Insights into overlapping brain networks for anxiety and alcohol use disorders

Abstract

Alcohol use disorder (AUD) is a debilitating condition affecting over 30 million Americans. AUD commonly co-occurs with other disorders, like other substance use disorders, trauma-related disorders, and anxiety disorders. Of the numerous co-occurring disorders, anxiety disorders are the most pervasive: anxiety disorders serve as a risk factor for developing AUD, emerge as co-occurring disorders that maintain alcohol drinking, and impede effectiveness of treatments for AUD. Anxiety, therefore, shapes the development, course, and treatment of AUDs. AUDs can also increase anxiety, suggesting a complex, bidirectional relation between alcohol use and anxiety. The intersection of AUDs and anxiety is also supported by their overlapping neural circuits, specifically neural circuits involved in stress responding, reward processing, and cognitive control. The current review highlights findings from several decades of research on how anxiety impacts the brain and treatment outcomes in AUDs. We also provide important considerations for future research, with the goal of reducing the shame and burden of alcohol use for individuals with AUD and their families.

I. Introduction

Alcohol use disorder (AUD) is a debilitating, chronic condition that is characterized by excessive or compulsive use of alcohol despite social, financial, or professional consequences. Prevalence rates of AUD have doubled in the past five years, with nearly 30 million adults qualifying for an AUD in 2022 (SAMHSA, 2022). Alcohol use disorder remains an elusive and costly disorder in the US, with researchers exerting tremendous effort to keep pace with understanding the development of AUD and potential treatments. The complexity of alcohol use extends further, as chronic alcohol use is closely connected to disorders like other substance use disorders, depressive disorders, and anxiety disorders (Grant et al., 2015). Of the co-occurring disorders, anxiety disorders represent the most pervasive: anxiety disorders serve as a risk factor for developing AUD, emerge as co-occurring disorders that maintain alcohol drinking, and impede effectiveness of treatments for AUD. Thus, identifying the role of anxiety in AUD serves as an important step towards identifying novel preventions and treatments.

A leading model of addiction recognizes negative affect (i.e., anxiety) as a cornerstone of AUD (Koob & Volkow, 2010, 2016). To date, most research in AUDs has focused on reward, craving, and cognition in chronic alcohol use, leading to important discoveries about the neurobiology and psychology of AUD. Thus, there remains a substantial gap in our knowledge about how acute and chronic anxiety may impact AUDs. Greater understanding of the role of anxiety in the development, maintenance, and relapse of alcohol use has great potential to inform the development of preventions and interventions.

The goal of this review is to characterize the role of anxiety in AUD, with a focus on how anxiety-related factors contribute to the development of AUD, maintenance of alcohol use, and relapse following treatment. This review will focus on the impact of anxiety and AUD on brain function in humans. The first section introduces how anxiety and stress impact the development, prevalence, and maintenance of alcohol use. The second section summarizes the distinct, but overlapping, neural networks involved in anxiety and AUD. The third section reviews current neuroimaging literature that investigates how anxiety and AUD shape brain function. Finally, the last section discusses how anxiety may impede the effectiveness of alcohol treatment. This review is guided by three primary principles: (1) anxiety shapes the development and maintenance of chronic alcohol use, (2) anxiety and AUD have significant overlap in their underlying neural networks, (3) anxiety should be a clinical focal point in AUD recovery.

II. How anxiety shapes alcohol use disorder

Anxiety contributes to the development and course of AUD. The impact of anxiety on AUD has been characterized by numerous longitudinal studies. The longitudinal studies we highlight below have three primary foci: (1) developmental impact of anxiety on AUD, (2) anxiety that co-occurs with AUD, and (3) anxiety’s role in maintaining alcohol use and impeding alcohol abstinence.

Developmental impact of anxiety on alcohol use

Adolescence.

Adolescence is a critical period of development when anxiety disorders are emerging, but is a period before AUDs typically emerge, making it ideal for studying longitudinal relations between anxiety and alcohol use. Klein and colleagues (2022) investigated whether lifetime or current adolescent anxiety predicted lifetime substance experimentation. Based on an initial wave of data collected from the Adolescent Brain and Cognitive Development dataset (n = 11,785), the authors found that children with an anxiety disorder were 31% more likely to have experimented with alcohol. Similarly, McCabe and colleagues (2023) investigated the relation between negative affect symptoms (anxiety) and binge drinking over five years using data from the National Consortium on Alcohol and Neurodevelopment in Adolescence-Adulthood (n = 831). Girls reported higher anxiety across the study period and had significant increases in binge drinking at follow-up visits compared with boys, but anxiety was not a significant predictor of binge drinking. Findings from these two studies suggest that adolescent anxiety predicted greater alcohol experimentation, but not binge drinking. One potential explanation for these findings is that anxious adolescents might be more cautious about alcohol use, such that they will use alcohol but not engage in binge drinking, whereas adolescents without anxiety may be more likely to more extraverted and risk-takers--traits associated with binge drinking.

