Uncoupling Immobility and Unconsciousness

Abstract

Anaesthetic endpoints like immobility and unconsciousness are defined at the population level through metrics like MAC and MAC-awake, yet how individuals behave at those concentrations remains poorly understood. Mandel and colleagues address this by characterizing the behavioural dynamics of mice held at fixed concentrations of isoflurane corresponding to either population MAC or MAC-awake, tracking whether an animal rights itself or responds to a paw pinch across repeated testing. Both endpoints flip states unpredictably yet resist switching and, interestingly, are wholly independent of one another.

General anaesthesia requires simultaneous maintenance of multiple endpoints, including immobility, unconsciousness, and amnesia. Of these, immobility is arguably the most operationally central in the clinical setting. Indeed, anaesthetic dosing is standardized around Minimal Alveolar Concentration (MAC), the concentration required to inhibit movement in response to noxious stimulation in 50% of subjects.¹ The related threshold, MAC-awake, captures the concentration at which patients no longer respond to commands, and occurs at substantially lower doses than MAC – roughly 30–60% of MAC depending on the agent.^2,3 In practice, the addition of neuromuscular blockers allows anaesthetic dosing to be maintained below MAC. However, studies using the isolated forearm technique,⁴ in which one arm is shielded from neuromuscular blockade, have revealed that approximately one third of anesthetized patients are able to produce purposeful responses to verbal commands with the spared arm despite being otherwise immobile.^5,6 Although neuromuscular blockers are often used to achieve immobility, many procedures are performed without these drugs, thus relying on volatile anesthetics to achieve this endpoint. Therefore, it is important to understand the dynamics of anesthetic-induced unconsciousness and immobility to determine how they relate within the same individual.

The dynamics of state transitions and individual thresholds can provide insight into the underlying neural mechanisms regulating those states. In psychophysics, perceptual thresholds have long been used to understand how physical stimuli are transformed into subjective experiences. The properties around such thresholds also reflect the probabilistic dynamics of their underlying neural processes. Bistable optical illusions are a useful analogy to illustrate this principle. A Necker cube, shown in figure 1, is an ambiguous visual stimulus that alternates between two mutually exclusive percepts, with transitions occurring stochastically despite constant visual input. Modeling the dynamics of these transitions has yielded testable predictions about the circuit-level processes underlying the two perceptions, including the roles of specific neurotransmitters in regulating perceptual switch frequency.⁷ Accordingly, quantifying the dynamics of state switching is not merely descriptive – it provides a foothold for inferring the mechanisms that regulate the transitions between these states. Doing so for MAC and MAC-awake within the same individual, then, could reveal whether immobility and unconsciousness are governed by shared or independent mechanisms.

Figure 1. Bistable perception of the Necker Cube.

The Necker Cube (middle) is an ambiguous image that can be perceived as either leftward or rightward facing. The two perspectives cannot be held simultaneously. Instead, perception switches between the two states stochastically despite the stimulus remaining constant.

Prior research has already begun deciphering the dynamics around MAC-awake using animal models. While exhaustive repeat testing around anesthetic thresholds is impractical in humans, animal models offer a tractable alternative. Across several species, the proxy for the return of consciousness is the righting reflex: when awake, many animals promptly return to a prone position after being placed supine. At a fixed concentration held at MAC-awake (i.e. MAC-righting), the same animal tested repeatedly will transition in and out of consciousness unpredictably.^8,9 This stochastic switching at a constant anesthetic concentration mirrors the properties of other bistable systems and raises the question: do the same dynamics hold true for responsiveness to noxious stimulation, and if so, how do the behaviours around MAC relate to the threshold of consciousness in the same individual?

In this issue of the British Journal of Anaesthesia, Mandel and colleagues systematically characterize state switches during stable concentrations of isoflurane anaesthesia within male mice.¹⁰ Two separate endpoints were assessed: (1) the righting reflex, a proxy for consciousness, and (2) movement in response to a standardized paw pinch. The goals were twofold: first, to describe how a single mouse responds to noxious stimuli when held at the population MAC for isoflurane, and second, to determine how MAC relates to MAC-awake within the same individual.

Understanding the behaviour of a system that switches between two states requires thinking along two dimensions. The first is how predictable the switching is, and the second is how readily the switching occurs (see Figure 2).

Figure 2. State transitions can be characterized by their stochasticity and resistance to state transitions (RST).

