The Validity of Rodent Models of Addiction

The Validity of Rodent Models of Addiction

Tom Arend Ruiter

Literature Thesis - MSc Brain and Cognitive Sciences

ABSTRACT

A major part of the body of research into addiction is performed in rodents. The question whether these rodent models of addiction are in fact a valid model of human addiction remains an ongoing debate in the field. Here, a summary is given of popular classical rodent models and their main criticisms, such as the lack of behavioural choices these rodents face and the high percentage of rats that adhere to these mod-els’ definitions of addiction. Recent improvements that address these shortcomings, like the 0/3 crit model and (social) choice models, are discussed. Furthermore, a critical look at the overall validity of rodent models is taken by assessing their homological-, face-, pathogenic-, mechanistic- and predictive validity. Fundamental differences between rodents and humans are discussed, including the inability of rodents to take long-term and abstract goals into account, differences in delay discounting between the species, and the importance of language in addiction. It is concluded that these models show increased motiva-tion of some rodents to self-administer drugs but miss key aspects of addiction.

KEYWORDS

addiction, animal models, rodents, validity

1

INTRODUCTION

Addiction has a high impact on the addicted, their families, and society. In the USA alone, yearly costs associated with abuse of alcohol, tobacco and illegal drugs are estimated to be over $700 billion from lost productivity and healthcare and crime related ex-penses, increasing when also taking into account the burden that an addicted person exerts on his or her family (Orford, Velleman, Natera, Templeton, & Copello, 2013; Volkow, Koob, & McLellan, 2016). Around 8.5% of the population has a substance use disorder, which is defined by the latest edition of the Diagnostic and Statisti-cal Manual of Mental Disorders (DSM) as adhering to at least two of the symptoms in table 1 where a severe substance use disorder ticks off at least six symptoms (American Psychiatric Association, 2013). Here, addiction is generally referred to as this most severe stage of substance use disorder where the disorder is chronic and accompanied by a considerable loss of self-control (Lipari, Hedden, & Hughes, 2013; Volkow et al., 2016).

This neat definition is contrasted by discussion among researchers and health care professionals about the nature of addiction. The central question in this discussion is if addiction is best conceptual-ized as a disease of the brain, or not (Heather, 2017). Proponents of the predominant brain disease model of addiction argue that prolonged substance use results in such dramatic changes to the brain’s reward-, cognitive control- and emotional circuitry that addicts are victim to these brain changes. Consequently, they argue that addiction is therefore an acquired disease caused by unhealthy (but voluntary) behaviour combined with a genetic predisposition,

like diseases such as diabetes (Volkow et al., 2016). Due to the neuro-biological changes, this voluntary drug-taking behaviour switches to compulsive drug seeking at a certain point, after which the stage of addiction is reached.

Much fundamental research into this neurobiological underpin-ning of addiction is being done using rodents, as these animal models allow for experimental control and manipulations that are not possible in human subjects (Spanagel, 2017; L. J. Vanderschuren, Minnaard, Smeets, & Lesscher, 2017; Venniro, Caprioli, & Shaham, 2016). Brain changes due to prolonged drug use in rodents are used as evidence in favour of the brain disease model of addiction (Hall, Carter, & Forlini, 2015). The underlying assumption in this line of research is that animal models of addiction are adequately analogous to human addiction that research results in rodents trans-late to the human condition. But is this indeed the case? Research into pharmaceuticals to alleviate addiction show that preclinical (animal) trials that looked promising do not translate to clinically relevant results in humans (Field & Kersbergen, 2019). Indeed, there is an ongoing debate going on in the field of addiction research about the extent to which rodent models of addiction can speak to the human condition (de Wit, Epstein, & Preston, 2018; Field & Kersbergen, 2019; Heilig, Epstein, Nader, & Shaham, 2016; Müller, 2018). A critical look at whether rodent models do indeed model human addiction adequately is warranted.

This leads to the central question asked here: how valid are rodent models of addiction? In this paper, I will describe the different types of animal models, their main criticisms, recent models that address these shortcomings, and critically look at these models using five different subtypes of validity. Lastly, the consequences of the limited validity of animal models for addiction research is discussed.

Table 1: DSM-5 criteria for substance use disorder (1) Hazardous use

(2) Social/interpersonal problems related to use (3) Neglected major roles to use

(4) Withdrawal (5) Tolerance

(6) Used for larger amounts or longer than intended (7) Repeated unsuccessful attempts to quit/control use (8) Spending extensive time using

(9) Continued use after problems related to use (10) Activities given up to use

(11) Craving

2

TYPES OF MODELS

First, I will broadly describe the most influential and widely used classical models of addiction. Important to note is that there is no one animal model of addiction that models all aspects of addictive behaviours in humans, but rather there are different models that try to model specific aspects of addiction. We can discern different categories of rodent models based on in what way they try to induce

substance dependence, and how they operationalize the concept of addiction in rodents (Sanchis-Segura & Spanagel, 2006; Spanagel, 2017).

Alcohol vapor models.Exposing rats intermittently to alcohol va-por for an extended period of time has been used as a model to produce alcohol dependence in rats (Vendruscolo & Roberts, 2014). Here, the experimenter tries to model the effects of long-term al-cohol consumption, with as dependent measures increased self-administrations of alcohol or signs of withdrawal when alcohol vapor administration is ceased. These include purely physical signs such as hyperactivity and tremors, and motivational aspects such as increased anxiety responses to stressors (Gilpin, Richardson, Cole, & Koob, 2008). Importantly, the rats do not have a choice in whether they receive the alcohol doses or not, as the exposure to alcohol is under the experimenter’s control.

Non-operant models.Non-operant models of addiction are mod-els where rodents are allowed to directly self-administer the drug, without needing to make an operant response. This type of ad-ministration is usually restricted to models of alcohol addiction, as other drugs such as morphine or amphetamine show reduced reinforcing properties when consumed orally (Sanchis-Segura & Spanagel, 2006). Usually, this is implemented by offering the rodent the choice between two bottles in their home cages: one filled with water, the other with alcohol (Spanagel, 2017). The dependent mea-sures used here, and how addiction is measured, are usually the alcohol consumption levels and the extent to which animals prefer the alcohol bottle to the water bottle.

Operant models.As opposed to non-operant models, in operant contexts lab animals need to perform an action in order to receive a dose of the drug. In the majority of studies this is done by letting the rodents associate a lever press with a drug award under a fixed ratio (where the same amount of lever presses produces a reward) or a progressive ratio (where an increasing amount of lever presses produces a reward) (Sanchis-Segura & Spanagel, 2006). As a dependent measure, under the progressive ratio the breaking point is often used, which is the maximum amount of lever presses an animal was willing to make in order to obtain a drug reward, where it is found that this breaking point is higher after escalated drug use (L. J. M. J. Vanderschuren & Ahmed, 2013). Another measure used in this context is the reinstatement of drug seeking, described below.

