Chasing the Shadows of Implicit Memory Under Anesthesia

Every day, thousands of patients across the globe emerge from general anesthesia to remember nothing of it. This amnesia, an inability to consciously recall what happens to the mind and body while under anesthesia, is so axiomatic that studies addressing a failure rate of only 0.1% are of sufficient public concern to reach the pages of the New England Journal of Medicine1 and Lancet.2 But the effects of anesthetics on cognitive and memory processes in the brain may be far more complex, and less monolithic, than the near-ubiquitous experience of amnesia might suggest. The amnesia of general anesthesia reveals only that at least one critical process in the formation and retrieval of hippocampal-dependent, declarative memory has failed. Largely, we assume that what has failed is consciousness itself: no perception occurs, and no conscious substrate exists to be either remembered or forgotten. But this is not necessarily true. A patient may have intact perception and consciousness and yet report amnesia if there is a failure of memory consolidation processes in the hippocampus. The most pure example of this was the famous amnesiac H.M., who after surgical removal of both hippocampi was unable to form new declarative memories, despite retaining otherwise normal cognitive function. At clinical doses, the amnesia produced by anesthetic drugs will not be so super-selective. But it is clear that many anesthetic drugs possess potent effects on memory function that are dissociable from their effects on arousal and consciousness.3,4 It is entirely plausible that the mind under general anesthesia may not be as dormant as it seems. Our patients may simply not remember. If some level of cognitive processing does occur under anesthesia, but is not consciously remembered, can it still leave a trace? This question is important because any evidence that it does demonstrates neural plasticity in response to external events, and it implies that the brain, and perhaps the mind, that returns from anesthesia is not absolutely identical to the one that enters. What we are searching for is nondeclarative or implicit memory, interchangeable umbrella terms used to collectively describe a cluster of discrete nonconscious memory systems unified only by their being non–hippocampal-dependent. Perhaps of greatest interest because of its role in fear-based psychopathologies, the amygdala-centered emotional learning system has an extensively elucidated circuitry and can be reliably assessed by robust fear-conditioning experimental paradigms in awake human subjects.5 Similarly, in awake subjects, caudate-dependent procedural memory can be assessed through performance improvement in motor tasks. Famously, H.M. was able to learn a hand–eye coordination skill, yet had no memory of ever practicing the task.6 However, in the noninteractive patient under general anesthesia, assessment of procedural memory is clearly not possible, and although some elements of fear-conditioning paradigms could theoretically be adapted, the methodological challenges are tremendous. What remains, then, is priming: an implicit memory process in which exposure to a stimulus influences the response to a later stimulus. As a simple and illustrative example of priming, amnesiacs can name pictures 100 milliseconds faster if they have seen them previously, despite having no declarative memory of the exposure.7 In studies of patients under general anesthesia, the methodology most frequently used has been an auditory adaptation of the word stem completion task (WSCT). In the classic, visual version of this task, presented in 1970 by Warrington and Weiskrantz in a study of four amnesic patients,8 subjects read a stimulus word (for example, CHEST). Later, when asked to complete the word stem CHE_ _, priming is exhibited by an increased likelihood to think of the stimulus word, as opposed to an alternate word, such as CHEAP. In the auditory adaptation, stimulus words are heard instead of read. There is no doubt that the WSCT can detect implicit priming, but the methodology is nuanced and has important vulnerabilities. Foremost, there is the need to dissociate any nondeclarative priming effect from the influence of any declarative memory for the stimulus word. This is usually achieved by some form of the process dissociation procedure,9 in which subjects are asked to exclude items that have been presented; conscious processes will drive avoidance of the target word, whereas nonconscious processes will drive a familiarity response favoring the target word. And, of particular relevance to anesthesia studies, the ability to detect implicit priming is substantially reduced if the modality used for stimulus presentation is different from that used for subsequent testing.10 This occurs because cross-modality priming requires conceptual processing, whereas dominantly perceptual processes are required for successful priming within a singular modality, with visual presentation being optimal. In this issue of Anesthesia & Analgesia, Lequeux et al.11 study implicit priming using the WSCT in patients receiving propofol and remifentanil anesthesia. What distinguishes this study from