Low stochasticity and low RST describe systems with highly predictable and frequent state changes, like the alternation of feet during a march (Left, Right, Left, Right). Low stochasticity and high RST systems are predictable but change states rarely, like the twice-yearly transition between standard time and daylight-saving time. High stochasticity and low RST systems are unpredictable and change states often, like a coin flip between heads and tails (H and T). High stochasticity and high RST systems are unpredictable yet tend to persist in their current state longer than would be expected by chance alone. Findings from Mendel et al. suggest that responsive-unresponsiveness to pain at MAC and righted-unrighted at MAC-awake (R and U for both) display stochastic responses that are also high in RST, with righting having greater RST than responsiveness to pain.

The first dimension, stochasticity, describes the predictability of the system. A highly stochastic system switches between states in a way that is essentially random, whereas a deterministic system switches in a way that is entirely predictable. Consider two examples that transition states frequently: a fair coin flip (heads or tails) and marching in step (left foot or right foot). Both alternate between two states and spend equal time in each, but they differ in predictability. A coin flip is highly stochastic, in that past coin flips are in no way predictive of the outcome for the next coin flip. Marching is predictable, in that the right foot always follows the left and vice versa.

The second dimension, resistance to state transitions (RST), describes how long a system tends to stay in one state before switching. A system with low RST switches frequently, while one with high RST tends to remain in the same state for extended periods. The twice-yearly clock change between daylight saving and standard time illustrates a system with high RST. The transition happens rarely, but the exact moment it occurs is known well in advance, making it both highly resistant to switching and highly predictable. High RST characterizes many bistable states in biology, where two stable states require effort to transition from one to the other.

To assess whether responsiveness to pain at population MAC displays a similar stochastic property as righting, Mandel and colleagues conducted rigorous experiments that were meticulously designed and executed. They exposed male mice to 0.95% isoflurane, the experimentally derived population MAC for this strain, for five hours. Following a two-hour equilibration period, movement to a painful paw pinch (binary outcome: responsive or unresponsive) was assessed every three minutes for three hours. To characterize stochasticity, the authors analyzed the distribution of dwell time lengths, which reflect the number of consecutive identical outcomes before a state change. For example, a responsive dwell time of 4 indicates the animal responded to four successive paw pinches but did not respond on the 5^th pinch. Dwell times were then pooled across animals to construct a population-level distribution.

Stochastic systems can exhibit a range of statistical properties. Mandel and colleagues asked whether responsiveness to pain was memoryless (regarding state memory, not memory of the painful stimulus). If so, event timing would be consistent with a Poisson point process, in which time between events follows an exponential distribution.¹¹ Dwell times map naturally onto this framework since they capture how long a system remains in each state before transitioning. The authors showed that both sets of dwell times were exponentially distributed, supporting the interpretation that responsiveness to pain at the population level is well-captured by a Poisson point process. In other words, MAC appears to exhibit memoryless dynamics.

A potential limitation of this approach is that pooling dwell times across animals yields a population-level metric that may not necessarily reflect the dynamics within any individual. An exponential distribution at the population level can arise even when individual behaviour is not memoryless, if there is sufficient heterogeneity among individuals. The sleep-wake cycle illustrates this conceptually. Circadian rhythms make an individual’s sleep timing highly predictable, but because those rhythms differ across individuals, aggregating sleep-wake states across a population can appear more stochastic than any single individual’s sleep patterns. The same ambiguity applies here. The population-level exponential fit is consistent with stochasticity but cannot distinguish between genuinely memoryless dynamics within individuals versus a more structured dynamic that varies across mice.

At population MAC, responsiveness to noxious stimulation varied more across mice than within the same mouse over repeated tests. This suggests that stable individual differences play a large role in the dynamics of responses around MAC. These findings also stand in contrast to the righting reflex at MAC-awake, where between-mouse and within-mouse variability were found to be roughly equal, suggesting less contribution from stable individual differences. However, it should be noted that in their earlier work with a larger sample size, individual differences were detectable in righting response.⁸ Together, this raises the possibility that MAC is better defined by individual characteristics than MAC-awake.

Finding individual differences along any parameter in a single-sex, age-matched, inbred mouse strain is notable. These animals represent a population where genetic sources of variability are already minimized by design. Yet pain processing is inherently variable even in the absence of genetic diversity, as environmental factors driving epigenetic modification can produce divergent pain thresholds even among genetically identical individuals.¹² If MAC varies substantially across individuals but MAC-awake does not, then individuals with high pain thresholds may require less anesthetic to achieve immobility, but not necessarily less to achieve unconsciousness.