Reinstatement models.Reinstatement procedures are a type of model that aims to resemble the drug seeking and relapse periods that are typical of human addiction. In rodent models, the lab animals are first exposed to a self-administration period where they perform an operant action to obtain drugs (e.g. lever pressing). Next, an extinction phase is introduced where the operant action does not produce the drug administration so that the rodent stops perform-ing the operant action. Lastly, a reinstatement-inducperform-ing action is performed to induce relapse. This reinstatement-inducing action generally consists of either (i) induction of stress by for example foot shocks, (ii) cues associated with drug taking being reinstated again (e.g. a sound was played every time the rodent self-administered by lever pressing, during extinction this sound was not present, and now the sound is played again when the lever is pressed), or

(iii) drug priming – a small dose of the drug being injected. De-pendent measures here are the increase in operant responses after the reinstatement-inducing action compared to during after extinc-tion (Sanchis-Segura & Spanagel, 2006; Spanagel, 2017). In this last phase, an operant response does not result in drug-administration in order to prevent any psychomotor effects of the drugs, and thus purely look at how much the rodents ‘want’ the drug. A variation on this is the conditioned place preference, where a conditioning box with different compartments is used, and entering one com-partment is paired with injecting a drug, while the other is paired with injecting an empty vehicle. Then, this association is extin-guished by injecting saline instead of the drug. After extinction, a reinstatement-inducing action (such as induction of stress) is performed, and the preference of the animal for the drug-paired compartment in comparison with the vehicle-paired compartment is assessed as dependent measure (Spanagel, 2017).

Abstinence-based relapse models.Abstinence-based models resem-ble reinstatement models, but differ in that they do not go through an extinction period but rather a period of abstinence in a differ-ent context (i.e. a differdiffer-ent cage). The dependdiffer-ent measure here is the number of nonreinforced operational responses after they are brought back into the original environment – again without these operational responses producing drug administrations (Venniro et al., 2016).

3

CRITICISMS ON MODELS AND RECENT

IMPROVEMENTS

Factors that classically have been criticized in these models are (i) the extremely high proportion of animals that adhere to the criteria of addiction of these models, and (ii) the lack of choice between other behavioural options except drug use. The overarching ques-tion between these two criticisms is whether the behaviour we see in rats can really be described as addiction, or if it is better described as non-problematic substance use that is to be expected when no other behavioural alternatives are provided. Furthermore, (iii) the lack of social interaction that these animals are provided during these experiments has been criticized, as rodents are social animals and the presence of other rodents has a clear effect on drug using behaviours.

To expand on the first criticism: in operant self-administration procedures, over 90% of the rodents will voluntarily self-administer the drugs they are provided if they are given no other behavioural choice (Heilig et al., 2016). This criteria of self-administration as a model for addiction is in stark contrast with the human condition, not in the least part because in humans a much smaller amount of initial users will eventually develop an abuse disorder for any particular drug (Perry, Westenbroek, & Becker, 2016).

Secondly, when rodents are given the choice between a non-drug reward (usually food or sweetened water) and administration of a drug that they were previously self-administering, an overwhelm-ing majority of rats will opt for the former option with only a small proportion of the lab animals showing a preference for the drug. This has been observed for cocaine, methamphetamines, nicotine and opioids (Ahmed, 2018). For cocaine, even after long-term drug

use at high doses – circumstances when animals would be consid-ered addicted in these models – the large majority of animals prefer a non-drug option to cocaine administration (Cantin et al., 2010).

Furthermore, this has not only been shown in self-administration choices, but also in reinstatement models (Ping & Kruzich, 2008). The main criticism that this line of research points to is that in ex-perimental setups where rats only have the choice to consume the available drug, it is impossible to distinguish actual substance abuse of drugs versus drug use that is done by lack of other available be-haviours – how ‘addicted’ are they if they would not self-administer in the presence of a food choice?

As for the last criticism, social interaction is an often ignored factor in rodent research, where this social interaction can play different roles: either as a non-drug alternative reward or as a protec-tive factor against addiction. A choice between self-administrating a drug (methamphetamine or heroine) or receiving contact with another rat stops the self-administration of these drugs, even in rats that met all criteria to be considered ‘addicted’ in the context of operant self-administrated models of addiction (Venniro et al., 2018). This is a similar result as described above with food rewards. This so called social-choice-induced abstinence also prevented typical methamphetamine craving and relapse (Venniro et al., 2018). Fur-thermore, even if social interaction is not provided as an alternative to drug administration, but rather as an addition to the lab animals’ environment (i.e. rats are kept together in cages where they make the drug-related choices), it has a moderating effect on drug use. The classic rat park studies are first examples of this, where rats housed in enriched environments with multiple behavioural op-tions and together with other rats – thus resembling their natural colonies - consumed less morphine than single-housed rats (Alexan-der, Beyerstein, Hadaway, & Coambs, 1981; Alexan(Alexan-der, Coambs, & Hadaway, 1978; Hadaway, Alexander, Coambs, & Beyerstein, 1979). Although the interpretation of these results has been called into question (Gage & Sumnall, 2019), recent studies based on similar ideas show similar results: in the presence of a non-drug-consuming peer, cocaine use is inhibited (Smith, 2012).

Combined, these three criticisms give clear points for improve-ment for older models, which recent models have attempted to address. For the first criticism, where the large majority of rodents that are exposed to the experimental procedures are meeting the model’s conditions for being addicted, alternative rodent models are suggested where only a smaller proportion of rats are considered addicted, with more of a focus on problematic use. The difference between these models is a different definition of addiction, instead of necessarily different ways to induce the ‘addiction’. One such example is the 0/3 crit model, which attempts to model problem-aticsubstance use - and thus addiction - more accurately. Here, analogues of three clinical features of substance abuse in humans are tested in rats to assess whether they are addicted: (1) difficulty stopping drug use, as assessed by continued cocaine seeking during periods where it was not available, (2) high motivation to take drug, as assessed with a progressive-ratio schedule to measure the break-ing point, and (3) substance use despite harmful consequences, as assessed by the persistence of animals to keep responding through a foot shock (Deroche-Gamonet, Belin, & Piazza, 2004). This last symptom is based on studies which found that rats which had limited exposure to cocaine were discouraged to take cocaine if it

was accompanied by a foot shock, but after prolonged cocaine self-administration this foot shock no longer functioned as a deterrent, thereby showing that this ignoring of foot shocks was an effect of long term drug use (L. J. M. J. Vanderschuren & Everitt, 2004). As experimental groups, the rats that adhere to zero of these criteria are compared to those that adhere to all three criteria, thereby mak-ing a comparison between non-addicted and addicted specimens possible. Another similar study where a definition of addiction is used that results in a smaller proportion of addicted rats is a recent study where rats are trained to self-administer alcohol in an operant setting (Augier et al., 2018). After ten weeks of self-administration the rats are given the choice between sweetened water and alcohol. In this choice, fifteen percent of the rats showed preference for the alcohol self-administration as opposed to a sac-charine solution. Furthermore, this fifteen percent showed more addictive-like behaviours for alcohol compared to the eighty-five percent saccharine-preferring rats, as assessed by higher breaking points for a progressive ratio reinforcement schedule for alcohol and a higher tendency to sustain negative consequences in the form of quinine adulteration and foot shock punishments in order to keep consuming alcohol. This fifteen percent ‘addicted’ rats cor-respond to the approximately fifteen percent of humans exposed to alcohol that develop substance-abuse related problems (Augier et al., 2018). The study mentioned above also addresses the second main criticism on classic models: it incorporates the choice between a drug and a non-drug alternative, and only considers the rats that prefer drugs as addicted.

