The role of stress-reactivity, stress-recovery and risky decision-making in psychosocial stress-induced alcohol consumption in social drinkers

Alcohol misuse is one of the leading avoidable risk factors for morbidity and mortality, and presents a significant global health challenge (Rehm et al. 2009). In the UK, for example, alcohol misuse is estimated to cost £3.5 billion per year to treat and manage (Williams et al. 2014). In some individuals, alcohol misuse can escalate into alcohol use disorder (AUD) (Skinner and Allen 1982). AUD and other substance use disorders (SUDs) have traditionally been described as diseases (i.e. the Brain Disease Model of Addiction [BDMA]; e.g. Volkow et al. 2016), in other words, as a chronic, relapsing disorders, characterised by withdrawal symptoms in the abstinence of alcohol, compulsive alcohol seeking, anhedonia and social/familial problems (American Psychiatric Association 2013). More recently, however, groups such as the Addiction Theory Network (ATN; Heather et al. 2018) have begun to question the BDMA and instead describe addiction as a ‘disorder of choice’ (e.g. Heather 2017; Heyman 2009). Regardless of which theory you subscribe to, the consequences of addiction are severe and often lead to adverse health outcomes. It has been postulated that 20% of people will meet the clinical characteristics for diagnosis with AUD in their lifetime (Hasin et al. 2007; Teesson et al. 2010). Considering only half the world’s population consume alcohol, this statistic is all the more alarming (WHO 2014). Despite the substantial current and projected prevalence of AUD, and decades of research into alcohol misuse and AUD, we are still unable to provide effective treatment programs for those affected, or put into place effective protective factors to predict and prevent AUD from developing in at-risk individuals (Harris and Koob 2017). Current treatments for AUD are inefficacious, and in order to find curative or preventative treatments or interventions we must identify and understand the risk factors which cause both the acquisition and maintenance of the disease.

One previously identified and well-established risk factor for alcohol misuse is psychological stress. Chronic alcohol use causes neuroadaptations in stress pathways, for example within the hypothalamic pituitary adrenocortical (HPA) and sympathetic adrenomedullary (SAM) axes (Sinha 2012). These adaptations have been identified by cortisol response dysregulation (Kreek and Koob 1998) and deficits in emotional regulation (Sinha 2001). As a result of these neuroadaptations, AUD patients commonly present with an increased stress-induced craving for alcohol. Nevertheless, it is still difficult to predict the latency to and the likelihood of relapse between individuals due to inconsistent findings between patient groups (Sinha et al. 2011). In addition, how adaptations to the stress pathways affect the early development of AUD, or alcohol misuse more generally, is not well characterised.

Incentive sensitisation theory (Berridge and Robinson 2016; Robinson and Berridge 1993, 2000, 2008) posits that addiction (e.g. AUD) results from of a gradual shift in the intensity of the incentive motivation to consume alcohol. Specifically, the ‘incentive salience’ of alcohol-related cues become heightened within at-risk populations. In turn, this increases the ‘incentive motivation’ to drink. In other words, alcohol-related cues grab the attention of the individual, increasing their craving and thus increases their likelihood of drinking. When considered in the context of stress as an additional risk factor, the increased likelihood of drinking during stress may be related to an increased psychological or physiological arousal, thus increasing the incentive salience of alcohol-associated cues (Kreek and Koob 1998). A recent study (Salemink et al. 2015) found links between neuroticism and risk for alcohol misuse. Their findings suggest that neuroticism in adultescents tends to act as a protective factor until the point of alcohol consumption; in other words, alcohol reduces the anxiety levels in neurotic adultescents, which in turn increases the incentive sensitisation of this reduced state of anxiety, thus increasing their risk of alcohol misuse. Under Incentive Sensitisation theory, it could be predicted that in healthy individuals, (1) stress will increase craving for alcohol, (2) craving will be positively correlated with stress-reactivity (e.g. physiological arousal caused by the stress), (3) craving will be the strongest predictor of drinking and (4) liking will not be correlated with craving.

