Translational approach to understanding momentary factors associated with alcohol consumption

Abstract

Multiple interindividual and intra‐individual factors underlie variability in drinking motives, challenging clinical translatability of animal research and limiting treatment success of substance use‐related problems. Intra‐individual variability refers to time‐dependent continuous and discrete changes within the individual and in substance use research is studied as momentary variation in the internal states (craving, stressed, anxious, impulsive and tired) and response to external triggers (stressors, drug‐associated environmental cues and social encounters). These momentary stimuli have a direct impact on behavioural decisions and may be triggers and predictors of substance consumption. They also present potential targets for real‐time behavioural and pharmacological interventions. In this review, we provide an overview of the studies demonstrating different momentary risk factors associated with increased probability of alcohol drinking in humans and changes in alcohol seeking and consumption in animals. The review also provides an overview of pharmacological interventions related to every individual risk factor.

Abbreviations

AA

ambulatory assessment

AUD

alcohol use disorder

CRF₁

corticotropin‐releasing hormone receptor 1

DD

daily diaries

DSM

Diagnostic and Statistical Manual of Mental Disorders

EMA

ecological momentary assessment

EMI

ecological momentary intervention

ICD

International Classification of Diseases

ILD

intensive longitudinal data

JITAI

just‐in‐time adaptive intervention

OX₁

orexin receptor 1

V_1B

vasopressin V1b receptor

1. INTRODUCTION

The probability of initiating, maintaining and losing control over substance use depends on the combined impact of multiple genetic/epigenetic factors and long‐lasting effects of environmental exposures. These factors define personality traits and other characteristics of an individual, such as cognitive abilities, reward/stress sensitivity and social skills. Chronic substance use alters personality and induces multiple impairments in brain function, which further impacts individual characteristics of a substance user (e.g. Robinson & Berridge, 2003). Finally, both preclinical and clinical research have recognized the role of momentary triggers, such as drug craving and stressful events, in leading to substance use (e.g. Morgenstern, Kuerbis, & Muench, 2014). New technologies offer various approaches to advance this research by collecting real‐time data on a variety of momentary environmental experiences and internal states that precede substance intake in real‐world settings.

In the context of developing precision medicine, it is not enough just to study interindividual differences but also intra‐individual factors and momentary stimuli associated with increased probability of substance use. For instance, both genetically defined stress sensitivity (Kudielka, Hellhammer, & Wüst, 2009; Meaney et al., 1991) and learned stress‐coping strategies (Corbin, Farmer, & Nolen‐Hoekesma, 2013) impact whether the exposure to a stressor leads to self‐medication through substance use. However, stressors differ in their intensity, controllability, predictability, magnitudes of stress responses caused and so forth, which have different behavioural consequences (Keyes, Hatzenbuehler, Grant, & Hasin, 2012). Traditionally, preclinical and clinical studies have examined group effects, or effects aggregated over time, leaving the processes occurring at the individual level and dynamically over time largely unexamined (Nelson, McGorry, Wichers, Wigman, & Hartmann, 2017). Intra‐individual variability is understood as time‐dependent continuous and discrete changes within the individual (Collins, 2006), and in substance use research is studied as momentary internal states (e.g. increased momentary impulsiveness, poor sleep quality, and experience of craving episodes) and response to external triggers (e.g. stressors and drug‐associated environmental and social cues). Momentary stimuli can directly impact behavioural decisions and therefore have the potential to predict substance consumption or re‐initiation after prolonged abstinence (Morgenstern et al., 2014).

In humans, researchers are increasingly employing the ambulatory assessment (AA) approach (for a review, see Trull & Ebner‐Priemer, 2013), which consists of using real‐time capture tools to assess ongoing human behaviour, physiology, experiences, and environmental factors in natural settings. Ecological momentary assessment (EMA), a type of ambulatory assessment, involves repeated sampling of a person's momentary behaviours (e.g. socializing) and related psychological parameters (e.g. excitement) in everyday life. Now often conducted on mobile devices with brief questionnaires, ecological momentary assessment enables us to understand the time course of momentary experiences and related factors together in naturalistic and unconstrained environments (Shiffman, Stone, & Hufford, 2008; Stone et al., 1998). For instance, ecological momentary assessment can examine numerous aspects of alcohol‐drinking behaviours, such as drinking over time (i.e. how often and what time), situational/contextual predictors and antecedents to drinking, and consequences of drinking episodes (Beckjord & Shiffman, 2014). To evaluate dynamic behavioural changes and other parameters at the intra‐individual level, ecological momentary assessment data can then be combined with biomedical intensive longitudinal data (ILD), such as passive sensing data (e.g. locomotor activity and transdermal alcohol sensors), information about environmental contexts (e.g. GPS location and companions) and clinical records (e.g. electronic health records; Kim, Marcusson‐Clavertz, Yoshiuchi, & Smyth, 2019). Analysis of high‐resolution intensive longitudinal data can identify behavioural and physiological biomarkers of specific psychopathologies (Nakamura, Kiyono, Wendt, Abry, & Yamamoto, 2016). These approaches are expected to have great utility when applied to the study of substance use‐related problems.

In preclinical research, the exposure of an animal to specific environmental triggers (e.g. stressors and drug‐associated cues) is controlled by an experimenter, and subsequent response is often monitored in a single test (e.g. reinstatement of drug seeking; Lê & Shaham, 2002; Figure 1a). Animal studies often benefit from a longitudinal design and a common approach is to measure long‐term consequences of environmental manipulation and time‐dependent changes in behaviour (e.g. Wolffgramm & Heyne, 1995). Intensive longitudinal data designs are presently underused but are expected to be informative in characterizing specific behavioural and physiological changes.

FIGURE 1.

Methodological approaches for momentary data collection. (a) Preclinical research; (b) human studies. In preclinical studies (a), exposure of animals to alcohol and specific environmental triggers (e.g. cues and stressors) are controlled by an experimenter, and response to these triggers is monitored in a single test and using either a longitudinal or intensive longitudinal data (ILD) design. ILD acquisition is also needed in studies that monitor behaviours controlled by internal states of an animal (e.g. circadian cycle‐dependent behaviours). Clinical studies (b) are focused on the significance of interindividual and intra‐individual variability with respect to substance consumption and monitor drinking behaviours by integrating ILD (e.g. physical activity) and ambulatory assessment methods in natural settings that can then be combined with biomedical ILD, such as passive sensing data (e.g. locomotor activity and transdermal alcohol sensors), information about environmental contexts (e.g. GPS location and companions), and clinical records (e.g. electronic health records and clinical visits). Preclinical research can contribute to refining development of novel strategies for individual therapeutic approaches (e.g. just‐in‐time adaptive intervention [JITAI] and ecological momentary intervention [EMI])

Medications currently used to treat alcohol use‐related problems are aimed at preventing heavy drinking, assisting with withdrawal symptoms and reducing probability of relapse. The aldehyde dehydrogenase inhibitor disulfiram, the opioid receptor antagonist naltrexone and acamprosate (calcium acetylhomotaurinate, a compound that does not have clearly defined mechanism of action) have been used for many years to manage alcohol misuse (Witkiewitz, Litten, & Leggio, 2019). Another μ‐opioid receptor antagonist and a partial κ‐opioid receptor agonist, nalmefene, have recently been approved for reduction of alcohol consumption. However, the effectiveness of these medications in reducing the probability of alcohol consumption remains limited or beneficial only in a subset of the population due to issues such as compliance, side effects, and lack of a personalized medicine approach (Walker & Lawrence, 2018). The identification of momentary state‐dependent factors and high‐risk situations associated with increased probability of drug use could suggest novel strategies for the development of individual therapeutic approaches. Insights derived from ambulatory assessment studies may be able to suggest targets for ecological momentary intervention (EMI; Heron & Smyth, 2010) and just‐in‐time adaptive intervention (JITAI; Spruijt‐Metz et al., 2015; Figure 1b). Ecological momentary intervention is the delivery of real‐time interventions to people in daily life especially when they are needed. Just‐in‐time adaptive intervention involves the setting of just‐in‐time interventions successively tailored to a prior stage of the intervention based on previous responses (obtained by ecological momentary assessment). Ecological momentary assessment‐based clinical studies can be used to determine treatment‐induced changes in drinking dynamics at both the interindividual and intra‐individual levels. Preclinical research can contribute to refining ecological momentary intervention and just‐in‐time adaptive intervention by identifying novel and more effective pharmacological treatments (e.g. Witkiewitz et al., 2019) and optimizing current ones (e.g. Foo et al., 2019).