The last criticism – on the lack of social factors incorporated in animal research – has been addressed in a recent model too. In a similar way as the aforementioned study gave rats a choice between sweetened water and alcohol, here the rats were given a choice between obtaining a drug (methamphetamine and heroine) and obtaining a social interaction (Venniro et al., 2018). This social interaction consisted of access to another rat for sixty seconds. In almost a hundred percent of choices the rats choose the social- over the drug option. In a follow-up experiment, the authors delayed the social reward to devalue it, so that the percentage of rats abstaining from drugs mimics the 40-50% of humans abstaining from drugs after being treated with the community reinforcement approach or contingency management (Stitzer, Jones, Tuten, & Wong, 2011) – both ‘social’ treatments. As a recommendation for future research they advise to use the lab animals that chose the drug over the delayed social reward as models for pharmacological interventions or fundamental research into the neuroscience of addiction.

Now that we have discussed classical animal models of addiction, their main criticisms and recent advances in the field to address these criticisms, I will discuss remaining differences between the (improved) animal models of addiction and the actual, human con-dition.

4

VALIDITY OF MODELS

In order to investigate the validity of these rodent models and structurally investigate differences between the rodent model and the human condition, first it has to be established what is meant by validity in the context of animal models.

Classically, the most widely used definition in this field is pro-vided by Willner, who identifies three separate types of validity (face-, construct- and predictive validity) and argues that these three constitute to the overall validity of an animal model (Will-ner, 1984). Willner defines face validity as the phenomenological similarities between the model and the psychopathology being modelled, construct validity as the match in theoretical rationale behind the abnormalities in the model and the real disease, and pre-dictive validity as whether the predictions made from the model are shown to be correct in humans too (e.g. preclinical trials translating to clinical significant results) (Willner, 1984). More recently, the shortcomings of these criteria have been criticized: for construct validity for example, a specific, widely accepted theory of the exact mechanisms of the psychopathology is needed - which we do not have – and even if we had there is no univocal way to translate such a theory directly to animal models (Treit, Engin, & McEown, 2010). Therefore, Belzung & Lemoine have developed more specific criteria that make for a clearer framework to discuss each individual aspect of animal models than Willner (Belzung & Lemoine, 2011). The rationale behind this framework is that an animal model is valid insofar as it is similar to the modelled human disease in all aspects that play a role in the disease: from the normal functioning of the specie, to the way a specimen develops the disease (risk-and triggering factors), to the mechanisms underlying the disease, to the symptoms that are being expressed, and the influence of (pharmacological) interventions on the disease.

Therefore, in order to assess the overall validity of an animal model, they propose to look at homological validity (does the specie resemble humans?), face validity (do the behavioural symptoms resemble that in humans?), mechanistic validity (do the theoretical cognitive or neurobiological mechanisms that produce the observ-able effects of the disease resemble that in humans?), pathogenic validity (does the way the animal transforms from a healthy into a pathological organism resemble that in humans?), and predictive validity (does the relation between a therapeutic agent and the observable effects of the disease resemble that in humans?). This provides a clear framework in which I will discuss different aspects and limitations of rodent models of addictions.

4.1

Homological validity

Homological validity refers to the ways in which the animal itself resembles and differs from humans. Rats are, like humans, social animals: they naturally live in large hierarchical communities (El-lenbroek & Youn, 2016). Furthermore, they show great behavioural complexity and are able to operantly learn task rules, and are able to attribute incentive salience to cues, making it possible to teach them about cues related to a drug (Parker et al., 2014).

The first key difference I want to discuss in this context is one in decision making, namely about the role of intent in addiction. If we take a look at the DSM criteria for substance use disorder (table 1), two of the criteria are essentially about intent: taking the substance in larger amounts or for longer than intended, and wanting to cease or reduce substance use but not being able to -these symptoms are core features of addiction, as they exemplify a loss of control over one’s behaviour. This focus on intent makes it very hard to model these symptoms in animals, as the key in both

symptoms is behaviour that is being performed against a previously formulated wish not to, and therefore can only be assessed by un-derstanding the animals subjective goals (Field & Kersbergen, 2019). How can we tell the difference between rodents intending to stop using because of negative consequences (e.g. foot shocks), or them making a deliberate choice to endure the negative consequences because the rewarding effects of the drugs are too high? This is an as of yet unanswered question, and it is doubtful that a rodent model can be developed that can take into account the intent of rodents (Field & Kersbergen, 2019). This role of intent is also im-portant in the recovery of addiction: most addicts and clinicians agree that the intent and aspirations to cease drug use are crucial for a successful recovery (McConnell, 2016). This extents to the importance of human’s self-awareness and the importance of one’s self-image in recovering from addiction. Realizing that one’s action do not stroke with the person they aspire to be is a strong motivator for overcoming addiction (McConnell & Snoek, 2018). This kind of high-level self-narration is unique to humans, and thus cannot be translated to rodent research. An example of such an abstract, long-term consideration that can be taken into account by humans but not rodents is the fact that people who perceive themselves as having a higher risk on addiction, are more likely to moderate their drinking behaviours (Haller & Chassin, 2010).

Furthermore, one perhaps obvious difference between human and rodent drug use, but one worth exploring, is the lack of language use in rodents while this is a ubiquitous part of human life and an important aspect of initiating, continuing and stopping drug use (de Wit et al., 2018). The choice for initial drug use is very much influenced by factors communicated through language: the expected short-term effects of the drug, the warnings for the long-term consequences of drug use, and of course the factor of social pressure which has been shown to have a sizable influence on initial drug use (J. A. Stein, Newcomb, & Bentler, 1987). The effects of language on initial use also extend to the experience of the drug’s effects: the user has information on the effects of the drug and, most likely, expects a positive experience. As the high of some drugs can be interpreted both positive and negative, depending on the social influences and expectations surrounding the drug use, these expectations are being shaped by the language and information we receive about it (de Wit et al., 2018). Furthermore, as nicotine and opiates can produce unpleasant effects when used for the first time, the user can be warned for this and take this information into account when deciding to use again (de Wit et al., 2018). Although rodents do have the ability to learn certain kinds of information from individuals of their own species, such as fear responses (Debiec & Olsson, 2017) or food preferences (Galef, 2002), this kind of social learning is limited to these broad mental states and preferences as opposed to the information-dense and context-rich language use of humans and, importantly, social learning is not used in the recent models discussed here. An animal in these models is not able to take this information into account and has to learn about the valence of the drug effects by itself, without knowing what exactly his happening to it or how to interpret the new experiences, resulting in an arguably different initial use. These factors call into question whether the same mechanisms underlie the selection of which humans develop addiction and which rats will. Furthermore, language is important in the ending of drug use. This ties in with

the fact that rodents cannot take long-term goals into account: in humans, these goals are communicated through language, such as warnings for the addictive effects of drugs or the harmful effects on physical and mental health caused by long-term use. Very rarely are the negative consequences of drug use in humans as direct as they are in animal models, where the rodents for example get direct foot shocks when they make the choice to consume the drug. Indeed, some effective therapies play directly into this language aspect, such as cognitive behavioural therapy, acceptance and commitment therapy, and motivational enhancement (de Wit et al., 2018).