Emerging young adults.

Emerging adulthood is an important stage of development for alcohol use—the average age of first drink is 18 years, consistent with the transition to college and other major life events (Maggs & Schulenberg, 2004; SAMHSA, 2023). Zimmerman et al. (2003) prospectively followed emerging adults in the Early Developmental Stages of Psychopathology study (n = 3,021). Participants with an anxiety disorder at baseline had higher levels of regular or hazardous alcohol use four years later, which was largely driven by social anxiety disorder, panic disorder, and agoraphobia. Fröjd and colleagues (2011) prospectively followed adolescents into adulthood in the Adolescent Mental Health Cohort Study (n = 2,070). Levels of general anxiety at baseline, but not social anxiety, were associated with a three-fold higher incidence of frequent alcohol use and drunkenness. At the two-year follow-up, baseline levels of general anxiety predicted a two-fold higher incidence of frequent alcohol use and drunkenness. Wolitzy-Taylor and colleagues (2012) utilized data from the Northwestern-UCLA Youth Emotion Project to determine whether adolescent emotional disorders predicted alcohol use, as well as whether alcohol use predicted adolescent emotional disorders (n = 627). Young adults with a baseline anxiety disorder were nearly three times more likely to develop an AUD, while baseline AUD did not predict future anxiety disorders. These data in emerging young adults give us greater insight into the direction of interaction between anxiety and alcohol—anxiety predicts alcohol use and AUD, while alcohol use and AUD do not predict anxiety disorders.

Adulthood.

Studies of adulthood offer a unique view on both lifetime history and current anxiety and alcohol use. Marquenie et al. (2006) aimed to determine the order of onset of anxiety disorders and AUD through the Netherlands Mental Health Survey and Incidence Study, a retrospective and prospective study conducted over three years. Participants were grouped according to baseline diagnoses (n = 7,076): anxiety disorder with an AUD (1.4%), anxiety disorder without an AUD (14.7%), AUD without an anxiety disorder (3.5%), and no anxiety disorder or AUD (80.4%). Anxiety disorders preceded development of AUD in 68% of people with both anxiety and AUD. People with a current anxiety disorder at baseline were over two times more likely to develop AUD at a follow-up assessment. AUD preceded anxiety disorders in 24% of people, while anxiety disorders and AUD co-emerged in 8%. Torvik and colleagues (2019) investigated whether anxiety disorders predicted development of an AUD in a twin cohort study (n = 2,284). Social anxiety disorder at baseline predicted the development of an AUD ten years later and social anxiety disorder was a stronger predictor of AUD than any other type of anxiety disorder. AUD at baseline did not predict development of a social anxiety disorder but moderately predicted other anxiety disorders at follow-up. Like studies in emerging adulthood, studies in adults suggests that anxiety disorders largely precede the development of alcohol use.

Together, these findings suggest that anxiety disorders largely precede the development of AUD. Interestingly, the data are most consistent in studies of emerging young adults and adults, with mixed findings in adolescence. While these differences may be the result of methodological variations, such as questionnaires vs. clinical interviews, they might also be the result of relatively lower rates of anxiety disorders and drinking during adolescence. Of interest, both generalized anxiety disorders and social anxiety disorders were predictors of later AUD, consistent with population-level studies investigating co-occurrence of anxiety disorder types and AUD (e.g., Grant et al., 2015).

Anxiety co-occurring with alcohol use

Prevalence of co-occurring anxiety disorders and AUD.

Recognition of anxiety as a common co-occurring disorder existed colloquially, but the relation between anxiety and AUD was only demonstrated with standard diagnostic methods 25 years ago using the DSM-III. Kessler et al. (1997) and Kushner et al. (1999) used epidemiological datasets to assess co-occurring disorders across the DSM-III, revealing a two-to-fivefold increase in the likelihood of having an anxiety disorder and an AUD. Additional epidemiological studies confirmed the co-occurrence of anxiety and AUD using later versions of the DSM: DSM-IV (Hasin et al., 2007; Schneier et al., 2010) and DSM-5 (Grant et al., 2015). These studies found that individuals with an anxiety disorder are four times more likely to have a co-occurring AUD (Hasin et al., 2007; Kessler et al., 1997). Notably, criteria for both anxiety disorders and AUDs changed across DSM-III, IV, and 5, which may influence the increase in co-occurrence rates—or suggests increased accuracy of capturing complex disorders. Despite changes across diagnostic criteria, co-occurrence of anxiety disorders and AUD is now a well-established association.