This review is an overview of studies demonstrating different momentary risk factors associated with increased probability of drinking in humans and changes in alcohol‐seeking and consumption in animals. Translation of preclinical experimental findings into successful clinical applications is a primary goal and has always been a major concern in the drug discovery field. Hence, our second aim is to review pharmacological intervention strategies that have been suggested, in the context of momentary triggers, by both preclinical research and using the ecological momentary assessment approach. The research domain criteria (RDoC; https://www.nimh.nih.gov/research/research-funded-by-nimh/rdoc/rdoc-updates.shtml) system will be used to maintain a translational perspective (Figure 2). The research domain criteria project was developed to reorient psychiatric research towards measurable behavioural dimensions and their underlying mechanisms, thereby increasing translational likelihood of preclinical findings (Cuthbert & Insel, 2013). At present, six domains of the research domain criteria matrix have been established, representing different aspects of emotional, cognitive, motivational, social and other functions: (i) negative valence, (ii) positive valence, (iii) cognitive systems, (iv) social processes, (v) arousal/regulatory processes and (vi) sensorimotor systems. Evidence showing the impact of sensorimotor systems on alcohol consumption is scarce, and they will be omitted from this review.

FIGURE 2.

Translational approach to understanding momentary triggers associated with alcohol consumption. (a) Preclinical research; (b) translation; (c) human studies. In preclinical studies (a), intra‐individual differences or effect of specific environmental triggers (e.g. cues and stressors) are often reported as a result of a single test (e.g. stress‐induced reinstatement of alcohol seeking) and once‐a‐day/averaged recordings of voluntary alcohol consumption changes in response to, for example, stress exposure. Behavioural patterns and states are identified to quantify the effect of environmental changes on animal drinking behaviour and locomotor activity using intensive longitudinal data (ILD) recordings. The basis of clinical translatability (b) is in finding common behavioural and physiological responses to challenging events between different species that associate with increased probability of drinking in humans and changes in alcohol‐seeking and consumption in animals. Subdivision of different physiological and behavioural functions into research domains might help with this task and lead to development of novel strategies for individual therapeutic approaches. Human studies (c) monitor drinking behaviours by integrating ILD (e.g. locomotor activity) and ambulatory assessment methods (e.g. ecological momentary assessment [EMA]) in natural settings. (b) Feeling stressed and state anxiety (negative valence), exposure to alcohol associated cues and elevated craving (positive valence), the impulse to act rashly and without thinking (cognitive systems), socializing and social rejection (social functions), poor sleep quality and high arousal in younger people and boredom in older people (sleep and regulatory systems) have been associated with increased probability of alcohol consumption in both animal and human studies

2. METHODOLOGICAL APPROACHES FOR MOMENTARY DATA COLLECTION

The translational value of preclinical research is based on similarities between different animal species. Non‐human primates are undoubtedly the best model organism to study human behaviour, but due to ethical considerations and practical reasons behavioural research nowadays largely uses rodents, in particular, mice and rats. The precept of ambulatory assessment is studying humans in “real‐time” in the natural environment. Animal field research involving observation of alcohol consumption by wild animals in their own habitat is possible (Wiens et al., 2008) but difficult to achieve and therefore seldom performed. Instead, animals are exposed to specific, relevant to human life, environmental challenges in the laboratory setting and behavioural, physiological, molecular, and so forth responses are studied. Differing experimental set‐ups and the evolutionary remoteness of the species should be considered when relating animal findings to human behaviour.

2.1. Animal studies

Monitoring time‐dependent substance‐related behavioural changes is based on paradigms giving animals long‐term free access to a substance. This can be achieved either by introducing a substance containing drinking solution(s) into the home cage of an animal (e.g. Spanagel & Hölter, 1999; Wolffgramm & Heyne, 1995) or by exposing animals to operant substance self‐administration paradigms incorporating instrumental and Pavlovian behavioural mechanisms (e.g. Bossert, Marchant, Calu, & Shaham, 2013; de Guglielmo, Kallupi, Cole, & George, 2017; Domi, Stopponi, Domi, Ciccocioppo, & Cannella, 2019; Figure 1a). In the case of alcohol, morphological and functional changes in the animal brain can be induced prior to behavioural testing by subjecting them to a chronic intermittent alcohol vapour exposure procedure (Vendruscolo & Roberts, 2014). Several other tests/paradigms are used whereby response to environmental exposures (e.g. stress and drug‐associated cues) can be measured (for a review, see Sanchis‐Segura & Spanagel, 2006). The most frequently used paradigm in the drug‐discovery field is reinstatement of substance seeking (Lê & Shaham, 2002). To measure seeking, an animal needs to establish stable operant alcohol self‐administration, followed by a drug‐free period (i.e. operant extinction), after which the resumption of the extinguished behaviour, in response to either stress, drug priming, conditioned cue or context, is tested. During this test, alcohol is not available (Figure 2a).

As animals cannot self‐report, quantitative analysis of behavioural patterns is required to study intra‐individual variability; the once‐a‐day/averaged recordings of substance consumption used so far often cannot provide the resolution for sufficiently informative analysis. In alcohol research, using intensive longitudinal data collected by automated instruments (Figure 1a), behavioural patterns and states can be identified to quantify the effect of environmental changes on drinking (Figure 2a). For example, locomotor activity, the timing and the frequency of approaches to an alcohol bottle, the number of licks or the amount of alcohol consumed per drinking approach can be recorded for an extended period of time (e.g. Foo et al., 2017; Vengeliene, Noori, & Spanagel, 2013). Recent advances in technology leading to the development of improved sensors, implemented in an automated “smart home cage” recording system, are beginning to make it possible to acquire multimodal intensive longitudinal data at high resolution in group‐housed rodents. This allows assessment of social interactions, which has advantages over single housing typically used in alcohol research. For example, in a home cage environment, continuous real‐time positional information can be recorded. Rodents are subcutaneously chipped with radio frequency ID (RFID) tags and their positions tracked and read by antennae beneath the cage. Furthermore, behavioural data can be collected using both video and IR cameras, and computer learning algorithms can be applied to automatically detect behaviour. Objective physiological measures by use of in vivo telemetry (e.g. heart rate, blood pressure and body temperature; Huetteman & Bogie, 2009) and brain electrophysiology can be complementary to behavioural studies.

2.2. Human studies

Various methods are used to monitor dynamic internal states and environmental contexts that vary differently across time. When clinicians and researchers use retrospective assessments such as trait questionnaires for alcohol use with related internal states and contexts, they assume that the assessment captures what participants believe or remember in a cross‐sectional assessment (Figure 1b). There are several frequently used questionnaires related to alcohol use (for a review, see Kaplan et al., 2018). Cross‐sectional questionnaires, although informative, do not necessarily comprehensively evaluate an (intra‐)individual's daily experiences as they are meant to be a single assessment for typical behaviours and related experiences in the past (e.g. last weeks, months or years).

A well‐designed ecological momentary assessment sampling scheme can precisely provide complete coverage of behaviours and experiences in an individual's everyday life (Shiffman et al., 2008). Signal‐contingent ecological momentary assessment characterizes continuous experience, for example, mood variations over time, while event‐contingent (i.e. user‐initiated) ecological momentary assessment monitors particular discrete events/episodes in subjects' lives, for example, smoking/drinking episodes and focuses on data collection around these events. Combining these strategies enables comprehensive analysis and the drawing of inferences about health‐related variables (Figure 2c). For example, internal states with environmental contexts can be collected multiple times per day using signal‐contingent designs, event‐contingent probes about target substance use and retrospective daily diaries (DD) that look back on that day overall. These designs allow exploration of how the variations of an individual's internal states and environmental contexts are related to target behaviours across the momentary (within‐day) and/or daily (across‐day) time scales.

Novel technologies enable scientists to collect data not only in larger amounts and at higher resolution but also more objectively. For instance, wearable devices (e.g. transdermal alcohol sensors) have recently become available that are capable of continuous monitoring of drug content in the body (Piasecki, 2019). Hence, detecting changes in, for example, cardiovascular parameters, electrodermal activity, wrist accelerometery, and skin temperature before and during self‐reported alcohol intake (or transdermally recorded intake) make it possible to evaluate not only internal states and substance use predictors but also individual differences in the pharmacological and behavioural effects of alcohol.