4.2

Face validity

Face validity refers to the question whether the behavioural symp-toms of the animal model resembles those seen in humans.

This aspect of the 0/3 crit model is easy to discuss, because it directly maps its criteria for addiction in rats to characteristics of substance abuse in humans; a strong point of this model. First, in this model, human difficulty in stopping drug use is modelled as continued cocaine seeking when a rodent is signalled with a light that cocaine was not available (Deroche-Gamonet et al., 2004). This is not a clear analogy with human reality: human addicts are not looking for the drug while it is impossible for them to get it. The key symptom that is trying to be modelled here is that humans who deal with addiction often intent to stop using drugs but cannot manage to refrain from drug use. This intention to stop but not being able to is hardly the same as continuing to seek a drug when one knows it is not available. Secondly, a high motivation to take the drug in humans is assessed with a progressive-ratio schedule, where addicted rats have a higher breaking point than non-addicted rats. In contrast to the first criteria, this is a clear analogue of an increased motivation in humans; the amount of labour an individual is prepared to invest to obtain a dose of the drug in question. Thirdly, continued substance use despite harmful consequences in humans is modelled as the persistence of animals to keep performing an operant response to obtain the drug, even though they are given a foot shock in addition to the drug dose. This analogy breaks apart when considering the different characteristics of these consequences faced by rodents and humans. For humans, the harmful consequences that they face are usually more long-term (e.g. in the end no longer being able to live a functioning life, ostracism by family and friends, or health consequences) (de Wit et al., 2018). For rodents, the consequences are direct and a simple cost-benefit calculation: if they value taking of the drugs more than they fear an electric shock, they will still take the drug; there is no conflict between direct- and long-term, more abstract goals. Therefore, in my view this measure more resembles an additional test for high motivation to take the drug, rather than of the dilemma human addicts face, namely that of harmful long-term consequences versus having a short term fix. Looked at in this light, the 0/3 crit model does measure higher motivation to consume a drug, but this alone does not constitute to addiction.

For choice models, where only animals that prefer the drug over an alternative reward are considered addicted, the face validity comes from the analogous situation of human addiction where taking or seeking the drug is preferred over other activities – a compelling analogy. At face value, one way in which this rodent

choice differs from those of humans is again the types of rewards that are taken into consideration: the rodent choice is simple, and direct. The human choice often includes long-term consequences, and abstract goals; it is never as simple as a short-term choice be-tween a drug being immediately administered or an alternative reward being offered. Furthermore, one aspect that is brought up as evidence for face validity in multiple papers of choice models are the comparable proportions of rats that prefer a drug above a non-drug choice compared to the proportion of humans that de-velop substance abuse problems out of all people that are exposed to the drug. For example, less than 15% of rats will continue to prefer cocaine over water sweetened with saccharin (Cantin et al., 2010), comparable with the proportion of people who develop co-caine addiction out of the whole population that tried it (Anthony, Warner, & Kessler, 1997). Although this apparent consistency of only a small proportion of drug users getting addicted across species seems like a strong point in favour of the validity of rodent models, there are two important points of criticism here. First, the exact proportion of drug-preferring rats versus non-drug-preferring rats is to a certain extent arbitrary, as it strongly depends on the height of the drug reward (i.e. how strong the drug doses are) and the height of the non-drug reward (e.g. how sweet the water is). In the example previously mentioned of 15% of rats preferring cocaine over sweetened water, an experiment with low concentration sweet water found that only half of the rats prefer cocaine over this less sweet water (Cantin et al., 2010); thus, this proportion is a function of how rewarding the alternative option is. This same phenomenon can be found more explicitly in a recent paper which offers rats a choice between social interaction and a dose of methamphetamine or heroine, where basically a hundred percent of the rats prefers social options to the drug options (Venniro et al., 2018). In order to decrease this percentage, they introduce a delay to the social reward, so that rats have to wait longer for and thus will prefer it less because of delay discounting. The authors note that this delay makes it possible to find a spot where 40-50% of rats prefer the social option of the drug option, to resemble the percentage of human addicts that respond to a community reinforcement ap-proach or contingency management – a rather arbitrary choice. This brings me to the second point, namely that the underlying reason why some rats prefer the drug option can differ from the underlying reason why humans cannot control their drug use. As described more extensively below, the reason that only a small proportion of rats and humans get addicted is possibly different: steep delay discounting in rats could predispose more impulsive rats to be less likely to prefer the drug option, while in humans this relationship is reversed (Ahmed, 2018). In my view, it is especially unlikely that the delaying of a social reward in rodents taps into the same individual differences that govern whether a community reinforcement approach or contingency management works in hu-mans, as these approaches are unarguably more complex than just the choice between drugs or social contact.

4.3

Mechanistic validity

Mechanistic validity refers to the question whether the theoretical mechanisms that produce the observable effects of the disease in the animal model are the same as those in humans. The observable

effects here are for example the three DSM-like criteria of the 0/3 crit model, or choosing the drug over another behavioural choice in alternative-choice models.