The most recent National Epidemiologic Survey on Alcohol and Related Conditions collected data from over 36,000 US adults on past-year and lifetime prevalence of AUD and psychiatric disorders using DSM-5 criteria (Grant et al., 2015). Individuals with any lifetime AUD diagnosis were 30% more likely to have an anxiety disorder, while those with severe lifetime AUD were 40% more likely to have an anxiety disorder. Controlling for sociodemographic variables (e.g., race, sex, income level) weakened the association between alcohol use and anxiety, ultimately suggesting important underlying factors that increase the likelihood of developing both AUD and anxiety disorders. Interestingly, controlling for other co-occurring disorders did not impact the relation between AUD and anxiety disorders—suggesting a specific link between anxiety and AUD.

Anxiety maintaining alcohol use.

In addition to anxiety co-occurring with AUD, anxiety can impact how and why an individual drinks. Motivations for alcohol use generally reflect positive and negative reinforcement: initially, individuals may drink for the rewarding properties of alcohol (positive reinforcement); however, as chronic drinking develops, individuals may drink to avoid negative emotions and withdrawal symptoms (negative reinforcement) (Koob, 2021). Considering that anxiety disorders often precede development of AUD, it is posited that those with anxiety disorders use alcohol for its negative reinforcing effects and/or transition faster to using alcohol for its negatively reinforcing properties. Several studies in humans support the negative reinforcing properties of alcohol in anxiety disorders. Kushner and colleagues (2011) found that individuals with an anxiety disorder transitioned more rapidly from first time regular drinking and getting drunk to severe AUD, compared with those without an anxiety disorder. Menary and colleagues (2011) found that among adults with an anxiety disorder, approximately 20% report self-medicating with alcohol. Adults who self-medicate with alcohol also drink significantly more than those who do not self-medicate, and those who self-medicate with alcohol are more likely to develop an AUD (Menary et al., 2011). Importantly, self-medication with alcohol in people with anxiety disorders also leads to greater utilization of mental health services and lower quality of life (Robinson et al., 2009). While not all people with anxiety engage in the negative reinforcing aspects of alcohol, motivations for alcohol drinking may be an important factor when looking toward improving treatments for AUD.

Anxiety development during alcohol withdrawal and abstinence.

Anxiety is common during alcohol withdrawal, as cessation of alcohol use results in increased stress reactivity and the emergence of anxiety symptoms (Koob, 2013; Koob & Volkow, 2010, 2016). During acute withdrawal, higher drinking frequency in adults with AUD predicts higher anxiety, craving, and cortisol levels (McCaul et al., 2017). In addition, adults with AUD experiencing alcohol withdrawal exhibit higher anxiety levels over a three-week study period, compared with healthy controls (Petit et al., 2020). Individuals with an AUD who have a co-occurring anxiety disorder also experience more intense withdrawal symptoms than those without a co-occurring anxiety disorder (Johnston et al., 1991). Beyond acute withdrawal symptoms, anxiety generally decreases during early abstinence (2–4 weeks; Zywiak et al., 1996). However, there is significant heterogeneity, with many people continuing to experience anxiety months to years into abstinence (Heilig et al., 2010). For example, individuals with an AUD and co-occurring anxiety disorder have higher anxiety and depression scores after 18 months of abstinence (Wolitzky-Taylor et al., 2015). Finally, and most importantly, several studies have shown that anxiety contributes to high relapse rates; individuals with high trait anxiety or an anxiety disorder are 2–5 times more likely to relapse within one year of abstinence (e.g., Schellekens et al., 2015; Willinger et al., 2002; Zywiak et al., 1996). The relation between anxiety and alcohol use continues during alcohol abstinence—anxiety increases during acute withdrawal and early alcohol abstinence.

Conclusion

Together, these data suggest anxiety shapes the development, prevalence, and maintenance of AUD. Findings from longitudinal studies reveal that anxiety disorders increase risk for developing an AUD, which is consistent across several diagnostic measures and study designs. Epidemiological studies also show a high co-occurrence between anxiety disorders and AUDs. Finally, anxiety plays an important role in the maintenance of alcohol use and symptoms experienced during alcohol withdrawal abstinence. Understanding alcohol use through the lens of anxiety reveals a complex and cyclical relation—anxiety disorders are a significant predictor of later alcohol use, and alcohol use can increase anxiety symptoms—suggesting an important interaction between alcohol and anxiety. To better understand the relation between alcohol use and anxiety, and to improve treatment for AUDs, we need to investigate their intersection in the brain.

