Commentary on Bechara et al.’s “A Neurobehavioral Approach to Addiction: Implications for the Opioid Epidemic and the Psychology of Addiction”

The article by Bechara et al. (2019; p. 96) rightly brings to prominence the importance of neurobehavioral research in the investigation of the underlying mechanisms of drug abuse and is particularly timely in the context of the U.S opioid epidemic. They refer to a number of theories of addiction that focus on different aspects of the phenomena associated with chronic drug misuse before a detailed account of the special properties of opioids in processing pain as well as reward and a survey of the role of “willpower” and related cognitive control mechanisms in the understanding and possible treatment of drug abuse.

Bechara et al.’s emphasis on theoretical mechanisms is much needed to make progress in reducing initiation, maintenance, and relapse in opioid addiction. In this spirit, I offer a few observations.

To begin, the opponent-motivational-processing perspective is rejected by Bechara et al. in its simple form as an account of opioid addiction because it appears to predict that “cold turkey” should eventually cure heroin addiction, whereas in fact it is a chronic, relapsing disorder. However, there is evidence that withdrawal symptoms can readily be classically conditioned (e.g., Goldberg & Schuster, 1967; Kenny, Chen, Kitamura, Markou, & Koob, 2006), which is consistent with the retrieval of such symptoms as aversive memories that sustain addictive behavior. Thus, it seems premature to discard this theory. There is also an important instrumental (voluntary behavior) component to the opponent theory, in that it is based on the negative-reinforcement principle that heroin-seeking occurs in anticipation of the withdrawal state, which can be postponed or escaped from by taking the drug. In that sense, heroin seeking is analogous to avoidance behavior, which is well known to be highly resistant to extinction, and exhibits the normal phenomena of extinction, such as spontaneous recovery, which again would readily promote relapse.

Instrumental appetitive behavior also plays an important role in drug seeking, as well as in drug use itself, which is most directly studied in animals in terms of schedules in which the drug is administered intravenously (or via other routes) consequent on instrumental responding, often in the presence of drug-related cues (discriminative stimuli and conditioned reinforcers). Such instrumental behavior has been demonstrated for virtually all drugs of abuse but has been most studied for stimulant drugs, such as cocaine and amphetamine. A modern cognitive account of such behavior (e.g., Balleine & O’Doherty, 2010) refers to “goal-directed behavior,” which is governed by specific neural circuits in the corticostriatal system; in the rat, this probably implicates the prelimbic cortex and dorsomedial striatum, equivalent to the caudate nucleus in primates. However, with training, instrumental behavior can become relatively autonomous of the goal and is then referred to as being habitual. This transition is associated with a devolution of control to other corticostriatal systems, notably to the sensorimotor cortex and the dorsolateral striatum (putamen in primates). Despite this transition, the flow of behavior depends on a constant interleaving and blending of goal-directed and habitual components.

This theory, based largely on studies using food as the reward, which attempts to integrate all of the major psychological mechanisms contributing to motivated behavior, has been applied to the problem of drug addiction (Everitt & Robbins, 2005; Robbins & Everitt, 1999) and related to the functioning of the cortex and dorsomedial striatum in goal-directed behavior and of the putamen in habit learning. The role of the nucleus accumbens and its mesolimbic dopaminergic innervation correspond to Pavlovian-instrumental transfer, the process by which classical conditioning can exert motivational influences on instrumental behavior. We believe this occurs most effectively when the stimuli (e.g., contextual cues accompanying drug experiences) also act as conditioned reinforcers and thus as predictive subgoals for future drug experiences.

Such contextual stimuli can provide an increasingly potent and steep gradient of incentive motivation for eliciting drug-seeking (e.g., opioid-seeking) behavior. For example, contextual stimuli could begin with familiar, if somewhat apparently arbitrary environmental features, even including specific people, that have in the past been associated with a heroin high, such as the sight of a street scene. Later, they would perhaps focus on a specific location such as a particular bar or café that elicits approach and entry by the drug user, perhaps to find an old social acquaintance with whom they previously “did” drugs, who happens to have a bag of white powdered heroin with them. After a hurried conversation and mounting excitement, the user might retreat to a familiar restroom and the use of a syringe. There are many anecdotal accounts of how cues, such as the sight of a needle, may cause a “euphoric” high or perhaps a profound feeling of relief in the case of opioid abuse (“needle freaks”). The point here is that all of these environmental stimuli, including people, can predict the drug and so in themselves, via this pairing, gain reinforcing properties via Pavlovian conditioning—probably both positive and negative (conditioned withdrawal) in the case of opioids.