Here, I will discuss the decision process that result in these ob-servable effects, with a focus on the difference in time scale between human- and rodent decision making. Where humans can take into account aspects not present in the immediate environment – such as the fact that long-term drug use will lead to harmful consequences - there is no evidence that rats possess this ability (Ahmed, 2018). Indeed, human capacity to adhere to long-term goals over alter-native short-term rewards is thought to be supported by specific cortical neurocircuitry (Heatherton & Wagner, 2011), which makes the mechanistic similarities between the short-term choices made by rats and those of human addicts questionable. In my opinion, this tension between long- and short term goals is one of the key as-pects of addiction, namely the inability to make decisions that lead to desirable outcomes in the long term but where a temptation on the short term has to be resisted. One way in how this difference be-tween humans and rats becomes apparent is by thinking about why certain rodents prefer saccharine water over drug choices (Ahmed, 2018). Rats are known to steeply discount rewards when they are delayed, meaning that small immediate rewards are quickly pre-ferred to larger future rewards, with a period of just 20-30 seconds leading rats to prefer a substantially smaller immediate reward over a bigger, but later available reward (Ahmed, 2017). As the rewarding effects of drugs on average take around fifteen seconds to have a first effect and a minute to reach their peak, a significant amount of delay discounting of the drugs is taking place during these trials – while the rewarding effects of saccharose water already takes place after approximately two seconds (Lenoir, Serre, Cantin, & Ahmed, 2007). This means there is an inherent lower value being placed on drug-rewards compared to food rewards if they are presented at the same time, as is done in many models discussed here (models which do introduce a longer delay show that delaying food pallets with two minutes results in oxycodone being preferred over the food choice) (Secci, Factor, Schindler, & Panlilio, 2016). It also means that the overall association between the operant choice and the reinforc-ing effects is inherently weaker in rats than in humans: where the operant choice in humans (i.e. intake of the drug) and the manifes-tation of its effects are very short on a human time scale, these tens of seconds to a minute are already long and affected by delay dis-counting in rodents (Ahmed, 2018). Furthermore, in models where the rewards are presented at the same time, this has the implication that the effect of impulsivity on addiction should essentially be reversed between humans and rodent models of addiction if the rewards are presented at the same time, as impulsivity involves preferring short-term rewards over long-term ones (Ahmed, 2018). Impulsive rodents have steeper delay discounting, and thus more impulsive rodents should value the late drug reward less than an immediately satisfying food reward and should have a tendency to prefer food over drugs if presented at the same time. For humans, this is precisely the other way around: having an impulsive nature is a risk factor for drug addiction (Ahmed, 2018). Here we see that a phenomenon that at first sight seems to be the same in rodents and humans, namely the fact that only a small part of the popu-lation exposed to drugs will prefer it above other choices, could have distinctly different underlying mechanisms leading to it. In

other words, face validity does not always constitute to mechanistic validity if closer inspected.

4.4

Pathogenic validity

Pathogenic validity refers to the question whether the way the species transforms from healthy to addicted resembles that in hu-mans.

In almost all recent models, this transformation from non-addicted rodent to models of addiction is achieved by escalating self-administration of the drugs over time (Ahmed, 2010). This escalating behaviour of self-administration is indeed a better analogy of human drug condi-tion than administracondi-tion of the drugs by the experimenter, outside of the animal’s control, although it is not shown to be possible for every drug of abuse: for example, rodents do not self-administer cannabis (Lefever, Marusich, Antonazzo, & Wiley, 2014).

However, it is still an oversimplified version of human circum-stance: many factors that play a role in the human road to addiction are not included in these rodent models. These include risk factors such as having certain personality traits, psychiatric disorders, or having challenging family- or environmental influences (Müller, 2018). The rat strains used are usually those of average phenotypes, instead of ones having these specific vulnerabilities. This calls into question whether the animal models capture the same pathogenic mechanisms that play a role in humans, or have different underly-ing reasons. This is similar to the point raised under mechanistic validityabout whether the mechanisms that underlie the choice for a drug reward are the same as in humans, which is in both cases crucial to know because having different underlying mechanisms means that interventions based on these mechanisms in rodents might not have the same effects in humans.

4.5

Predictive validity

Predictive validity refers to the question whether the relation be-tween therapeutic agents and the observable effects of the disease resembles that in humans, with a specific interest in if results ob-tained from pharmacological interventions in the animal model are predictive of clinical results in humans.

There are a few known examples of therapeutic agents against addictions to specific substances that were first discovered in ro-dents and later shown to be effective in humans as well (Pierce, O’Brien, Kenny, & Vanderschuren, 2012). First of all, there is acam-prosate, a GABA receptor agonist that was shown to reduce the voluntary intake of ethanol by rats (Boismare et al., 1984). After these rodent trials, it was shown to produce clinically relevant re-sults in humans regarding alcohol abstinence, and it is still being used in the treatment of addiction today (Mason & Heyser, 2010). Another success story for animal models of addiction is that of nal-trexone, an opioid receptor antagonist. This substance was shown to reduce alcohol preference in hamsters (Ross, Hartmann, & Geller, 1976) and, after trials in rhesus monkeys, human studies found clinically relevant effects too (Volpicelli, Alterman, Hayashida, & O’Brien, 1992). The last drug that shows clinically relevant efficacy in halting substance use and that was first researched in rodents and only later in humans, is varenicline. This partial nicotinic re-ceptor agonist was found to lessen the amount of nicotine taken during self-administration and to diminish nicotine seeking after

reinstatement (Coe et al., 2005), and was thereafter shown to be effective in humans too (Gonzales et al., 2006; Jorenby et al., 2006). These are the main therapeutic agents that were first developed and tested in animals and only later in humans (Field & Kersbergen, 2019), and thus the main examples of positive predictive validity of animal models of addictions.

On the other hand, there are a plethora of examples of drug trials that were found to be effective in rodent studies, but whose results did not translate to human clinical trials. For example, for decades research has been done using animal models to find an effective therapeutic agent to alleviate addictions of psychostimulants such as cocaine, but none did produce clinically relevant results in hu-mans (Pierce et al., 2012). This failure of translation is not only present for psychostimulants: rodent studies found strong effects of antagonists of the corticotropin-releasing factor receptor 1 (CRFR1) in reducing alcohol intake, alleviating withdrawal symptoms and re-ducing stress-induced reinstatement, and CRFR1-antagonists were thought to be a great potential pharmaceutical (Heilig et al., 2016). However, these results did not translate to recent human clinical trials (Kwako et al., 2015).

5

CONCLUSIONS AND CONSEQUENCES FOR

RESEARCH

Here, I have discussed five different aspects to the validity of rodent models and critically looked if the analogy with human addiction holds ground. Some of these criticisms are related to the specific models: the 0/3 crit model claims to model three symptoms of problematic drug use, but ultimately only shows a heightened moti-vation for the rodents in taking drugs – not problematic or addicted use. Choice models only present the rats with short-term and direct choices, while human drug users are having problems with long-term and abstract goals. Also, the mechanisms that underlie these short-term choices might be fundamentally different in humans and rodents due to, among other reasons, time scale differences in delay discounting. Furthermore, these models do not take into account factors that place individuals at risk of addiction, casting doubt on whether the rodents that adhere to these model’s criteria of addiction are being selected due to the same mechanisms as in humans.