others in the field is the argument that the probability of implicit memory is modulated by dynamics of arousal related to surgical stimulation. Lequeux et al. therefore compare implicit priming across cohorts with high and low levels of noxious stimulation, which they achieve by altering the remifentanil-to-propofol ratio used to target a constant Bispectral Index value of 50. The principle is that higher opiate concentrations should attenuate spinothalamic tract-mediated activation of the ascending reticular arousal system in response to noxious stimulation, gating the activation of networks associated with perception and learning, including those involving the amygdala. An audio loop of target words is played throughout surgery, and the WSCT and other memory tests performed two to four hours later. Notably, the authors do not include the process dissociation procedure and rely on a test of free recall to determine the existence of any declarative memory. However, because free recall is not a perfectly sensitive detector of all declarative memory, it can be argued that failure to spontaneously remember target words does not eliminate the possibility that declarative processes are influencing performance on the WSCT. Questions regarding this distinction are somewhat immaterial, as in short what Lequeux et al. report is a comprehensively negative result. They detect no difference between the high- and low-opiate groups, and indeed they find no evidence for either declarative memory or implicit priming in either cohort; performance on all memory tests is no different from a control group that heard no words. What interpretations can be drawn from these findings? It would certainly be excessive to conclude that implicit priming does not occur under anesthesia. The literature in the field is quite remarkable for the inconsistency of results, perhaps highlighting more than anything the fragility and nuances of the methodology. But several well-conducted studies have found evidence for implicit priming.12–15 Even allowing for prior negative studies, the results of Lequeux et al., and critiques that the assumptions of the process dissociation procedure produce false positive results,16 it is difficult to view the body of literature in toto and conclude that implicit memory formation simply does not occur. The methodologic difficulties encountered in studying implicit priming in the perioperative environment cannot be understated. The need for cross-modality testing hinders the detection of purely perceptual priming. The wording of instructions and the linguistic characteristics of the test items are critical. Pharmacologic agents necessary for clinical management may have effects on test performance. The memory itself is transient and may decay in strength to below the detection threshold in as little as two hours.10 It would be excessive to attribute the negative results to methodology, although it certainly would have been reassuring if Lequeux et al. had demonstrated that their paradigm was able to produce and detect successful priming in a control group. However, some considerable confidence can be gained from noting that the same investigators have been able to do so in a previous study using a similar technique.17 One lesson from the existing body of literature is that relatively subtle characteristics of the anesthetic state may be important in establishing the appropriate conditions for priming. Lequeux et al. designed their study arms to have different levels of ascending activation related to noxious stimulation, and based on the lower heart rate and blood pressure in the high-opiate group, they likely achieved that goal. It may be that it is not spinothalamic activation per se that is important, but rather the extent to which it activates the ascending reticular arousal system. The remifentanil-to-propofol ratios in the two groups were markedly different, but by targeting a constant Bispectral Index value of 50, the authors may have simply interchanged opiate-mediated modulation of arousal systems with γ-aminobutyric acid-mediated modulation by propofol. Noxious stimulation may be of importance principally when it causes dynamics in arousal that are not counterregulated. The Sebel group, which has most consistently demonstrated implicit priming, found that it is abolished when tight control of Bispectral Index values is successfully executed,18 but not when there are breakthrough events.14 What Lequeux et al. may have observed is the effectiveness of tight control of the Bispectral Index, however, achieved. After a period of considerable interest during the 2000s, there has been little to move the field over the last five years. The study of Lequeuex et al. may not provide the definitive breakthrough, but it should reignite interest and help promote a new generation of work on this difficult and important question. DISCLOSURES Name: Kane O. Pryor, MD. Contribution: This author prepared the manuscript. Attestation: Kane O. Pryor approved the final manuscript. Name: James C. Root, PhD. Contribution: This author prepared the manuscript. Attestation: James C. Root approved the final manuscript. This manuscript was handled by: Gregory J. Crosby, MD.