An alternative theory suggests, in predisposed individuals, that addiction arises following a ‘shift’ from controlled, voluntary substance use to a habitual, compulsive (addicted) state. According to this theory, compulsive drug taking develops as a result of an imbalance between goal-directed (mediated by the dorsomedial striatum [DMS]) and habitual behaviour (mediated by the dorsolateral striatum [DLS]) (Everitt and Robbins 2005). It is suggested that addicted individuals have a more dominant DLS-based ‘habit-forming’ neural circuitry, and this appears to be linked strongly to trait impulsivity (Everitt and Robbins 2005). It has been suggested that traits such as impulsivity represent sub-clinical ‘intermediate’ phenotypes (termed endophenotypes; Gottesman and Gould 2003) that are state-independent (i.e. present in at-risk individuals even when the disease is not). Trait impulsivity, defined as the propensity to proceed without forethought and take risks, despite potential adverse consequences, is one such neurocognitive endophenotypes. In humans, trait impulsivity is linked both to addiction and to relapse; addicts have higher levels of impulsivity, as do their (non-addicted) first-degree relatives (Bowden-Jones et al. 2005; Lawrence et al. 2009). Impulsivity has been causally, in pre-clinical studies, linked to compulsive drug-seeking in rodent models (Belin et al. 2009). Furthermore, previous research from our laboratory suggested that risk-taking (an individual tendency closely related to impulsivity) was positively correlated with stress-induced craving; i.e. people with greater risk-taking tended to have a larger alcohol craving post-stress (Clay et al. 2018). Interestingly, stress has been shown to modulate the shift between goal-directed and habitual responding, potentially as a result of stress activating the amygdala, which then influences the shift to habitual behaviour (Koob 2008). Under this theory, it could be predicted that, in healthy individuals, higher rates of stress-induced drinking will be observed in impulsive individuals.

There were two aims for this study. The first was to test the hypothesis that acute psychosocial stress (Trier Social Stress Test [TSST]; Kirschbaum et al. 1993) would increase voluntary alcohol consumption in a sample of social drinkers. The second aim of this study was to test how well variability in stress-induced voluntary alcohol consumption was predicted by incentive sensitisation theory or neurocognitive endophenotype theory. Based on neurocognitive endophenotype theory, it would be predicted that, even in non-addicted individuals, early signs of a loss of control over drinking (i.e. drinking more when feeling stressed) might be more common in impulsive individuals. However, under incentive sensitisation theory, there is no reason to suggest that impulsive individuals would drink more than others. Towards our second aim, we tested the hypothesis that, in a sample of social drinkers, impulsive/risk-taking individuals would consume more alcohol following exposure to stress.

The primary aim of this study was to examine the effects of psychosocial stress on craving and voluntary alcohol consumption in healthy social drinkers. The study was developed with incentive sensitisation and neurocognitive endophenotypes as the theoretical framework and aimed to test predictions from both theories. Specifically, we tested the hypothesis that neurocognitive endophenotypes, such as impulsivity and risk-taking, affect stress-induced alcohol consumption in healthy social drinkers. First, we found that a mild psychosocial stressor increased voluntary alcohol consumption in healthy social drinkers. We also found that, although stress increased both craving and drinking, these two measures were statistically independent, with those that showed higher craving not necessarily consuming more alcohol. Instead, we found that risky decision making (IGT) and physiological stress-reactivity (SDNN and HR) were the most important predictors of alcohol consumption following psychosocial stress, and that those who drank more reported higher levels of subjective enjoyment of the drink.

Previous research has provided evidence that stress-reactivity plays a crucial role in the onset of relapse within abstinent alcoholics (Sinha et al. 2011; Sinha 2012), as well as an increase in craving (alcohol-seeking) and self-administration of alcohol in translational (animal) studies (Cleck and Blendy 2008; Koob and Kreek 2007; Koob et al. 2004; Rasmussen et al. 2006). Our data support and extend these findings, showing that participants with greater HRV reactivity and poor HR and HRV recovery following stress tended to drink more. This is, to the best of our knowledge, the first empirical demonstration of how individual differences in stress reactivity and recovery affect voluntary drinking. This finding is supported by the increasing body of work focusing on vagal dysfunction, and its inhibitory effect on the management of allostatic load (Ingjaldsson et al. 2003; Rechlin et al. 1996; Shively et al. 2007; Thayer and Sternberg 2006), suggesting that those with poor vagal tone may be at increased risk of stress-induced craving for alcohol (Koob and Le Moal 2001). Further assessments of stress-reactivity (i.e. sC and sAA) did not reveal any significant effects of the TSST. Other studies have consistently found the TSST to increase neuroendocrinal responses to stress (Het et al. 2009; Kelly et al. 2008; Kirschbaum et al. 1993; Oswald et al. 2006). The blunted sC and sAA responses observed in our data could have been the result of the omission of the mental arithmetic test from our TSST procedure (on the basis of participant feedback acquired during preliminary studies). The reduction of the TSST running time from 10 to 5 min may not have given adequate time for cortisol and alpha-amylase to be detected through saliva. Regardless of this, the shorted TSST used in this study still provoked a significant increase in HR (both during anticipation and during the speech) and increased both craving and alcohol consumption.