Some of these shortcomings can be alleviated in future models, for example by using strains of rodents which show genetic or be-havioural characteristics that are known to put humans at-risk for developing addiction, which are indeed already included in some studies (Koob & Volkow, 2016). Likewise, the recent models dis-cussed above did attempt to include a broader range of factors such as social elements and behavioural choices, adequately addressing some previous shortcomings of rodent models. However, the fact that each individual model only models a few aspects of addic-tions remains a fundamental problem. The importance of this, and the complexity of substance use disorder, is highlighted by recent developments in the field of network theory in psychopathology, where mental disorders are seen as casually connected symptoms rather than being caused by one (but usually as of yet unidentified) single source (Borsboom, Cramer, & Kalis, 2019). This network view of psychopathological disorders gives an alternative to the usual reductionist explanations, and has been proposed for substance use

disorder too (Field, Heather, & Wiers, 2019; Rhemtulla et al., 2016). In this light, animal models that only model a certain symptom of addiction are problematic, as the interaction between the symptoms is missing here - which in the network view is fundamental to the development and manifestation of the disorder (Field & Kersbergen, 2019). To give an example, in the first network analysis of substance abuse and dependence symptoms, researchers found a strong pre-dictive relationship between more-than-planned substance use and tolerance (Rhemtulla et al., 2016), and this interaction cannot be investigated using animal models as some models do try to model tolerance, but none can infer whether animals use more drugs than they intended. In this framework, the validity of animal models for modelling addiction is limited.

Furthermore, most of the previously mentioned criticisms are not specific to these models, but touch upon inherent limitations of the rodent as a model of human addiction. These inherent limitations of rodent models include the inability of rodents to take long-term goals into account, while this is an essential part of loss of control addicted people experience: we often conceptualize self-control as the ability to resist temptations in order to achieve abstract or long-term goals, such as having a successful career, being a good person, maintaining your health, or keeping up social relations with people we care about. Addiction is characterized by a loss of this self-control, where these long-term and abstract goals are not weighed for short-term decisions, and this is precisely what cannot be measured in rodents. Another limitation is the inability to measure intent of rodents, which is a key mechanism in addiction: using drugs against formerly formulated intentions not to. Together, these limitations lead me to the conclusion that rodent models of addiction do not model human addiction accurately, as many of the higher order cognitive processes playing key roles in the psychopathology of addiction are not present. What do these animal models measure if not addiction? In my view, these models show heightened motivation of some rats to use drugs: undoubtedly, there is a propensity of rodents to self-administer them, a preference of some to prefer the drug to alternative choices, a willingness to perform operant tasks to obtain them, and a perseverance to keep using the drug even though it is accompanied by foot shocks. However, the fact that there is some heightened propensity to use drugs does not constitute to addiction in humans, and should not be enough to claim to model addiction in rodents.

What do these shortcomings of rodent models mean for the way in which addiction is conceptualized? The brain disease model of addiction is in large part built upon research performed in rodents due to the option to use experimental procedures and manipulations which are not possible with human subjects. This is a fundamentally reductionist mission given that these rodents at best model a few symptoms and omit key higher-order cognitive factors such as intent, self-control and long-term goals. This is in line with the overall reductionism of the brain disease model of addiction with the goal to find neuroscientific root causes of addiction. Given the shortcomings in predictive validity and the difference in (cognitive) mechanisms underlying drug-related behaviour between humans and rodents, it is perhaps time to reconsider the interpretations of these rodent studies as strong evidence for the brain disease model of addiction, thereby leaving room for other views - such as the previously mentioned network theory of addiction.

Given the questionable predictive validity of animal models of addiction, some have argued that funding that goes into research relating to the brain disease model of addiction might be better spent elsewhere (Hall et al., 2015). Indeed, the pharmaceutical industry has also largely turned away from investing in the use of animal models (Hyman, 2016). Does this mean that the rodent models discussed here are ultimately useless and should be abandoned? That might be too simplistic of a conclusion. As discussed before, in my view these rodent models do not model addiction, but rather a heightened motivation to take a drug, and some factors in this might be relevant to human addiction too. However, if these animal models are to be used, their limitations and what they do and do not model should clearly be taken into account.

Even though recent improvements in rodent models of addic-tion have addressed important shortcomings, the fundamental dif-ferences between humans and rodents and the uniquely human elements of addiction seem to stand in the way of understanding many of its underlying mechanisms and subsequent treatments through the use of rodent models.

6

REFERENCES

Ahmed, S. H. (2010). Validation crisis in animal models of drug addiction: Beyond non-disordered drug use toward drug addiction. Neuroscience & Biobehavioral Reviews, 35(2), 172–184. https://doi.org/10.1016/j.neubiorev. 2010.04.005

Ahmed, S. H. (2017). Individual decision-making in the causal pathway to addiction: Contributions and limitations of rodent models. Pharmacology Biochemistry and Behavior, 164, 22–31. https://doi.org/10.1016/j.pbb.2017. 07.005

Ahmed, S. H. (2018). Trying to make sense of rodents’ drug choice be-havior. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 87, 3–10. https://doi.org/10.1016/j.pnpbp.2017.09.027

Alexander, B. K., Beyerstein, B. L., Hadaway, P. F., & Coambs, R. B. (1981). Effect of early and later colony housing on oral ingestion of morphine in rats. Pharmacology Biochemistry and Behavior, 15(4), 571–576. https: //doi.org/10.1016/0091-3057(81)90211-2

Alexander, B. K., Coambs, R. B., & Hadaway, P. F. (1978). The effect of housing and gender on morphine self-administration in rats. Psychophar-macology, 58(2), 175–179. https://doi.org/10.1007/BF00426903

American Psychiatric Association. (2013). Diagnostic and Statistical Man-ual of Mental Disorders (DSM-5®). American Psychiatric Pub.

Anthony, J. C., Warner, L. A., & Kessler, R. C. (1997). Comparative epi-demiology of dependence on tobacco, alcohol, controlled substances, and inhalants: Basic findings from the National Comorbidity Survey. https: //doi.org/10.1037/10248-001

Augier, E., Barbier, E., Dulman, R. S., Licheri, V., Augier, G., Domi, E., ... Heilig, M. (2018). A molecular mechanism for choosing alcohol over an alternative reward. Science, 360(6395), 1321–1326. https://doi.org/10.1126/ science.aao1157

Belzung, C., & Lemoine, M. (2011). Criteria of validity for animal models of psychiatric disorders: Focus on anxiety disorders and depression. Biology of Mood & Anxiety Disorders, 1, 9. https://doi.org/10.1186/2045-5380-1-9

Boismare, F., Daoust, M., Moore, N., Saligaut, C., Lhuintre, J. P., Chretien, P., & Durlach, J. (1984). A homotaurine derivative reduces the voluntary intake of ethanol by rats: Are cerebral GABA receptors involved? Pharma-cology Biochemistry and Behavior, 21(5), 787–789. https://doi.org/10.1016/ S0091-3057(84)80020-9

Borsboom, D., Cramer, A. O. J., & Kalis, A. (2019). Brain disorders? Not really: Why network structures block reductionism in psychopathol-ogy research. Behavioral and Brain Sciences, 42. https://doi.org/10.1017/ S0140525X17002266

Cantin, L., Lenoir, M., Augier, E., Vanhille, N., Dubreucq, S., Serre, F., ... Ahmed, S. H. (2010). Cocaine Is Low on the Value Ladder of Rats: Possible Evidence for Resilience to Addiction. PLOS ONE, 5(7), e11592. https://doi. org/10.1371/journal.pone.0011592

