and functional impairments, including neuronal and glial apoptotic damage, mitochondrial dysfunction, impaired synaptic transmission, and changes in behavioral and cognitive outcomes. The key points are that the effects in animals are reproducible and that although they may be a potential foundation for concern in humans, the relevance of the animal findings to humans remains to be confirmed. The comprehensive review of human evidence by Drs. Andrew Davidson and Lena Sun summarizes large retrospective population-based studies and prospective clinical trials. Population studies suggest weak evidence for associations between anesthesia and surgery in early childhood and poor neurodevelopmental outcomes (a very small added risk of slightly worse subsequent academic achievement or later diagnosis of a behavioral disability). Clinical evidence suggests strongly that single, relatively short anesthetic exposures are not associated with increased risk. There may be some evidence for a greater association with multiple exposures. Davidson and Sun7 conclude there is very little if any human evidence to support a recommendation that anesthesia and surgery at a particular age is either safe or unsafe.
What then is the translation of the cellular and animal research to clinical anesthesia? Interpretations vary. One is that there is positive translation of nontoxicity, concluding that “animal models have not demonstrated conclusively that short duration or single-drug exposures consistently cause neurodegeneration or behavioral effects, congruent with clinical trials showing that short-duration, single-drug anesthetics do not pose a significant risk to relatively healthy children.”17 Another is that there is a positive translation of toxicity.9 A third is that there is a translational disconnect between the developmental neurotoxicity of multiple drugs in several animal species and little evidence of increased risk of short exposures in humans.
Why might there be a disconnect? Is it the laboratory methods or the animal models? The cellular and behavioral findings in animals appear based on sound methods and are consistent and reproducible across multiple drugs and species. Do the animal models accurately reflect patients? This is particularly important because developmental neurotoxicity concerns arose first from cellular and animal studies, which were extrapolated from those of fetal alcohol and antiepileptic drug exposure, not based on clinical evidence of neurocognitive deficits associated with neonatal or pediatric anesthetic exposure. Indeed, despite the fact that the FDA warnings were driven largely by animal data,7 safety signals from preclinical studies over more than two decades have not led to the identification of a well-defined and specific clinical phenotype associated with such anesthetic exposure,5 and we face the limitation that clinical studies cannot randomize children to anesthesia versus no anesthesia.
This is not the first experience of the anesthesiology community with concerns of drug toxicity, driven largely or exclusively through animal data, and FDA warnings or drug labels that were based on absent or even contravening clinical evidence. The specialty of anesthesiology has experienced and responded magnificently to such translational challenges before. In response to signals, preclinical scientists established various in vitro and animal models and poked and prodded them to yield mechanistic insights, while understanding that such models were somewhat “idealized.” Clinical scientists investigated whether these problems did indeed exist in patients, and if so, what were the consequences and implications? Was there translation? And if not, why not? What could be learned? What did we learn? Three examples are instructive.
Shortly after isoflurane introduction in 1981, there were suggestions that isoflurane could cause myocardial ischemia due to coronary vasodilation and resultant coronary steal.18,19 Animal models showed that isoflurane could indeed cause steal, ischemia, and regional wall motion abnormalities. These initial observations sparked a controversy. Based on animal models and some clinical data, isoflurane safety was questioned, and restrictions on clinical use were proposed. Nevertheless, no warnings or restrictions were mandated by the FDA based on animal (or clinical) data. Subsequent thorough clinical studies showed no ischemic toxicity and exonerated isoflurane. Hundreds of millions of isoflurane anesthetics were subsequently administered, isoflurane was for a while the most widely used volatile anesthetic in the United States, and isoflurane coronary steal is of no concern today.
During clinical development of sevoflurane, it was identified that sevoflurane could be degraded by carbon dioxide absorbents containing strong base to a haloalkene (compound A), which was nephrotoxic in rats.20 Based on several rat studies showing unambiguous compound A nephrotoxicity and one human volunteer study of sevoflurane, sponsored by a competing anesthetic manufacturer, purporting kidney injury, sevoflurane safety was questioned, a heated controversy ensued, and regulatory approval was imperiled. Sevoflurane was eventually found not to be nephrotoxic21 and was approved, albeit with considerable warnings and restrictions in the drug label required by the FDA (and regulatory agencies in other countries), based largely on the animal data. Subsequent thorough patient studies showed no evidence of any renal toxicity and vindicated sevoflurane. An estimated billion sevoflurane anesthetics have been administered without evident nephrotoxicity. Sevoflurane nephrotoxicity is today a historical footnote. Nevertheless, the warnings and restrictions remain in the label, reminding us that such warnings rarely go away, even in the light of evidence.
Methadone undergoes extensive hepatic metabolism, leading to drug inactivation and elimination.22 Initial in vitro experiments showed that the cytochrome P450 isoform CYP3A metabolized methadone. Based on extrapolation of this test tube data, clearance of methadone in patients was attributed to CYP3A. The drug label and numerous practitioner guidelines warned about potential CYP3A-mediated methadone drug interactions. Nevertheless, neither the role of CYP3A in clinical methadone clearance nor CYP3A drug interactions were ever clinically tested. When well-controlled clinical studies were ultimately performed, they showed that CYP3A was not at all involved in methadone clearance or drug interactions. Rather, another P450 isoform, CYP2B6, was found to be responsible.22 Nevertheless, the information about CYP3A remains in the methadone label.23
These three examples illustrate that in vitro and animal data may not always translate to clinical relevance. The methods were sound, but the models were “off target.” Although no regulatory warnings, restrictions, or clinical recommendations resulted from the isoflurane experience, the same is not true for sevoflurane or methadone. Indeed, the isoflurane experience was said to “underscore the importance of allowing the passage of time before assessing the clinical and scientific impact of a research finding.”19
Scientists endeavor to traverse a linear trajectory from observation to data to information to knowledge. But when do scientific findings achieve sufficient certainty to become “fact,” “knowledge,” or even “conventional wisdom”? When are they certain enough to be enshrined in guidance or regulation? And what do we do when such certainty becomes evanescent? Or even wrong? Scientific or medical certainty is a tenuous concept, if not an oxymoron.24 What we “know” must have the tractability to respond to what we find.
Anesthesiology the specialty has hosted shibboleths before,25,26 and has and can change them through investigation. No amount of animal data on anesthetic developmental neurotoxicity will improve its clinical relevance or answer the ultimate, pertinent clinical question. If we are to resolve whether anesthetics cause clinically significant deleterious effects on the developing brain and address the uncertainties of translation, we must view our operating rooms (and intensive care units) as clinical laboratories, function as clinical investigators, gather the data, and report the findings.
The critical review and helpful comments of several Anesthesiology editors are gratefully appreciated.
Dr. Kharasch is the Editor-in-Chief of Anesthesiology and his institution receives salary support from the American Society of Anesthesiologists for this position.