We found that risky decision-making (in the IGT) was a significant predictor of alcohol consumption. Risky or impulsive decision-making has been previously associated with AUD, with abstinent alcoholics in particular showing deficits in their decision making ability (Fein et al. 2004). Risky-decision making is also observed in those who misuse other substances (Barry and Petry 2008), including cocaine (Bolla et al. 2003; Verdejo-Garcia et al. 2007) and cannabis (Hermann et al. 2009; Verdejo-Garcia et al. 2007). Furthermore, our previous research has shown that risk-taking was positively correlated with alcohol craving (Clay et al. 2018) and others have demonstrated links to reported alcohol use in social drinkers (Fernie et al. 2010). We did not find that any other measures of inhibitory control (stop-signal task, BIS) were predictive of voluntary alcohol consumption. McGrath, Jones and Field (2016) examined the effects of a psychosocial stressor on voluntary drinking in a semi-naturalistic ‘mock bar’ environment. They hypothesised that alcohol consumption in this setting would be increased following stress owing to changes in inhibitory control. Similar to our study, they found that stress increased drinking. Also similar to our findings, they did not find any evidence that impulse control (via the stop-signal task) was predictive of drinking. However, here we observed that one measure of inhibitory control, risky-decision making, was predictive of drinking. Collectively, these data suggest that stress-induced lapses in inhibitory control are not sufficient to explain variability in drinking, but instead lend support to the theory that subtle neurocognitive endophenotypes may be more important in predicting the move from controlled to uncontrolled drinking.

Our data provides evidence that explicit craving was increased following a psychosocial stressor in a sample of healthy social drinkers, replicating our prior work (Clay et al. 2018). Interestingly, alcohol craving was not found to be a significant predictor of alcohol consumption, although explicit craving (DAQ) change was greater in the stress group. The literature suggests that craving is the prominent risk factor in the onset of relapse, and there is a large surrounding research output in support of this idea (e.g. Litt et al. 2000; Robinson and Berridge 1993; Schneekloth et al. 2012; Sinha et al. 2011). The fact our data did not substantiate this could be due to craving only being an important predictor of alcohol consumption in those already addicted, i.e. alcoholics. Consequently, our data suggests that decision-making is of a higher importance when predicting alcohol consumption in healthy social drinkers. If this is the case, underlying personality traits could be key in predicting those most at risk of moving from a social (habitual) state to a dangerous (addicted) one. It should be considered that had we used different cues that grabbed the attention or a different assessment of implicit craving, there may have been a different outcome in terms of implicit craving change within social drinkers. Clinical research indicates that both addicted cocaine and alcohol users tend to show an increased craving following presentation of stress-inducing imagery (Sinha et al. 2011). Incentive sensitisation theory (e.g. Robinson and Berridge 1993) predicts that the physiological response to stress acts as a cue that induces incentive motivation to take the drug; in other words, incentive sensitisation predicts that stress will increase craving. However, incentive sensitisation theory suggests that craving will also be increased by a number of environmental cues (drug stimuli, for example), and this has been demonstrated in many studies (e.g. Field and Powell 2007; Field and Quigley 2009; Wiers et al. 2006). For future studies, it would be useful to include a number of additional cues (e.g. drug-related visual stimuli, or olfactory/gustatory cues) to examine their role of impulsivity/risk-taking on craving/consumption of alcohol. It may be that the stressor used here was insufficient to induce incentive motivation to consume alcohol, thus explaining why craving did not predict drinking. Finally, although we found differences in explicit craving, we did not find differences in our measure of implicit craving. It may be that the AAT was not sensitive enough to detect changes in craving in our healthy sample. Others (e.g. Field and Powell 2007) have used attentional bias protocols as a measure of implicit craving, and this may be more sensitive. The AAT was also not intercorrelated with our assessment of explicit craving (DAQ), suggesting that the AAT may not have good psychometric properties for the assessment of implicit craving of alcohol. In future studies, other measures of implicit craving, such as the attentional bias measures (Field and Powell 2007), could be included in addition to the AAT. However, the attentional bias task take a long time to carry out, and in the current context (i.e. following stress), this is not ideal. Future research should strive to validate a novel, less temporally limited, assessment of implicit craving.