The ambulatory assessment approach provides novel developments in the modelling of health‐related data. Studies using ambulatory assessment have examined how a target phenomenon (e.g. drinking behaviour) co‐varies with related variables that often vary across different levels (i.e. data resolution), including moments (e.g. craving), days (e.g. weekday or weekend), persons (e.g. male and female) or other levels (e.g. continuously measured physiological factors). Sophisticated models based on such integrated behavioural, physical, social, biological and environmental monitoring, together with ecological momentary assessment self‐reports, may produce more opportunities to quantify target behaviours. To effectively combine multiple data streams researchers will need to draw from disciplines dealing extensively with intensive longitudinal data, for example, signal processing, hierarchical and mixed effects modelling, and dynamic systems theory. Such data‐driven model‐based techniques are expected to lead to novel treatments/interventions that are personalized, contextualized, and delivered when and where they are most needed in daily life (Spruijt‐Metz et al., 2015).

3. NEGATIVE VALENCE

Negative valence refers to states (e.g. anxiety and fear) and behavioural responses (e.g. avoidance and aggression) triggered by aversive situations or contexts, such as traumatic or stressful experiences and threat anticipation or exposure. Stressful and traumatic life experiences are considered key environmental contributors to trigger the onset and recurrence of many known psychiatric conditions listed in the Diagnostic and Statistical Manual of Mental Disorders (DSM) and International Classification of Diseases (ICD), including substance use disorder (SUD; e.g. María‐Ríos & Morrow, 2020). Stressors can facilitate drug consumption in humans by increasing activity of the central stress system (the hypothalamic–pituitary–adrenal axis) that subsequently alters activity of the brain reward/reinforcement system (Sinha, 2008; Uhart & Wand, 2009). Inborn biological qualities define the sensitivity of the stress system and ability to adapt (Kudielka et al., 2009; Meaney et al., 1991; Regev & Baram, 2014). Assessment of the meaning and significance of environmental stressors relies on the limbic‐affective processing circuits, including amygdala interactions with subcortical and prefrontal cortical areas, further diversifying individual response to stress (Sinha, 2008). The stress response also depends on characteristics of the stressor itself: predictability, duration, controllability and so forth (Keyes et al., 2012). Studies on the association between stress experiences and substance consumption have focused on stressor‐specific effects and intra‐individual/interindividual moderators of stress‐drug association.

Trait and state anxiety are closely interconnected with stressful experiences (Daviu, Bruchas, Moghaddam, Sandi, & Beyeler, 2019). Trait anxiety and anxiety disorder (adhering to DSM/ICD classification) are known to be associated with substance use disorders (for a review, see Belin, Belin‐Rauscent, Everitt, & Dalley, 2016), suggesting that momentary anxiety state may be one predictor of increased probability of substance intake. Unfortunately, ecological momentary assessment‐related studies frequently include “negative affect” and “negative mood” as momentary internal states instead of defining them as specific states. Mood and affect are complex DSM/ICD oriented states (e.g. negative mood encompasses anxious, nervous, jittery, irritable, angry, frustrated, down, blue, depressed and sad) cannot be studied in animals and are not included in the research domain criteria system. Fortunately, some studies (e.g. Simons, Dvorak, Batien, & Wray, 2010) have investigated the association of anxiety with alcohol consumption as a subdimension of affect.

Finally, studies indicate that during alcohol deprivation in animals (Spanagel & Hölter, 1999) and acute withdrawal (and likely beyond) in humans (Heilig, Egli, Crabbe, & Becker, 2010; Higley, Koob, & Mason, 2012), anxiety levels and stress sensitivity are elevated, which may contribute to increased probability of stress‐induced relapse.

3.1. Animal studies

Preclinical researchers use several protocols to study the effect of stressors on alcohol‐related behaviours, some of which implement a longitudinal design. In these studies, alcohol consumption is recorded for several days/weeks, and changes in consumption are measured in response to short (i.e. 5‐ to 15‐min duration) once‐a‐day exposures to a stressor (e.g. cold swim, foot shock and restraint). Widely varying outcomes are observed: under stressful conditions, alcohol intake has been found to not only increase but also decrease and remain unaffected (e.g. Vengeliene et al., 2003; Boyce‐Rustay, Janos, & Holmes, 2008; for a review, see Becker, Lopez, & Doremus‐Fitzwater, 2011), demonstrating that, similar to in humans, stress and alcohol‐drinking interactions in animals are complex. Interindividual differences in the stress response have been found using different rodent strains/lines (e.g. Boyce‐Rustay et al., 2008; Vengeliene et al., 2003). Studies using voluntary long‐term (i.e. several months) drinking rats have demonstrated that stress‐associated increases in alcohol consumption are modified not only by differences in the genetic background but also depend on the age of alcohol drinking initiation and type of stressor (Siegmund, Vengeliene, Singer, & Spanagel, 2005; Vengeliene et al., 2003). Furthermore, stress exposure during a drug‐free period was shown to promote relapse‐like behaviours, including alcohol seeking (e.g. Funk, Vohra, & Lê, 2004; Lê & Shaham, 2002); this association has also been shown to be stressor‐specific. A meta‐analysis (Noori, Helinski, & Spanagel, 2014) demonstrated that the choice of experimental paradigm and type of stressor are among the most significant variables in the induction of stress‐related alcohol consumption, concluding that exposure to forced swim and foot shock stress were most likely to increase free‐choice alcohol consumption in rats. Although these are different stressors than those experienced by humans and animals in the wild, the translational potential of these studies is based on similarities in physiological response to stressors. For instance, the above‐mentioned meta‐analysis demonstrated that pharmacologically induced stress‐like responses in animals (by administration of non‐selective α₂‐adrenoceptor antagonist yohimbine) and foot shock stress are equally effective in reinstating alcohol‐seeking in rodents (Noori et al., 2014).

Alcohol deprivation in voluntary drinking rodents (depending on species and genetic background) may lead to increased alcohol consumption upon its re‐exposure (Vengeliene, Bilbao, & Spanagel, 2014) and even can irreversibly change behaviour (Foo et al., 2017; Vengeliene et al., 2003). An intensive longitudinal data‐based study (alcohol consumption and locomotor activity recorded for several months at a resolution of 1 min) indicated that irreversible changes in rat behaviour occur during the very first alcohol deprivation (Foo et al., 2017). It is unclear which internal state of an animal during deprivation is associated with subsequent higher/lower alcohol consumption during an alcohol relapse‐like situation. One candidate is anxiety; during alcohol withdrawal, higher anxiety levels have been reported in voluntary drinking or vapour self‐administering rats (de Guglielmo et al., 2017; Spanagel & Hölter, 1999), and repeated stress exposures have been shown to sensitize animals to withdrawal‐induced anxiety (Breese, Overstreet, & Knapp, 2005). A positive correlation between trait‐anxiety and higher alcohol consumption has been found in some studies (e.g., for a review, see Belin et al., 2016), and a study in mice selectively bred for high alcohol drinking demonstrated that the anxiolytic effect of alcohol is not responsible for high alcohol intake (Barkley‐Levenson & Crabbe, 2015). The above findings suggest that the association between momentary anxiety states and probability of alcohol consumption require further investigation.

3.2. Human studies

Momentary factors related to negative valence are studied as perceived stress caused by different daily negative events (e.g. work, money, household, social and family) and daytime anxiety (studied as a subdimension of negative affect). Unfortunately, these factors have not yet been studied in abstinent alcohol users (Table 1).

TABLE 1.