III. Neural networks involved in anxiety and alcohol use disorder

Studies investigating the neural networks underlying anxiety disorders or AUD suggest distinct, yet overlapping, neural networks. Comparing the anxiety and AUD networks highlights overlap in both stress and reward brain systems. We previously proposed an intersection of anxiety and addiction, based on evidence that there is substantial overlap between anxiety and addiction neurocircuitry (Avery et al., 2016). Here, we expand that review to discuss major overlaps between anxiety and AUD neural networks, including those related to reward and cognition.

Brain networks in anxiety

Anxiety is characterized by hypervigilance related to distant, vague, or unpredictable threat ( T. A.Brown et al., 1998; Zinbarg & Barlow, 1996). Processing and coordinating responses to unpredictable threats involves a broad network of brain regions, which including the bed nucleus of the stria terminalis (BNST), amygdala, hippocampus, hypothalamus, insula, ventromedial prefrontal cortex (vmPFC), dorsomedial and dorsolateral prefrontal cortex (PFC), and mid-cingulate and anterior cingulate cortex (ACC) (Abend, 2023; Chavanne & Robinson, 2021). Within this network, different brain regions coordinate anticipation, responses, and processing of unpredictable threats (Abend, 2023; Y. Somerville & Abend, 2024). Many studies have investigated brain differences between anxious and healthy people; we review a select number of meta- and mega-analyses to provide a broad understanding of anxiety’s impact on the brain.

Structural brain measures.

Measures of brain structure, like gray matter volume, can provide clues about structural abnormalities that may be associated with anxiety. The Enhancing NeuroImaging Genetics through Meta-Analysis (ENIGMA)- Anxiety Working Group reported a mega-analysis of structural neuroimaging data (Harrewijn et al., 2021). The sample (n = 4,394) included people with a lifetime or current generalized anxiety disorder (25%) and healthy controls. The mega-analysis revealed a group x sex interaction, such that men with a generalized anxiety disorder had a larger gray matter volume in the ventral diencephalon than control men; no differences were detected between women. The ventral diencephalon is comprised of multiple small brain regions, including the hypothalamus, lateral and medial geniculate, and subthalamic nuclei, that are involved in reward and coordinating responses to anxiety. An additional mega-analysis from the ENIGMA Anxiety Working Group (n = 3,890) compared people with a lifetime or current social anxiety disorder (29%) and healthy controls (Groenewold et al., 2023). The mega-analysis focused on volumetric differences between subcortical structures, revealing that individuals with social anxiety disorder have smaller putamen. The putamen is part of the dorsal striatum, which helps coordinate reward-related learning and movements associated with the reward learning. A post-hoc analysis by age (adolescents vs. adults) revealed that adults with social anxiety drove the effects in the putamen; adolescents showed no significant differences in subcortical volumes. These structural brain findings were intriguing because the differences did not emerge in the putative anxiety brain network, but instead were in brain regions involved in processing stimuli and coordinating responses to the environment.

Resting-state functional connectivity measures.

Resting state functional connectivity provides a measure of correlated activity between brain regions in the absence of a task and is thought to provide a measure of the intrinsic structure of brain networks. Zugman and colleagues (2023) conducted a meta-analysis on 14 resting-state connectivity studies that used the amygdala as a seed region, based on the amygdala’s role in fear and anxiety disorders (n = 572 subjects, 51% anxiety disorder). The primary finding revealed that people with an anxiety disorder had lower connectivity between the amygdala and the medial PFC, cingulate gyrus, and ACC. The authors also did an additional meta-analysis of 29 resting-state connectivity studies using any brain region as a seed, revealing no significant connectivity differences between groups (n = 1,506, 49% anxiety disorder). Xu et al. (2019) conducted a meta-analysis with 20 studies comparing healthy controls to individuals with an anxiety disorder (n = 974, 48% anxiety disorder). There were multiple group differences in resting-state functional connectivity across the brain. Relevant to anxiety-related brain regions, individuals with an anxiety disorder exhibited weaker connectivity between the following regions: cerebellum-dorsolateral PFC, posterior cingulate cortex-dorsolateral PFC, anterior insula-ACC, and anterior insula-dorsal ACC. Overall, resting-state functional connectivity studies reveal weaker connectivity between corticolimbic regions, which may reflect impaired ability to regulate fear and anxiety.

Task-based measures.