The authors then investigated whether responsiveness to pain at MAC exhibits resistance to state transitions (RST). To do this, they compared how often state changes were observed in each mouse (i.e. how often a responsive trial was followed by an unresponsive one and vice versa) against what would be expected from a Bernoulli process with the same overall responsiveness rate. A Bernoulli process serves as the baseline where each trial is an independent draw with no memory of the previous outcome, like a coin flip. An animal that transitions between states significantly less often than this Bernoulli benchmark (given the same underlying response rate) can be said to exhibit high RST. In contrast, if a state changes more often than that Bernoulli benchmark (ex. marching, probability of a state change = 1), then it can be said to have low RST.

The authors found that both righting and responsiveness to noxious stimulation show greater RST than a Bernoulli process, indicating that mice are more likely to remain in their current state than expected by chance alone. Comparing the two behaviours further showed that loss of righting exhibits higher RST than immobility. One interpretation is that transitions between consciousness and unconsciousness are more strongly constrained and require larger perturbation than transitions between responding and not responding to noxious stimulation. From a biological plausibility standpoint, this distinction makes sense. Pain responses need to remain adaptable to environmental demands, whereas conscious states tend to be more stable, supported by network dynamics that favor persistence.

The difference in RST between MAC and MAC-awake offers a window into the mechanisms that stabilize each state. Near the threshold for consciousness, opposing wake- and unconsciousness-promoting circuits resist switching, producing longer dwell times. In sleep neurobiology, this phenomenon is consistent with reciprocal inhibition between wake-active and sleep-active pathways, or the “flip-flop” theory of sleep.^13,14 In this framework, whichever state is dominant actively suppresses the other, and the high RST near the threshold for consciousness is the consequence of this hysteresis. Responsiveness to noxious stimulation, by contrast, lacks an analogous stabilizing architecture in the spinal cord. Nociceptive signals do not inhibit a corresponding “non-nociceptive” network, and so the reciprocal inhibition that prolongs dwell times near MAC-awake is absent near MAC, reflected by the lower observed RST.

Finally, the authors tested whether pain responsiveness at MAC and the righting reflex at MAC-awake were correlated within animals, and they found no relationship. Mice more responsive to noxious stimulation at MAC were not more likely to right at MAC-awake, and mice with higher RST for pain responsiveness did not show higher RST for righting. The results suggest that MAC and MAC-awake are independently regulated processes.

The observed independence between righting and immobility aligns with distinct biological mechanisms driving MAC-awake and MAC. Although the righting reflex originates in the rodent brainstem and does not exist in humans, it is often used as a surrogate for MAC-awake because both endpoints indicate loss of arousal and are produced by comparable doses of inhaled anesthetics. MAC is largely mediated by spinal processes. For instance, compared to systemic delivery, anesthetic delivery restricted to the brain results in a substantially higher dose requirement to achieve immobility.^3,15 Similarly, decerebrated rats exhibit the same MAC as intact animals, providing additional evidence that the spinal cord mediates anesthetic-induced immobility, rather than the brain.¹⁶ The righting reflex, by contrast, depends on multiple brain regions. A rapidly expanding area of research is focused on defining the many neural circuits that modulate the righting reflex with the goal of accelerating return to consciousness without altering anesthetic pharmacokinetics.^17,18 Further supporting this dissociation, genetic variants that affect pain sensitivity, such as the redhead gene, influence MAC but not the incidence of intraoperative awareness.^19,20 However, unlike animal studies investigating fundamental mechanisms of immobility and unconsciousness induced by a single anesthetic, in clinical practice these endpoints are typically achieved by a combination of agents including volatile anesthetics, opioids and neuromuscular blockers. Therefore, the observed independence between righting and immobility in this animal study, while mechanistically informative, is not directly applicable to the clinical setting.

Taken together, the findings reported by Mandel and colleagues provide a compelling quantitative approach to describing the behavioural state transitions around MAC. The memoryless dynamics, at least at the population level, and measurable resistance to state changes provide a basis for understanding how spinal reflexes are modulated. Some ambiguity remains, and whether stable individual differences in a heterogenous population contribute to what appears to be a stochastic dynamic remains an area for future exploration. Describing these processes at the individual level could open the door to identifying the underlying mechanisms that drive a surprisingly dynamic system. As it turns out, during steady-state anaesthesia, the only truly steady variable is the anesthetic.