Some conditioned stimuli will simply be incidental and not the outcome of the drug abuser’s actions—this is formally akin to Pavlovian-instrumental transfer, whereby the occurrence of such stimuli add to the motivation of an already initiated drug-seeking sequence. Alternatively, the stimuli can be deliberately generated by opiate users themselves and can even be quite abstract in nature. For example, they could be the thought of obtaining money via theft to buy a new supply of heroin, which generates the search for likely cues that would result in a successful mugging on the basis of previous experience. This process would result in very long, complex sequences of behavior to obtain drugs that can appear to be quite novel (and “flexible”) but are, in fact, part of a rather limited repertoire, maintained (or “bound together”) by a sequence of conditioned stimuli, many of them generated by the drug users themselves. In the laboratory, this drug-seeking behavior can be simulated by so-called second-order schedules of drug reinforcement, whereby rodents or monkeys will emit long sequences of instrumental lever-pressing that only eventually lead to intravenous drug infusions but may intermittently produce response-related stimuli such as brief lights that ultimately occur just before and during intravenous drug infusions (see Everitt & Robbins, 2000). This is then equivalent to a classic Pavlovian-to-instrumental transfer scenario. In my experience, however, the response-produced cues are much more effective as conditioned reinforcers than simply presenting the cues independently of the animal’s behavior.

Additional complexity is provided by the familiar training of such a sequence, whereby the behavior may gain ritualistic elements (even if these are ostensibly novel sequences). Experimental work in rodents has shown that in theory such conditioned stimuli can lose their “value” and yet sustain their response-eliciting (and hence their “drug-seeking”) functions. This can be demonstrated by showing that devaluation of the cues—for example, by poisoning for food (Holland, 2004; Parkinson, Roberts, Dickinson, Everitt, & Di Ciano, 2005), or hypothetically, for example, by pharmacological tolerance (i.e., reduction) of the drug’s subjective effects—fails to blunt the associated drug-seeking behavior, which has attained an habitual, or automatic, quality—quite autonomous of the original subjective sequelae it produced, perhaps several months or even years ago. There may be some subtle differences in the ways in which drugs affect this conditioning. For example, early work by Taylor and Robbins (1984, 1986) showed that stimulant drugs such as amphetamine enhanced the efficacy of conditioned reinforcers (lights and noises) for water reward (to control responding) when infused into the nucleus accumbens in a dopamine-dependent fashion, in apparently early manifestations of “incentive salience.” However, while producing analogous effects, opioids may work better to potentiate more general contextual cues than discrete stimuli such as lights or noises. The general point is that not only do stimuli associated with the drug potentially predict its reinforcing effects, but also the action of the drug itself then potentiates this very action, thus providing powerful control over drug-seeking behavior.

Thus, most drugs of abuse act on the mesolimbic dopamine system, enhancing the efficacy, for example, of conditioned reinforcers (“incentive salience”), but as Bechara et al. (2019) indicate, these drugs also exert effects at other sites, often in a manner dependent on the drug under investigation. Evidence supporting the goal-directed-to-habit hypothesis was recently summarized (Everitt & Robbins, 2016) and will not be repeated in detail here.

The overall implications for addiction are that goal-directed behavior is disrupted by drugs of abuse, partly as a consequence of impaired prefrontal cortical functioning. This also leads to a progressive narrowing of goals in the chronic drug abuser; the tendency to rely on habits is enhanced. This has been shown in studies indicating that cocaine abusers exhibit impairment in appetitive and aversive goal-directed behavior but enhancement in appetitive (though not aversive) habitual responding (Ersche et al., 2016) or, in computational terms, a drift from “model-based” to “model-free” appetitive behavior (Voon et al., 2015). Whether opioid abusers would similarly show a drift to habits, albeit for aversively motivated habitual behavior, would be of considerable interest.

However, following Bechara et al.’s commentary, several points need clarification. First of all, the role of “motivation” in habits requires further definition. It is actually sometimes problematic, and indeed can be quite perilous, to infer motivational processes from subjective accounts. My recent experience in asking chronic drug abusers why they take drugs more often than not does not result in the abusers claiming that they enjoy drugs as hedonic experiences; rather, they often say instead that “they don’t know.” And subjective reports of craving (“wanting”) often apparently occur as post hoc rationalizations of their urges to take the drug. “Must do!” responses may also be considered as possible rationalizations rather than motivational influences per se. Patients with Gilles de Tourette’s syndrome similarly have compulsive urges to perform certain behaviors, as a consequence of dysregulated corticostriatal circuitry; but these are probably not “motivated” in the sense in which Bechara et al. imply.