Coe, J. W., Brooks, P. R., Vetelino, M. G., Wirtz, M. C., Arnold, E. P., Huang, J., ... O’Neill, B. T. (2005). Varenicline: An 42 Nicotinic Receptor Partial Agonist for Smoking Cessation. Journal of Medicinal Chemistry, 48(10), 3474–3477. https://doi.org/10.1021/jm050069n

de Wit, H., Epstein, D., & Preston, K. (2018). Does human language limit translatability of clinical and preclinical addiction research? Neuropsy-chopharmacology, 43. https://doi.org/10.1038/s41386-018-0095-8

Debiec, J., & Olsson, A. (2017). Social Fear Learning: From Animal Models to Human Function. Trends in Cognitive Sciences, 21(7), 546–555. https: //doi.org/10.1016/j.tics.2017.04.010

Deroche-Gamonet, V., Belin, D., & Piazza, P. V. (2004). Evidence for Addiction-like Behavior in the Rat. Science, 305(5686), 1014–1017. https: //doi.org/10.1126/science.1099020

Ellenbroek, B., & Youn, J. (2016). Rodent models in neuroscience research: Is it a rat race? Disease Models & Mechanisms, 9(10), 1079–1087. https: //doi.org/10.1242/dmm.026120

Field, M., Heather, N., & Wiers, R. W. (2019). Indeed, not really a brain disorder: Implications for reductionist accounts of addiction. Behavioral and Brain Sciences, 42. https://doi.org/10.1017/S0140525X18001024

Field, M., & Kersbergen, I. (2019). Are animal models of addiction useful? Addiction, 0(0). https://doi.org/10.1111/add.14764

Gage, S. H., & Sumnall, H. R. (2019). Rat Park: How a rat paradise changed the narrative of addiction. Addiction, 114(5), 917–922. https://doi.org/10. 1111/add.14481

Galef, B. G. (2002). Social Learning of Food Preferences in Rodents: Rapid Appetitive Learning. Current Protocols in Neuroscience, 21(1), 8.5D.1-8.5D.8. https://doi.org/10.1002/0471142301.ns0805ds21

Gilpin, N. W., Richardson, H. N., Cole, M., & Koob, G. F. (2008). Vapor Inhalation of Alcohol in Rats. Current Protocols in Neuroscience, 44(1), 9.29.1-9.29.19. https://doi.org/10.1002/0471142301.ns0929s44

Gonzales, D., Rennard, S. I., Nides, M., Oncken, C., Azoulay, S., Billing, C. B., ... Group, for the V. P. 3 S. (2006). Varenicline, an 42 Nicotinic Acetyl-choline Receptor Partial Agonist, vs Sustained-Release Bupropion and Placebo for Smoking Cessation: A Randomized Controlled Trial. JAMA, 296(1), 47–55. https://doi.org/10.1001/jama.296.1.47

Hadaway, P. F., Alexander, B. K., Coambs, R. B., & Beyerstein, B. (1979). The effect of housing and gender on preference for morphine-sucrose so-lutions in rats. Psychopharmacology, 66(1), 87–91. https://doi.org/10.1007/ BF00431995

Hall, W., Carter, A., & Forlini, C. (2015). The brain disease model of ad-diction: Is it supported by the evidence and has it delivered on its promises? The Lancet Psychiatry, 2(1), 105–110. https://doi.org/10.1016/S2215-0366(14) 00126-6

Haller, M. M., & Chassin, L. (2010). The Reciprocal Influences of Perceived Risk for Alcoholism and Alcohol Use Over Time: Evidence for Aversive Transmission of Parental Alcoholism. Journal of Studies on Alcohol and Drugs, 71(4), 588–596. https://doi.org/10.15288/jsad.2010.71.588

Heather, N. (2017). Q: Is Addiction a Brain Disease or a Moral Failing? A: Neither. Neuroethics, 10(1), 115–124. https://doi.org/10.1007/s12152-016-9289-0

Heatherton, T. F., & Wagner, D. D. (2011). Cognitive neuroscience of self-regulation failure. Trends in Cognitive Sciences, 15(3), 132–139. https: //doi.org/10.1016/j.tics.2010.12.005

Heilig, M., Epstein, D. H., Nader, M. A., & Shaham, Y. (2016). Time to con-nect: Bringing social context into addiction neuroscience. Nature Reviews Neuroscience, 17(9), 592–599. https://doi.org/10.1038/nrn.2016.67

Hyman, S. E. (2016). Back to basics: Luring industry back into neuro-science. Nature Neuroscience, 19(11), 1383–1384. https://doi.org/10.1038/nn. 4429

Jorenby, D. E., Hays, J. T., Rigotti, N. A., Azoulay, S., Watsky, E. J., Williams, K. E., ... Group, for the V. P. 3 S. (2006). Efficacy of Varenicline, an 42 Nicotinic Acetylcholine Receptor Partial Agonist, vs Placebo or Sustained-Release Bupropion for Smoking Cessation: A Randomized Controlled Trial. JAMA, 296(1), 56–63. https://doi.org/10.1001/jama.296.1.56

Koob, G. F., & Volkow, N. D. (2016). Neurobiology of addiction: A neuro-circuitry analysis. The Lancet Psychiatry, 3(8), 760–773. https://doi.org/10. 1016/S2215-0366(16)00104-8

Kwako, L. E., Spagnolo, P. A., Schwandt, M. L., Thorsell, A., George, D. T., Momenan, R., ... Heilig, M. (2015). The Corticotropin Releasing Hormone-1 (CRH1) Receptor Antagonist Pexacerfont in Alcohol Dependence: A Ran-domized Controlled Experimental Medicine Study. Neuropsychopharmacol-ogy, 40(5), 1053–1063. https://doi.org/10.1038/npp.2014.306

Lefever, T. W., Marusich, J. A., Antonazzo, K. R., & Wiley, J. L. (2014). Eval-uation of WIN 55,212-2 self-administration in rats as a potential cannabinoid abuse liability model. Pharmacology Biochemistry and Behavior, 118, 30–35. https://doi.org/10.1016/j.pbb.2014.01.002

Lenoir, M., Serre, F., Cantin, L., & Ahmed, S. H. (2007). Intense Sweetness Surpasses Cocaine Reward. PLoS ONE, 2(8). https://doi.org/10.1371/journal. pone.0000698

Lipari, R. N., Hedden, S. L., & Hughes, A. (2013). Substance Use and Mental Health Estimates from the 2013 National Survey on Drug Use and Health: Overview of Findings. In The CBHSQ Report. Retrieved from http: //www.ncbi.nlm.nih.gov/books/NBK385055/

Mason, B. J., & Heyser, C. J. (2010). Acamprosate: A Prototypic Neuromod-ulator in the Treatment of Alcohol Dependence [Text]. https://doi.org/info: doi/10.2174/187152710790966641