Association between different momentary factors and probability of drinking

Factor	Participants		Methods		Results	Reference
	N	Age	Measures (factor, drinking)	Study period (days)	Probability of drinking
Negative valence
Stress exposure/feeling stressed	S1: 83 S2: 88	S1: 37 a S2: 34 a	S1: DD, DD S2: EMA/DD, EMA/DD	S1: 60 S2: 30	↓ in individuals with low coping‐related motives to drink; no association in individuals with high coping‐related motives to drink	Todd, Armeli, Tennen, Carney, & Affleck (2003)
	386	18–70	EMA/uEMA/DD, EMA/uEMA/DD	21	↑ when faced with work/school stressors; decreased when faced with financial stressors coupled with hangover	Epler et al. (2014)
	200	40 a	DD, DD	7	No association in heavy SMM drinkers with high addiction severity; ↓ in heavy drinkers with low addiction severity	Mereish, Kuerbis, & Morgenstern (2018)
	S1: 85 S2: 28	18–65	DD, DD	S1: 14 S2: 56	No direct association between stress and drinking in AUD (i.e. a stressful event predicted craving)	Wemm, Larkin, Hermes, Tennen, & Sinha (2019)
Anxiety	102	18–24	EMA, DD/EMA	21	↑ in individuals with high negative urgency	Simons et al. (2010)
	100	18–25	EMA, DD/EMA	21	↑ in men with high cognitive abilities	Dvorak & Simons (2014)
	1,636	19 a	DD, DD	30	2 SD ↑ in daily anxiety predicting an increase of approximately 0.02 drinks consumed	O'Hara, Armeli, & Tennen (2014)
	35	50–74	EMA, EMA	14	↑ alcohol and cannabis use	Paolillo et al. (2018)
Positive valence
Cue exposure	37	18–70	uEMA, a phone interview	10	No association with relapse	Witteman et al. (2015)
Craving	102	18–	EMA, EMA	14	No association with relapse	Cooney et al. (2007)
	112	21–	uEMA, uEMA	5	↑ during the first 2 drinks	Ray et al. (2010)
	29	45 a	EMA, EMA	28	No association with relapse	Holt, Litt, & Cooney (2012)
	42	15–20	EMA, DD	7	↑ alcohol use associated with higher average daily craving levels	Ramirez & Miranda (2014)
	159	18–65	EMA, EMA	14	↑ alcohol, tobacco and cannabis use (no association with opiate use)	Serre, Fatseas, Denis, Swendsen, & Auriacombe (2018)
	172	18–66	EMA/uEMA, EMA/uEMA	7	↑ (stronger association in individuals with lower self‐control)	Remmerswaal, Jongerling, Jansen, Eielts, & Franken (2019)
	S1: 85 S2: 28	18–65	DD, DD	S1: 14 S2: 56	↑ in AUD (craving directly mediates the association between stress and alcohol use)	Wemm, Larkin, Hermes, Tennen, & Sinha (2019)
	100	25–65	EMA, EMA	14	↑	Jones, Tiplady, Houben, Nederkoorn, & Field (2018)
Cognitive processes
Inhibitory control	100	25–65	EMA, EMA	14	↑ with worsening of inhibitory control	Jones, Tiplady, Houben, Nederkoorn, & Field (2018)
Impulsivity (urgency)	211	21–35	EMA, BQ	10	↑ in ADHD	Pedersen, King, Louie, Fournier, & Molina (2019)
Impulsivity (LP)	211	21–35	EMA, BQ	10	↑ in ADHD	Pedersen, King, Louie, Fournier, & Molina (2019)
Impulsivity (SS)	211	21–35	EMA, BQ	10	No association	Pedersen, King, Louie, Fournier, & Molina (2019)
Impulsivity (LPE)	211	21–35	EMA, BQ	10	No association	Pedersen, King, Louie, Fournier, & Molina (2019)
Social processes
Rejection	77	18–44	uEMA, uEMA	14	↑ when experiencing more social rejection by close friends/relatives	Laws, Ellerbeck, Rodrigues, Simmons, & Ansell (2017)
Social communication (DR)	149	15–18	EMA, EMA	2 WE	↑ with more people and with mixed gender composition at party; ↑ with friends at non‐party events	Lipperman‐Kreda, Finan, & Grube (2018)
	83	18–30	EMA, EMA	21	9× ↑ when drinking with others than alone	O'Donnell et al. (2019)
Arousal and regulatory systems
Poor sleep quality	139	20–73	EMA, EMA	7	↑ in younger age; ↓ in older age	Kuerbis et al. (2018)
Low arousal	139	20–73	EMA, EMA	7	No association in younger age; ↑ in older age	Kuerbis et al. (2018)
High arousal	53	18–60	EMA, uEMA	10	↑	Peacock, Cash, Bruno, & Ferguson (2015)

Note: Factors are classified into five domains that represents different aspects of emotional, cognitive, motivational, social and other functions: negative and positive valence, cognitive and social processes and sleep/arousal related functions. Only those studies which reported intra‐individual variation of a risk factor are included in the table. Columns provide information on participants: number (N) and age (range or mean); study methods: which method was used to collect data on target variables, factor and drinking (i.e. once a day in case of DD, multiple times a day in case of EMA or a single assessment of baseline questionnaires prior to study in case of phone/face interview during BQ); study period (days) and probability of drinking in association with the momentary factor experienced prior to drinking.

Abbreviations: ↑, increased; ↓, decreased; ADHD, attention‐deficit hyperactivity disorder; AUD, alcohol use disorder; BQ, baseline questionnaires; EMA, signal‐contingent ecological momentary assessment; DD, daily diaries; DR, drinking reference group; LP, lack of planning; LPE, lack of perseverance; S1/S2, study 1/study 2; SMM, sexual minority men; SS, sensation seeking; uEMA, user‐initiated ecological momentary assessment; WE, weekend, Friday evening through Sunday morning.

^aMean age due to no information for the range of age.

Although stress experience has been linked to increased probability of substance use, stress has not been extensively studied using momentary assessment methods in alcohol research. One study (Epler et al., 2014) found that decreased time to next drink was related to work/school stressors; however, another study did not find a clear association between stress and alcohol drinking (i.e. a stressful event did not predict drinking directly, but predicted increased craving that in turn predicted drinking; Wemm et al., 2019). Mereish et al. (2018) showed that, depending on drinking history, stressful experiences can even reduce alcohol consumption. This inconsistency may be at least partly explained by Todd et al. (2003), which showed that alcohol may not have a stress‐response dampening effect in some people (e.g. those with low self‐reported drinking‐to‐cope).

Simons et al. (2010) showed that negative affect could not be generalized as having a clear direction of association with changes in alcohol use. In this study, state anxiety (i.e. nervous, jittery and anxious) was associated with alcohol consumption for individuals higher in trait negative urgency (i.e. tendency to act rashly when distressed). Age may also play a role in determining the effects of anxiety on drinking; a study in college students found an association between daytime anxiety and evening drinking (O'Hara et al., 2014), whereas in middle‐aged adults, higher anxiety levels were not linked to the current alcohol (and cannabis) use event but predicted an upcoming one (Paolillo et al., 2018; Table 1). These studies demonstrate that in some individuals, anxiety might be associated with increased alcohol use. However, individual characteristics of participants, such as differences in cognitive abilities and impulsivity, as well as their gender, might interfere with this association (Dvorak & Simons, 2014; Simons et al., 2010). Further research will be needed to better characterize the complex relationships between anxiety and alcohol use.

3.3. Pharmacological interventions

In animal studies, reduction of stress‐induced alcohol seeking has been well‐studied especially using a stress‐induced reinstatement of alcohol seeking paradigm. Stress‐induced alcohol seeking in animals is reduced by agents modulating activity of the stress system, such as CRF₁ and glucocorticoid antagonists, α₁ antagonists and α₂ agonists, several opioidergic compounds, as well as by administration of orexin receptor 1 (OX₁) antagonists, different serotonergic, glutamatergic and GABAergic modulators, and others (for a review, see Mantsch, Baker, Funk, Lê, & Shaham, 2016). In preclinical studies, CRF is the most studied neuropeptide system in the context of stress‐related behaviours. Compounds that decrease CRF signalling are demonstrated not only to reduce alcohol withdrawal symptoms (Higley et al., 2012) but also have an anxiolytic‐like effect (Griebel & Holmes, 2013). Unfortunately, multiple clinical trials of CRF₁ antagonists were negative (Spierling & Zorrilla, 2017). This might be explained by the stress system being susceptible to neuroplastic adaptive changes (Regev & Baram, 2014), leading to rapid development of tolerance to stressful life challenges as well as pharmacological treatments targeting this system. The stress‐responsive arginine‐vasopressin system and specifically the V_1B receptor might be a new promising target to treat stress‐ or anxiety‐related drinking (for a review, see Zhou & Kreek, 2018), however more research is needed to support this notion.