An additional note from our data was that ‘drink enjoyment’ was not significantly correlated with craving. This finding is consistent with the predictions of incentive sensitisation theory (Robinson and Berridge 1993), which states that that incentive motivation to drink (wanting) and subjective enjoyment (liking) are independent. However, interestingly, we observed that there was a strong correlation between subjective ‘liking’ of alcohol and voluntary alcohol consumption. This finding further suggests that, in healthy social drinkers that are primed with an acute risk-factor for AUD (psychosocial stress), the predictions of incentive sensitisation theory do not stand. Instead, those at highest risk of stress-induced alcohol consumption did appear to be those that are more stress reactive, have poorer vagal tone (lower stress-recovery) and are higher in risky decision making. This may have significant implications in the search for sub-clinical early identification of those at risk of AUD.

There were several limitations to the present study that should be considered. First, the generalisability of the findings is limited by our sample, who were students and staff at the University of Portsmouth and may not have been truly representative of the general population. Second, we used a modified version of the TSST, which did not induce significant changes in either salivary cortisol or saliva alpha amylase, as shown by numerous previous studies. However, we did observe significant increases in HR (and in HRV parameters) following the modified version of the test, suggesting that it did induce stress in the participants. It may be that a longer version of the TSST would have increased craving and this may have been linked to drinking. A further limitation of this study was the fact that both stress and control groups completed cognitive trials. Dickerson and Kemeny (2004) suggest that cognitive tasks stimulate the HPA axis enough to detect changes cortisol levels. Therefore, the use of cognitive tasks as a means of quantification within stress research could act as a confounding variable. Ideally, research in this area should aim to assess stress variability and individual differences on separate occasions. Unfortunately, due the nature of this study, mitigating against such a confound in this way was not an option due to the need to measure craving/consumption following stress. In the same vein, the IGT has received some criticism of its validity (Dunn et al. 2006; Steingroevera et al. 2013). Bechara et al. (1994) make several assumptions about the decision-making processes in healthy participants: ‘Healthy participants learn to prefer the good options over the bad options, healthy participants show homogeneous choice behaviour and healthy participants first explore the different options and then exploit the most profitable ones’. A recent study (Steingroevera et al. 2013) suggests instead that ‘Healthy participants often prefer decks with infrequent losses, healthy participants show idiosyncratic choice behaviour and healthy participants do not show a systematic decrease in the number of switches across trials’. These findings suggest that decision-making, when assessed using the IGT, in healthy participants is heterogeneous. Therefore, future research using should err on the side of caution when using the IGT. Nevertheless, our data suggests that this heterogeneity can predict variability in alcohol consumption amongst social drinkers. Additionally, both groups of participants drank less than half of the 12 available drinks, meaning that on average < 1 unit of alcohol was consumed, which may indicate an unsuccessful stress manipulation following the modified TSST. However, as previously stated, we believe that the stress manipulation was successful due to the significant HR group × time interaction and that the insignificant sC and sAA responses were due to shortening of the TSST. Furthermore, our method of assessing drink enjoyment, i.e. via 15-point Likert scale, may have limited subject interest. Finally, the alcohol used in the voluntary drinking procedure was vodka mixed with a mixer of choice. It may be that if we had given a greater range of drinks as reinforcers, this would have increased the amount consumed in some participants. We chose to use vodka and mixer owing to the very precise amount (and v/v of alcohol) this afforded us. In future studies, we may include a choice of beer, wine and spirit to examine if there are any differences observed. However, subjectively, no participants voiced this as a reason for not consuming the alcohol, so we do not think this represented significant confound.

In conclusion, we have demonstrated that psychosocial stress causes an increase in voluntary alcohol consumption. We also demonstrated, for the first time, that the number of drinks consumed was predicted by greater physiological reactivity and slower physiological recovery following stress, and by risky decision making. Collectively, our findings may provide a useful translational framework through which we can further study the early risk factors that predict the shift from controlled recreational drinking and alcohol misuse, leading (in some) to AUD. In particular, the role of variation in personality traits and physiological reactivity that predict cue-induced alcohol consumption needs to be further studied in order that we can be in a stronger position to identify those at risk of developing AUD in the future. A useful target of study in the first instance may be role of variation in efficacy of neural circuits that underlie both stress-reactivity and impulse control (i.e. cortico-amygdala-striatal projections; Koob 2008) in cue-induced drinking.