The drift to habitual control can be seen to parallel the shift to compulsive drug-seeking behavior, which now dominates many of the criteria used by the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (e.g., DSM–5; American Psychiatric Association, 2013; see Clark, Cuthbert, Lewis-Fernandez, Narrow, & Reed, 2017) for substance-use disorders. However, the question of the precise relationship between compulsive behavior and habits is an important one (Everitt & Robbins, 2005). One approach would be to suggest that drug-related mechanisms (e.g., via dopamine release) exaggerate habitual control; this might include drug-induced sensitization. As mentioned by Bechara et al. (2019), sensitization affects the dorsal as well as the ventral striatum. However, there is rather little direct evidence to date to suggest that sensitization plays a major role in human drug abuse. Another plausible mechanism might involve stress, which has been shown to accelerate the development of habitual behavior (Schwabe & Wolf, 2010). Given that stress is a frequent antecedent and concomitant of drug abuse, this might be especially important in the case of opioid abuse. However, as Bechara et al. have reviewed, a disruption of top-down control by the prefrontal cortex is an obvious additional way in which habits can become perseverative and “out of control,” presenting as compulsive drug seeking and encouraging ritualistic drug use. It is perhaps important to emphasize that the prefrontal cortex is a heterogeneous structure, and its “top-down control” mechanisms may take many forms, including the representation of goals in associative (“model-based”) structures, leading to the formulation of plans and the control of impulsive as well as compulsive behavior, probably via different corticostriatal pathways.

Understanding whether drug addiction arises from an exaggeration of bottom-up subcortical processes such as reinforcement learning or a diminution of top-down cortical regulation, such as willpower, or (most likely) both of these, is still to be resolved. Given the nature of drug addiction, the causal influences on these neural mechanisms are also unclear; for example, they could include possible neurotoxic effects of the drugs themselves or predisposing tendencies such as impulsivity, which may explain the large individual variability in vulnerability to addiction. It is also difficult to unravel separate effects of drugs in the face of polydrug abuse; it is possible that the propensity to opioid addiction via prescription opioids may provide new avenues for exploring the specific neural effects of chronic opioid abuse.

In humans, researchers are often restricted to indirect methods for inferring predisposing influences such as endophenotypes, involving study of first-degree relatives of drug-abusing individuals, although increasingly, longitudinal studies of children before (and after) drug abuse are likely to be most useful. The current Adolescent Brain Cognitive Development (ABCD) project of the National Institute of Mental Health (2016), involving longitudinal study of nearly 12,000 U.S. 10-year-olds, looks likely to be informative.

A more economic and rapid means for addressing causality involves the use of experimental animals, a strategy that has engaged considerable research interest. My own work (Dalley et al., 2007), relevant to the emphasis in the target article on impulsive decision making, showed that impulsive behavior in rats, together with reduced dopamine D₂ receptors and a thinner insular cortex (Belin-Rauscent et al., 2016), were predictors of future compulsive cocaine-seeking and risk-taking behavior (Belin, Mar, Dalley, Robbins, & Everitt, 2008). However, impulsive responding was not a predictor of future heroin seeking (McNamara, Dalley, Robbins, Everitt, & Belin, 2010), consistent with different predisposing etiologies for different drugs.

Bechara et al. (2019) speculate imaginatively on future possible treatments for addiction, focusing on training of “top-down” models of self-regulation, such as economic decision making and working memory, as well as “circuit interventions,” such as repetitive transcranial magnetic stimulation (rTMS), which clearly deserve to be tried. The utility of such methods may depend on the stage and severity of the substance-use disorder. If habits are indeed prominent components of compulsive behavior, the latter could be difficult to change (especially as punishment seems to lack efficacy in stimulant abusers; e.g., Ersche et al., 2016). The conventional wisdom is that it is easier to substitute habits than to train habit self-regulation. More radical interventions such as deep-brain stimulation (DBS) may be necessary to convert drug-related habits to a healthier breadth of goal-directed behavior, although animal research indicates that this may be theoretically feasible (Coutureau & Killcross, 2003). The radical interventions might require the use of subcortical interventions such as DBS to appropriate nodes of the corticostriatal circuitry (e.g., ventral capsule, subthalamic nucleus, or the nucleus accumbens; Luigjes et al., 2012), and this approach is being adopted in preliminary trials. A precise neurocircuit-based intervention in humans may eventually conceivably use some of the new neurobiological tools, such as optogenetics or Designer Receptors Exclusively Activated by Designer Drugs (DREADDS), but this is rather far in the future and would depend on ethical debate as well as on technical development.

Overall, the article by Bechara et al. (2019) captures much of the vitality and social significance of contemporary research on addiction, while focusing the endeavor on the remarkable and unique problems posed by the opioid epidemic.