4. POSITIVE VALENCE

Positive valence refers to response to positive motivational situations or contexts, such as initial response to reward, anticipation and valuation of reward, reward‐related learning, seeking, craving and habit formation. Behavioural functions of this domain are highly dependent on the stage of drug use (i.e. whether drug use is in initial stages, chronic, or if the user is experiencing withdrawal). Initial response to drug is generally thought to be rewarding/positively reinforcing in most users, and response to drug‐predicting stimuli may induce reward anticipation, at least in some individuals. However, chronic drug use not only leads to the development of tolerance and decreased hedonic effects but can also permanently change brain circuits normally regulating the perception of reward‐related stimuli. Specifically, decreased sensitivity to reward may occur, hypersensitized response of reward/reinforcement systems to drugs and drug‐associated stimuli may develop, drug‐related action–outcome associations may become excessively optimistic, and maladaptive automatic stimulus–response habits may form (for a review, see Robinson & Berridge, 2003). These changes in positive valence‐related functions have been studied in the context of different theories of substance use disorder (Bickel et al., 2018).

From the perspective of momentary processes, two factors have mainly been studied—response to drug‐associated stimuli and craving. The presence of drug‐associated stimuli (including internal cues) elicits cue‐reactivity response, which often correlates with increased craving (Drummond, 2000). Sinha et al. (2003) demonstrated that cue‐induced and stress‐induced craving were associated with increased activity in similar brain regions (i.e. the hypothalamic–pituitary–adrenal axis and noradrenergic/sympatho‐adreno‐medullary system), suggesting some degree of overlap between stressful experiences and response to environmental stimuli predicting drug effects. Craving may be responsible for the maintenance and re‐initiation of substance use in some individuals (Tiffany & Conklin, 2000, please see Serre, Fatseas, Swendsen, & Auriacombe, 2015 for a comprehensive review on EMA‐based craving studies). In others, exposure to drug‐associated stimuli may trigger an automatic drug‐seeking response leading to re‐initiation of substance use (Tiffany & Conklin, 2000).

4.1. Animal studies

Multiple protocols have been established to assess the positive reinforcing effects of alcohol (and other substances) and reward‐learning‐related behaviours (e.g. conditioned reinforcement, Pavlovian and instrumental conditioning, and cue‐/context‐induced seeking) and to study the impact of positive valence‐related functions in the acquisition, maintenance, and re‐acquisition of drug use (for a review, see Sanchis‐Segura & Spanagel, 2006). It has been demonstrated that similar to humans, alcohol‐associated stimuli (cues and contexts) in animals also acquire incentive‐motivational properties and become conditioned reinforcers. Cue‐conditioning studies have played a major role in the study of momentary risk factors associated with changed drug consumption. Craving (albeit extensively studied in clinical research) cannot be reliably identified in animals (Littleton, 2000), but it is assumed that elevated craving at least partly drives increased cue‐/context‐induced drug seeking. This assumption is based on testing predictive validity of the cue‐induced reinstatement of drug seeking paradigm, for instance naltrexone is known to decrease alcohol craving in humans (Hendershot, Wardell, Samokhvalov, & Rehm, 2017) and has also been effective in preventing cue‐induced alcohol seeking in rats (Ciccocioppo, Lin, Martin‐Fardon, & Weiss, 2003).

As mentioned in Section 3, it is unclear what (alcohol deprivation‐caused) internal state of an animal is responsible for the subsequent higher/lower alcohol consumption during a relapse‐like situation. However, increased post‐abstinence alcohol consumption may reflect changes in the rewarding/reinforcing value of alcohol (Lê & Shaham, 2002). Conditioned reinforcers in animal experiments are typically introduced as a contingent consequence of operant responding for drug‐reward. Although several protocols to study conditioned reinforcement processes have been established (Sanchis‐Segura & Spanagel, 2006), cue‐induced reinstatement of alcohol‐seeking behaviour is the most frequently used paradigm (Lê & Shaham, 2002). During this test, re‐exposure of animals to alcohol‐associated cues increases operant responding, and the introduction of a withdrawal phase before the reinstatement test leads to incubated cue‐induced alcohol seeking (Pickens et al., 2011). The conditioned drug self‐administration training phase to test alcohol seeking is relatively short (i.e. 10–12 sessions); the shortness of this exposure might have a negative impact on success of clinical translatability of alcohol‐seeking‐related findings. The value of long‐term (i.e. 44–55 sessions) training was demonstrated by Domi et al. (2019) who showed that during long‐term training procedures, interindividual variability between animals can be detected with respect to motivation and persistence to self‐administer. Persistence of alcohol seeking was recorded during every self‐administration training session during a period where responding was neither reinforced by conditioned stimuli nor alcohol delivery. It was found that one third of rats progressively increased responding for alcohol during these alcohol‐free periods independently of whether they were selectively bred for high alcohol preference (Domi et al., 2019). The impact of inborn individual differences on cue‐conditioned behaviours has also been demonstrated using different rodent strains (Lederle et al., 2011).

Administration of alcohol in rats is known to elevate extracellular dopamine concentration in both the ventral and dorsal striatum (Di Chiara & Imperato, 1988); consequently, lower mesolimbic dopaminergic neuronal activity is measured in rats withdrawn from chronic alcohol administration (Diana, Pistis, Carboni, Gessa, & Rossetti, 1993). This hypo‐dopaminergic state may be responsible for the development of transient reward deficiency (i.e. decreased sensitivity to reward) reported in animal studies using an intracranial electrical self‐stimulation paradigm (e.g. Schulteis & Liu, 2006). These studies also showed that magnitude and duration of decreased sensitivity to reward depended on the duration of alcohol exposure. However, a recent study in long‐term alcohol vapour‐exposed rats demonstrated that withdrawal causes dynamic, oscillatory‐like changes in the activity of brain reward/reinforcement circuits (Hirth et al., 2016). This finding was confirmed by a meta‐analysis conducted on dopamine levels in the ventral striatum demonstrating that in alcohol withdrawn rodents, regardless of the alcohol exposure paradigm (free‐choice, forced administration, etc.) and duration, the hypo‐dopaminergic state is short‐lasting and followed by a long‐lasting hyper‐dopaminergic state (Hirth et al., 2016). The dynamic regulation of the mesolimbic dopamine system during deprivation suggests state‐dependent changes in relapse risk factors, related to either reward deficiency (hypo‐dopaminergic state) or multiple other behavioural impairments associated with hyper‐dopaminergic states (Cinque et al., 2018; Vengeliene et al., 2017). Interestingly, hyper‐dopaminergic states in animals are also shown to be associated with decreased sensitivity to reward (Cinque et al., 2018; Vengeliene et al., 2017), suggesting that reward sensitivity follows an inverted U‐shape corresponding to dopaminergic system activity. The impact of duration of alcohol deprivation on characteristics of post‐abstinence drinking has not been well studied in animals. However, a few studies using the free‐choice post‐abstinence drinking approach demonstrated that a protracted deprivation phase in rats (i.e. 2–3 weeks) increased frequency of approaches to alcohol bottles during the first post‐abstinence days compared to baseline drinking conditions (Vengeliene et al., 2013; Vengeliene, Noori, & Spanagel, 2015), demonstrating increased “wanting” to consume alcohol. The amount of alcohol consumed per approach remained similar to baseline amounts, suggesting that hedonic value of alcohol did not change.

4.2. Human studies

Response‐to‐alcohol‐associated cues and craving have been studied as momentary factors related to positive valence (Table 1). Unfortunately, there are too few EMA studies related to cue exposure to draw definite conclusions on the behavioural impact of cues. A study in adolescents demonstrated that the presence of alcohol cues in the natural environment was associated with higher amounts of craving, and subsequently, higher daily average craving levels predicted greater volumes of alcohol consumption (Ramirez & Miranda, 2014). In another study, the number of self‐reported exposures to alcohol cues in patients discharged from detoxification treatment did not predict relapse (Witteman et al., 2015), suggesting that other factors may play a significant role. In general, studies have demonstrated that craving was a predictor of increased “baseline” alcohol use in both adolescent and adult drinkers (Ramirez & Miranda, 2014; Ray et al., 2010; Remmerswaal et al., 2019; Serre et al., 2018; Wemm et al., 2019). Like cue exposure studies, ecological momentary assessment studies have suggested that craving alone may not be a good predictor of relapse (Cooney et al., 2007; Holt et al., 2012; Witteman et al., 2015).