McConnell, D. (2016). Narrative Self-Constitution and Recovery From Addiction. American Philosophical Quarterly, 53(3), 307–322. McConnell, D., & Snoek, A. (2018). The Importance of Self-Narration in Recovery from Addiction. Philosophy, Psychiatry, & Psychology, 25(3), E-31-E-44. https: //doi.org/10.1353/ppp.2018.0022

Müller, C. P. (2018). Animal models of psychoactive drug use and ad-diction – Present problems and future needs for translational approaches. Behavioural Brain Research, 352, 109–115. https://doi.org/10.1016/j.bbr.2017. 06.028

Orford, J., Velleman, R., Natera, G., Templeton, L., & Copello, A. (2013). Addiction in the family is a major but neglected contributor to the global burden of adult ill-health. Social Science & Medicine, 78, 70–77. https://doi. org/10.1016/j.socscimed.2012.11.036

Parker, C. C., Chen, H., Flagel, S. B., Geurts, A. M., Richards, J. B., Robin-son, T. E., ... Palmer, A. A. (2014). Rats are the smart choice: Rationale for a renewed focus on rats in behavioral genetics. Neuropharmacology, 76, 250–258. https://doi.org/10.1016/j.neuropharm.2013.05.047

Perry, A. N., Westenbroek, C., & Becker, J. B. (2016). Chapter 6—Sex Differ-ences and Addiction. In R. M. Shansky (Ed.), Sex DifferDiffer-ences in the Central Nervous System (pp. 129–147). https://doi.org/10.1016/B978-0-12-802114-9. 00006-8

Pierce, R. C., O’Brien, C. P., Kenny, P. J., & Vanderschuren, L. J. M. J. (2012). Rational Development of Addiction Pharmacotherapies: Successes, Failures, and Prospects. Cold Spring Harbor Perspectives in Medicine, 2(6), a012880. https://doi.org/10.1101/cshperspect.a012880

Ping, A., & Kruzich, P. J. (2008). Concurrent access to sucrose pellets decreases methamphetamine-seeking behavior in Lewis rats. Pharmacology Biochemistry and Behavior, 90(3), 492–496. https://doi.org/10.1016/j.pbb. 2008.04.009

Rhemtulla, M., Fried, E. I., Aggen, S. H., Tuerlinckx, F., Kendler, K. S., & Borsboom, D. (2016). Network analysis of substance abuse and dependence symptoms. Drug and Alcohol Dependence, 161, 230–237. https://doi.org/10. 1016/j.drugalcdep.2016.02.005

Ross, D., Hartmann, R. J., & Geller, I. (1976). Ethanol preference in the hamster: Effects of morphine sulfate and naltrexone, a long-acting morphine antagonist. Proceedings of the Western Pharmacology Society, 19, 326–330. Sanchis-Segura, C., & Spanagel, R. (2006). REVIEW: Behavioural assess-ment of drug reinforceassess-ment and addictive features in rodents: an overview. Addiction Biology, 11(1), 2–38. https://doi.org/10.1111/j.1369-1600.2006.00012. x

Secci, M. E., Factor, J. A., Schindler, C. W., & Panlilio, L. V. (2016). Choice between delayed food and immediate oxycodone in rats. Psychopharmacol-ogy, 233(23), 3977–3989. https://doi.org/10.1007/s00213-016-4429-0

Smith, M. A. (2012). Peer influences on drug self-administration: Social facilitation and social inhibition of cocaine intake in male rats. Psychophar-macology, 224(1), 81–90. https://doi.org/10.1007/s00213-012-2737-6

Spanagel, R. (2017). Animal models of addiction. Dialogues in Clinical Neuroscience, 19(3), 247–258.

Stein, J. A., Newcomb, M. D., & Bentler, P. M. (1987). An 8-year study of multiple influences on drug use and drug use consequences. Journal of Personality and Social Psychology, 53(6), 1094–1105. https://doi.org/10.1037/ 0022-3514.53.6.1094

Stitzer, M. L., Jones, H. E., Tuten, M., & Wong, C. J. (2011). Commu-nity Reinforcement Approach and Contingency Management Interventions for Substance Abuse. Handbook of Motivational Counseling: Goal-Based Approaches to Assessment and Intervention with Addiction and Other Problems, 549–569. https://doi.org/10.1002/9780470979952.ch23

Treit, D., Engin, E., & McEown, K. (2010). Animal Models of Anxiety and Anxiolytic Drug Action. In M. B. Stein & T. Steckler (Eds.), Behavioral Neurobiology of Anxiety and Its Treatment (pp. 121–160). https://doi.org/ 10.1007/7854_2009_17

Vanderschuren, L. J. M. J., & Ahmed, S. H. (2013). Animal Studies of Addictive Behavior. Cold Spring Harbor Perspectives in Medicine, 3(4), a011932. https://doi.org/10.1101/cshperspect.a011932

Vanderschuren, L. J. M. J., & Everitt, B. J. (2004). Drug Seeking Be-comes Compulsive After Prolonged Cocaine Self-Administration. Science, 305(5686), 1017–1019. https://doi.org/10.1126/science.1098975

Vanderschuren, L. J., Minnaard, A. M., Smeets, J. A., & Lesscher, H. M. (2017). Punishment models of addictive behavior. Current Opinion in Be-havioral Sciences, 13, 77–84. https://doi.org/10.1016/j.cobeha.2016.10.007

Vendruscolo, L. F., & Roberts, A. J. (2014). Operant alcohol self-administration in dependent rats: Focus on the vapor model. Alcohol, 48(3), 277–286. https://doi.org/10.1016/j.alcohol.2013.08.006

Venniro, M., Caprioli, D., & Shaham, Y. (2016). Chapter 2 - Animal models of drug relapse and craving: From drug priming-induced reinstatement to incubation of craving after voluntary abstinence. In H. Ekhtiari & M. P. Paulus (Eds.), Progress in Brain Research (pp. 25–52). https://doi.org/10. 1016/bs.pbr.2015.08.004

Venniro, M., Zhang, M., Caprioli, D., Hoots, J. K., Golden, S. A., Heins, C., ... Shaham, Y. (2018). Volitional social interaction prevents drug addiction in rat models. Nature Neuroscience, 21(11), 1520–1529. https://doi.org/10. 1038/s41593-018-0246-6

Volkow, N. D., Koob, G. F., & McLellan, A. T. (2016). Neurobiologic Ad-vances from the Brain Disease Model of Addiction. New England Journal of Medicine, 374(4), 363–371. https://doi.org/10.1056/NEJMra1511480

Volpicelli, J. R., Alterman, A. I., Hayashida, M., & O’Brien, C. P. (1992). Nal-trexone in the Treatment of Alcohol Dependence. Archives of General Psy-chiatry, 49(11), 876–880. https://doi.org/10.1001/archpsyc.1992.01820110040006

Willner, P. (1984). The validity of animal models of depression. Psy-chopharmacology, 83(1), 1–16. https://doi.org/10.1007/BF00427414