4.3. Pharmacological interventions

In animal research, reduction of alcohol‐associated cue‐/context‐induced reinstatement of alcohol seeking by administration of different compounds has been the main strategy in discovery of novel pharmacological targets to reduce cue reactivity in humans (for reviews, see Spanagel, 2009; Bossert et al., 2013). Cue‐induced seeking in rats is reduced by compounds acting on dopamine, opioid, cannabinoid, OX₁, glutamate receptors and others (Spanagel, 2009). Similar targets have been tested in reducing voluntary alcohol consumption, operant self‐administration and relapse‐like consumption (Lê & Shaham, 2002; Spanagel, 2009), however translational success of novel anti‐relapse compounds has been limited (Bossert et al., 2013). This may be related to the fact that alcohol craving correlates with seeking and relapse only in specific individuals (Tiffany & Conklin, 2000), highlighting the necessity of studying interindividual variation in preclinical research. For instance, a recent intensive longitudinal data study (Foo et al., 2019) demonstrated that nalmefene was the most effective in rats that consumed greater amounts of highly concentrated alcohol per drinking approach prior to drug treatment.

Modification of the impact of positive valence‐related factors has been attempted in ecological momentary assessment studies. A study in adolescents demonstrated that naltrexone reduced alcohol cue‐potentiated craving and drinking (Miranda et al., 2014). Another study using the anticonvulsant topiramate showed blunting of alcohol craving while drinking and consequent reduction of the number of drinks consumed across drinking episodes (Miranda et al., 2016). Interestingly, a study in rats demonstrated that in a relapse‐like situation, another anticonvulsant drug, lamotrigine, did not affect drinking frequency but dramatically reduced alcohol intake during a drinking approach, demonstrating that this compound affected hedonic value of alcohol but not “wanting” to consume alcohol (Vengeliene et al., 2013). In addition, ecological momentary assessment‐based studies have confirmed the role of interindividual differences in drug response showing that naltrexone reduced craving and heavy drinking in carriers of the dopamine receptor DRD4‐L allele variant (Miranda, Treloar Padovano, Gray, Wemm, & Blanchard, 2018), while topiramate reduced drinking and increased confidence in avoiding heavy drinking later in the day in GluK1 kainate subunit rs2832407 C‐allele homozygotes (Kranzler et al., 2016).

5. COGNITIVE SYSTEMS

Cognitive systems are concerned with processes such as attention, external environmental perception, working memory, and behavioural control. Cognition provides a means to adapt to an ever‐changing environment. Hence, it is not surprising that impaired cognitive performance is the most common characteristic in ICD/DSM symptom‐based definitions of mental disorders. Among the constructs of the cognitive domain (i.e. perception, declarative memory, cognitive control, and working memory), cognitive control has been linked to substance use (e.g. Bechara, 2005). Specifically, the trait “impulsivity” has garnered attention due to its association with compulsive drug taking and seeking behaviours (Belin et al., 2016; Dalley, Everitt, & Robbins, 2011). Impulsivity has also been given a central role in several addiction theories, such as self‐control failure models and theories suggesting a dual‐systems model of decision making and cognitive control (Bickel et al., 2018). These theories emphasize the role of choosing in situations where two systems have overlapping control over behaviour—the reward‐driven impulsive system and the executive system responsible for self‐regulatory processes. From the perspective of a state‐dependent change in cognitive performance, increased impulsiveness/failed impulse control can be seen as a momentary factor associated with the increased probability of substance consumption (de Wit, 2009).

5.1. Animal studies

The heterogeneity of impulsive actions in humans (see below) requires development of complementary tests in animals in order to maintain the translational value of preclinical findings. In animals, impulsivity testing can be subdivided into testing impulsive choice (i.e. a human analogue of non‐planning impulsivity in delay‐discounting paradigms), motor impulsivity, and general attentional abilities (i.e. behavioural spontaneity in a stop‐signal reaction time task, the continuous performance test, and the go/no‐go paradigms; for a review, see Dalley et al., 2011; Winstanley, 2011). Rats characterized as highly impulsive in the delay‐discounting paradigm consumed higher amounts of alcohol (Poulos, Le, & Parker, 1995). Rats selectively bred for high alcohol preference demonstrated higher behavioural spontaneity and impulsive choosing (Beckwith & Czachowski, 2014, 2016), suggesting that an inborn impulsivity trait may contribute to higher alcohol consumption. Furthermore, a rat study demonstrated that individual differences in alcohol‐induced impulsivity were positively correlated with subsequent voluntary alcohol consumption (Poulos, Parker, & Lê, 1998).

The association between momentary impulsivity and alcohol consumption has yet to be established. Provoking a temporal impulsive state in animals or identifying naturally occurring variation in impulsivity due to the dynamic nature of neurochemical systems could serve as potential approaches to study momentary impulsivity in animals. A behavioural model of negative urgency (i.e. tendency to act rashly under distress) has recently been established in rats, whereby a momentary behavioural state is induced by unexpected reward omission (Gipson et al., 2012). Concerning the dynamic nature of neurochemical systems during alcohol withdrawal, as mentioned above, oscillatory‐like changes in dopaminergic activity were observed in both the ventral and dorsal striatum (Hirth et al., 2016). Several studies indicate that impulsivity is related to dysfunction of the frontostriatal circuits (for a review, see Dalley et al., 2011), and Carvalho et al. (2017) reported that in rats depleted of mesolimbic and nigrostriatal dopamine, a sudden rise in brain dopamine concentration led to choice impulsivity.

5.2. Human studies

Chronic alcohol consumption increases impulsive behaviours and impairs cognitive functions, for example, deficits in attention, reaction time, mental flexibility, learning, and short‐term memory. These effects may remain even after long‐term abstinence and continue to contribute to the maintenance of alcohol use (e.g. de Wit, 2009; Fein, Bachman, Fisher, & Davenport, 1990). Impulsive actions can be caused/triggered by various factors and are subdivided into negative or positive urgency (i.e. acting rashly when distressed or in response to positive affect), sensation seeking, lack of planning or inability to remain focused, and lapses in attention (de Wit, 2009). As a result, impulsivity is usually studied as a multidimensional construct. A momentary self‐report impulsivity scale (MIS) has been developed to capture different aspects of impulsivity—motor, urgency, non‐planning, lack of perseverance, and attentional impulsivity (Tomko et al., 2014) and it is reported that alcohol use is associated with increased mean impulsivity scores on the same day (Trull, Wycoff, Lane, Carpenter, & Brown, 2016). Meta‐analyses have suggested that all facets of impulsivity are associated with alcohol consumption to a certain extent and report that problematic alcohol consumption is most highly related to both negative and positive urgency (e.g. Coskunpinar, Dir, & Cyders, 2013).

The association between momentary state‐dependent impulsivity and alcohol use has not been studied systematically (Table 1). Several ecological momentary assessment‐based studies have reported moderating effects of trait impulsivity on alcohol consumption (Black, Cooney, Sartor, Arias, & Rosen, 2018; Gaher et al., 2014; Simons et al., 2010). A study by Pedersen et al. (2019) suggested that variability in urgency (positive and negative) and lack of planning may be associated with alcohol‐related problems. In addition to these subjective measures, testing objective impulsivity measures (i.e. stop‐signal reaction time) shows that worsening of inhibitory control over the day predicts increased alcohol consumption later that day (Jones et al., 2018).

5.3. Pharmacological interventions

Cognitive behavioural therapy in treatment‐seeking individuals is based on strengthening aspects of executive control over learned behavioural patterns and reduction of impulsive responding to, for example, cue exposure (Sofuoglu, DeVito, Waters, & Carroll, 2013). Decreasing trait impulsiveness has also been a focus of pharmacological studies. For instance, Rubio, Martínez‐Gras, and Manzanares (2009) demonstrated reductions in craving, and alcohol consumption associated with topiramate treatment may be related to changes in trait impulsivity. It should be mentioned, however, that anticonvulsants, such as topiramate, have been shown to cause cognitive decline (Walker & Lawrence, 2018). Preclinical studies showed that the noradrenaline/5‐HT reuptake inhibitor atomoxetine, several 5‐HT_2A antagonists, a 5‐HT_2C agonist WAY‐163909 and several other monoaminergic compounds decreased impulsivity in both animals and humans (for a review, see Winstanley, 2011). Two monoaminergic compounds, modafinil and guanfacine, are currently under development for treatment of alcohol use disorder and have been demonstrated to significantly improve self‐report measures of impulsivity (Walker & Lawrence, 2018). Given the crucial role of the frontal cortex in mediating impulsive actions (Dalley et al., 2011), the glutamatergic system and especially metabotropic glutamate receptors are considered promising targets (Winstanley, 2011). Interestingly, the glutamatergic system is also considered as a promising target in preclinical alcohol research due to the effectiveness of compounds targeting this system in other alcohol use associated behaviours (e.g. cue‐ and stress‐induced alcohol‐seeking; Holmes, Spanagel, & Krystal, 2013). These targets have the potential to be used to evaluate changes in momentary impulsivity using an ecological momentary assessment approach.

6. SOCIAL PROCESSES

The social domain includes such processes as understanding of the self and others, the development of social bonds, and the ability to integrate and interact in the social environment. An interplay of genetic and environmental factors (e.g. early disruption of healthy social environment) contributes to the formation of social personality traits and, in some cases, impairments of social functions (Tomalski & Johnson, 2010). Although DSM‐/ICD‐based behavioural syndromes are not the main focus of this review, it is worth mentioning that some syndromes related to social impairments (e.g. antisocial and conduct) are found to be associated with drug‐related problems (Goldstein et al., 2007), suggesting that impaired social functions may lead to increased drug seeking and consumption. Oxytocin is released in response to a variety of social stimuli and activates, among others, the mesocorticolimbic reward/reinforcement system (Love, 2014), suggesting that social encounters may be protective against drug use due to their rewarding and stress‐buffering effects (Smith & Wang, 2014; Tops, Koole, IJzerman, & Buisman‐Pijlman, 2014). On the other hand, stressful social interactions are known to activate the central stress system (Dickerson & Kemeny, 2004), which is thought to be associated with substance use, and social exclusion might be one factor associated with increased substance use (Heilig, Epstein, Nader, & Shaham, 2016).

6.1. Animal studies

While the impact of social processes on drug use has been extensively studied in animals, it is important to note that human social organization is different (e.g, human “social norms” are absent in other species). This should be kept in mind when considering the translational value of preclinical findings. At the same time, animal societies have evolved their own particular organizational aspects. For instance, mice are a highly territorial and hierarchically organized species (Poole & Morgan, 1976) and, accordingly, lack certain features in social behaviour that are characteristic to other species (e.g. elaborate social play; Pellis & Pasztor, 1999) but they can be used in studies involving attack/defence interactions (e.g. social defeat).

Non‐human primate studies have better translational value owing to their similar social organization to humans. These studies indicate that early maternal separation may cause higher alcohol consumption under normal living conditions later in life (Barr et al., 2009; Fahlke et al., 2000); like humans, non‐human primates usually give birth to single offspring and invest large amounts of energy and time into caring and protecting them. Therefore, early maternal separation in primates may have similar impact on behaviour as that observed in neglected children. Peer‐reared non‐human primates are found to have higher neuroendocrine responses to social separation stress than mother‐reared control animals and these responses are associated with higher alcohol consumption (Barr et al., 2009; Fahlke et al., 2000), demonstrating the moderating effect of interindividual differences on the relationship between momentary social stressors and alcohol use. Both long‐term maternal separation and chronic social isolation during adolescence were also found to be associated with higher alcohol consumption in rodents (for a review, see Becker et al., 2011).

In rodents, the impact of momentary variation in the social environment has been studied as short‐term changes in housing conditions (e.g. brief social isolation and overcrowding), intermittent social interaction, and exposure to social cues. Changes in housing conditions in rats have not produced consistent results with respect to alcohol consumption (for a review, see Becker et al., 2011). Exposure to social cues has been studied using the social transmission paradigm, whereby it was demonstrated that social cues (i.e. interacting with an intoxicated sibling) can lead to increased alcohol consumption (Maldonado, Finkbeiner, & Kirstein, 2008). Similar results were obtained by providing rats with opportunities for social interaction, demonstrating that alcohol consumption was higher when access to alcohol was given combined with the opportunity for social interaction (Tomie, Gittleman, Dranoff, & Pohorecky, 2005). However, when rats were given continuous access to social interaction, alcohol intake was lower compared to rats that were provided with intermittent access opportunities (Tomie, Lewis, Curiotto, & Pohorecky, 2007), indicating that social communication‐related factors can affect probability of alcohol use. That social stimuli function as contextual cues in rats has also been demonstrated by the study of Browning and Shahan (2018).

Another area of social domain‐related research is social rank. Studies show that lower social rank is correlated with increased alcohol consumption in rodents (for a review, see Becker et al., 2011). However, a clear relationship between social rank and alcohol consumption has not been established in non‐human primates (Becker et al., 2011). The impact of social rank on alcohol consumption has also been studied longitudinally. Wolffgramm and Heyne (1995) found that dominant rats were more sensitive to changes in their social environment (i.e. reduction of social contact and appearance of unknown conspecifics) with respect to alcohol consumption, causing them to increase consumption while these changes did not affect alcohol consumption in subordinate animals. Transient state‐induced social rank (i.e. induced during short‐term interaction with a conspecific) has been studied in a resident–intruder procedure as a response to a social defeat (Becker et al., 2011). In these studies, the intruder decreased alcohol consumption immediately after defeat, but depending on genetic background (e.g. CRF₁ knockout mice), delayed increase in alcohol intake was observed (Becker et al., 2011).

6.2. Human studies

The impact of social processes on substance use in humans is exceptionally complex. Substance use is a learned, context‐dependent behaviour and human “social norms” introduce many biases toward drug‐related behaviours (e.g. stigma against drinking in women, being “cool”/”uncool” among youth and parental supervision in adolescents). On the other hand, social structure (e.g. drug user's social status; de Vries et al., 2019) and social environment (e.g. drinking alone vs. pleasant social situation; Armeli et al., 2003) modify drug effects and change a drug user's behaviour. In addition, due to its licit status, alcohol use is indirectly promoted by society, and this should be taken into consideration when interpreting findings related to social processes and translatability of animal research. Most alcohol drinking in humans occurs in social contexts and social motivation plays significant role in drinking (Lipperman‐Kreda et al., 2018). In contrast, in animals, reduction of social contact is usually associated with increased alcohol consumption (Tomie et al., 2007; Wolffgramm & Heyne, 1995).

The effect of social communication on drinking has been studied using the ecological momentary assessment approach (Table 1). One study showed that social contexts with friends at non‐party events and larger numbers of people and mixed gender composition at parties are associated with increased probability of drinking in adolescents (Lipperman‐Kreda et al., 2018). Another study found that young adults surrounded by drinking people are more likely to show both an increased probability of drinking initiation and higher levels of consumption (O'Donnell et al., 2019). Further ecological momentary assessment‐based studies are needed to assess if/how momentary social interactions (e.g. seeking social comfort during distress, managing recurrent social threats, forming new social bonds) correlate with increased alcohol use. To date, only one study has reported that social rejection by close people (e.g. family members and close friends) is associated with increased drinking, whereas rejection by strangers and less familiar individuals had no effect on drinking (Laws et al., 2017).

6.3. Pharmacological interventions

Several brain systems regulate social behaviours (e.g. opioid, cannabinoid and monoamine; Vanderschuren, Niesink, & Van Ree, 1997), but two neuropeptide systems, oxytocin and opioid, have earned the majority of attention. Oxytocin regulates a number of behaviours involving social interactions and has prosocial, anxiolytic, anti‐depressive and stress‐buffering effects (Neumann & Landgraf, 2012). Contextual variability (e.g. valence of social stimuli) and individual differences (e.g. attachment anxiety trait) may moderate effects of exogenously applied oxytocin (Bartz, Zaki, Bolger, & Ochsner, 2011), and these variables could be studied using an ecological momentary assessment approach. Importantly, findings related to pharmacological manipulation of social behaviour are species‐specific (for a review, see Loseth, Ellingsen, & Leknes, 2014). For instance, social play behaviour in rats is increased by administration of different μ‐opioid agonists and decreased by antagonists (for a review, see Vanderschuren et al., 1997), but in non‐human primates, opioidergic compounds have the opposite effect on motivation for social interaction (Loseth et al., 2014). In human studies, the opioid receptor antagonist naltrexone increased attention to emotional expressions but slowed identification of sadness and fear (Wardle, Bershad, & de Wit, 2016). To further explore relationships between social behaviour and alcohol use, medications like naltrexone and nalmefene should be tested in alcohol using individuals within an ecological momentary assessment approach.

7. AROUSAL AND REGULATORY SYSTEMS

Arousal and regulatory systems are responsible for the regulation of motor activity levels, sensory and emotional reactivity, and wakefulness and sleep. These systems are controlled by motivationally significant environmental stimuli (e.g. threat), internal stimuli (e.g. emotions), and homeostatic drives (e.g. hunger and circadian rhythms). The impact of arousal states (e.g. alert, active and energetic vs. calm, relaxed and bored) on substance use has not been well studied, as these states are integrated in “mood” and “affect” concepts and are seldom investigated separately. In contrast, sleep quality is extensively studied with respect to its association with substance use. Alcohol is used by some for its sedative, sleep‐promoting effects (Brower, Aldrich, Robinson, Zucker, & Greden, 2001). From this perspective, initiation of alcohol use may be an attempt to self‐medicate to reduce sleep latency. Moderate and high blood alcohol levels, however disturb the sleep architecture and chronic drinking has a long‐lasting negative impact on sleep quality (Chakravorty, Chaudhary, & Brower, 2016). In addition, chronic drinking‐induced sleep problems can persist into alcohol abstinence and predict relapse (Brower et al., 2001; Chakravorty et al., 2016). Hence, low sleep quality can be seen as a neurophysiological predictor of alcohol drinking in all situations ‐ initiation and maintenance of drinking, and renewed drinking after abstinence.

7.1. Animal studies

Animal research has yielded various insights into the effects of alcohol on sleep and sleep processes. In rodents, several weeks of forced alcohol exposure led to shortening of the circadian period (Rosenwasser, 2015). Free‐choice alcohol consumption had similar but less consistent effects on animal behaviour (Rosenwasser, 2015). It was found that, depending on the alcohol access schedule, gender, and genotype, tolerance to these alcohol‐induced changes did occur (for a review, see Rosenwasser, 2015). Furthermore, alterations in sleep architecture were observed in adult rats forcibly exposed to alcohol during adolescence (Criado, Wills, Walker, & Ehlers, 2008), suggesting that early life alcohol exposure induces long‐lasting changes in sleep quality. Taken together, these studies suggest that reliable long‐lasting changes in sleep occur when high blood alcohol levels (achieved by forced exposure) are maintained.

Aalto and Kiianmaa (1984) found that forced sleep deprivation in rats led to a transient increase in alcohol intake. However, other manipulations of the regular light–dark cycle (e.g. shorter/longer dark phase, constant light/dark phase, non‐24‐h periods) do not reveal a clear relationship between circadian disruptions and alcohol consumption (for a review, see Rosenwasser, 2015). These experiments also show the importance of individual differences; shifts in circadian cycle or other changes in lighting regimen can cause either an increase or decrease in alcohol consumption, depending on rodent strain (Becker et al., 2011; Rosenwasser, 2015). In several longitudinal rodent studies, changes in circadian activity were observed in response to alcohol deprivation and re‐exposure using long‐term free‐choice drinking (Vengeliene et al., 2013, 2015). At baseline, alcohol and water consumption followed a diurnal pattern; alcohol deprivation dramatically increased frequency of drinking upon re‐exposure, leading to a loss of normal diurnal drinking patterns (Vengeliene et al., 2013, 2015). Another intensive longitudinal data study found a decrease in circadian power, increased instability of circadian locomotor rhythms and increased ultradian rhythms during the first week of deprivation (Foo et al., 2017). Decreased circadian power persisted after re‐exposure of rats to alcohol, pointing at a key long‐term effect of withdrawal on circadian rhythms. Changes in arousal states can also be studied in terms of increased/decreased locomotor activity. For example, intermittent access to an alcohol sipper tube increases locomotor activity and consumption in rats (Tomie et al., 2005), intermittency‐induced high alcohol consumption is a basis of many animal drinking models (e.g. binge‐like ethanol drinking; Barkley‐Levenson & Crabbe, 2015).

7.2. Human studies

Human studies also show that chronic alcohol use can lead to reduced sleep quality (Chakravorty et al., 2016). Effects on sleep quality have been confirmed by ecological momentary assessment studies, which find that heavy drinking may increase sleep latency (Van Reen et al., 2016) and decrease sleep quality (Lydon et al., 2016). Another ecological momentary assessment study (Kuerbis et al., 2018) found that poor sleep quality predicted increased drinking in young adults but in older adults was associated with a reduced number of subsequent daily drinks, highlighting the importance of age differences (Table 1). Low arousal (specifically boredom) had no association with drinking in young adults but was reported to be a strong predictor for daily alcohol drinking in older adults (Kuerbis et al., 2018; Table 1). Another study in a predominantly young adult sample showed that higher mean daily arousal and lower arousal variability predicted the likelihood of alcohol use (Peacock et al., 2015).

7.3. Pharmacological interventions

Although sleep disturbances can be a marker of relapse to alcohol use (Brower et al., 2001), improving sleep quality by either behavioural or pharmacological means is not sufficient to reduce probability of alcohol relapse (Brower, 2015). In humans, insomnia is usually treated with sedating antidepressants, anticonvulsants, antipsychotics and benzodiazepines (Brower, 2015). Except for anticonvulsants, which have some efficacy in reducing drinking, insomnia medications appear ineffective in reducing alcohol consumption and preventing relapse (Witkiewitz et al., 2019). Recent animal research suggests using compounds targeting orexinergic and melatonergic systems. Orexin neurons coordinate arousal, energy homeostasis and regulate sleep/wakefulness cycle (Inutsuka & Yamanaka, 2013), while melatonin synchronizes behaviours and physiological functions of the body to the presence/absence of daylight (Buijs, van Eden, Goncharuk, & Kalsbeek, 2003). Compounds targeting these systems are shown to be effective in reducing alcohol seeking and relapse‐like drinking in rats (Lawrence, Cowen, Yang, Chen, & Oldfield, 2006; Vengeliene et al., 2015). Vengeliene et al. (2015) also suggested that restoring normal sleep architecture has the additional beneficial effect of reducing alcohol consumption. Orexin receptor antagonists and melatonin receptor agonists are approved for human use to treat insomnia and could be tested for their effectiveness in reducing probability of human alcohol consumption.

8. SUMMARY AND CONCLUSIONS

The impact of momentary triggers preceding substance consumption has long been recognized. However, only recently has it become possible to assess ongoing human behaviour, physiology, experiences and environmental factors in natural settings at high resolution, enabling evaluation of the multi‐dimensional dynamic changes of behaviours and other parameters at the intra‐individual level, over time. Crucial differences between human and animal studies are caused by evolutionary proximity of some species and lack of ecological validity in animal research. However, as technology improves and can be implemented in the automated smart home cage setting, it is becoming possible to acquire multi‐modal intensive longitudinal data at high resolutions in group‐housed rodents. Importantly, considering that animal studies in the past have used longitudinal study designs, efforts could be made to reanalyse existing data sets and search for mechanisms with a longitudinal/dynamic perspective. Analysis (and where possible, meta‐analysis) of these data is expected to yield new insights into underlying mechanisms of substance‐related behaviours.

This review showed the diversity of momentary triggers associated with increased probability of alcohol consumption. Different aspects of emotional, cognitive, motivational, social, and other dimensions appear to contribute to momentary trigger‐alcohol associations (Figure 2b). However alone, none of these factors seems to be a reliable predictor of alcohol relapse in humans, suggesting that combination of several triggers (e.g. stress‐/cue‐induced craving and rejection by close people) might be necessary to provoke alcohol relapse, accounting for inconsistencies in results.

Finally, the translation of preclinical experimental findings into successful clinical applications has lately become a serious concern in the drug discovery field. Multiple compounds to reduce stress‐/cue‐induced alcohol seeking and voluntary alcohol consumption have been suggested in animal studies, but translation success of these findings is low. Both animal and human studies support studying interindividual differences in drug response. Since aggregate data may give a misleading picture of what occurs at both the interindividual and intra‐individual levels, using a longitudinal, ambulatory assessment ‐style approach combined with biomedical intensive longitudinal data may help to more robustly identify and characterize common behavioural and physiological responses to challenging events between different species.

8.1. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander et al., 2019).

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

Vengeliene V, Foo JC, Kim J. Translational approach to understanding momentary factors associated with alcohol consumption. Br J Pharmacol. 2020;177:3878–3897. 10.1111/bph.15180