Task-based neuroimaging studies are an important avenue for understanding how people process and respond to stimuli. While responses to tasks likely result from differences in brain structure and resting state connectivity, tasks provide a direct probe of psychological processing and thus may reveal additional insights. A seminal meta-analysis from Etkin & Wager (2007) revealed that anxiety disorders were associated with higher activation of the amygdala, insula, hippocampus, and mid-cingulate cortex during negative emotional conditions. A meta-analysis by McTeague et al. (2020) similarly showed that people with anxiety disorders exhibited greater activation in the amygdala and hippocampus during emotional tasks.

In addition to emotion-processing tasks, tasks that provoke anxiety in humans (e.g., threat of shock, N-P-U tasks, CO₂ challenges) consistently recruit the BNST, amygdala, thalamus, hippocampus, insula, dorsal ACC, and dorsomedial PFC—all displaying greater activation in those with higher anxiety (e.g., Geng et al., 2018; Radoman et al., 2021; Somerville et al., 2010). These studies also report stronger functional connectivity from the amygdala to the dorsal ACC and dorsomedial PFC, as well as stronger connectivity from the dorsomedial PFC to the insula and hippocampus (e.g., Geng et al., 2018). Taken together, these findings suggest that altered anxiety network interactions underly anxious processing.

Anxiety disorders are also characterized by avoidance behaviors related to potential threat or cues that trigger anxiety. Therefore, additional processes are at play when the brain interprets and responds to its environment, like evaluation of risk versus reward related to approaching or avoiding anxiogenic stimuli. Risk is over-estimated in anxiety, causing an ‘inflation’ of potential threat (Grupe & Nitschke, 2013). Tasks that tap into risk versus reward evaluation include monetary incentive delay tasks and approach-avoidance conflict tasks. Risk versus reward evaluations require prefrontal cortex integration (dorsomedial PFC, dorsal ACC) with reward-processing regions, like the nucleus accumbens, caudate, and putamen; anxious individuals display less activation of reward-related regions (nucleus accumbens, caudate, putamen) when receiving a reward (e.g., Auerbach et al., 2021). In approach-avoidance conflict tasks, anxious individuals exhibit lower medial PFC activation (Aupperle & Paulus, 2010). Overall, reward-related processing is altered in anxiety disorders, which may lead people with anxiety to be more susceptible to seeking reward through alcohol.

In summary, weaker corticolimbic connectivity at rest, coupled with stronger connectivity during tasks, suggests that there may be compensatory mechanisms at large. Specifically, weaker connectivity at rest may be due to poor baseline emotion regulation, while stronger connectivity during tasks may result from over-compensation to regulate emotion.

Brain networks in alcohol use disorder

Chronic alcohol use results in substantial, widespread impact across the brain. Alcohol is a neurotoxin that degrades gray and white matter structure and impairs coordination between brain regions associated with cognition (lateral and medial PFC, hippocampus) and emotion processing and regulation (BNST, amygdala, striatum, hypothalamus) (Koob & Volkow, 2010, 2016). Many studies have investigated brain differences between people with AUDs and controls; we review a select number of seminal studies and meta-analyses to provide a broad understanding of alcohol’s impact on the brain.

Structural brain measures.

Chronic alcohol use results in widespread structural changes in the brain, which was demonstrated in some of the earliest studies on alcohol use in humans (e.g., Courville, 1955; Jernigan et al., 1982; Pfefferbaum et al., 1988). Xiao and colleagues (2015) conducted a meta-analysis of nine studies assessing gray matter differences between people with AUD and controls (n = 655, 45% AUD). People with AUD had significantly lower gray matter in the dorsal and rostral ACC, posterior ACC, striatum, and insula, compared with controls. Yang and colleagues (2016) conducted a meta-analysis of 12 studies that revealed more widespread gray matter changes (n = 931, 47% AUD). AUD was associated with lower gray matter volume in the insula, superior temporal gyrus, striatum, precentral gyrus, dorsolateral PFC, dorsal ACC, thalamus, and hippocampus, compared with controls. Recently, Li and colleagues (2021) conducted a meta-analysis of 23 studies comparing people with AUD with controls (n = 1,724, 49% AUD) revealing lower gray matter volume of the dorsolateral PFC, insula, and dorsal ACC in people with AUD. These findings suggest that chronic alcohol use is associated with consistent gray matter loss in regions associated with emotion processing and regulation.

Resting-state functional connectivity measures.

Resting-state functional connectivity is also altered in AUD. Rodríguez et al. (2024) systematically reviewed studies that analyzed resting-state data using seed-based functional connectivity (n = 7). Overall, these studies found that people with AUD, compared with controls, exhibited stronger connectivity of the anterior insula with the hippocampus and medial orbitofrontal cortex, and between the posterior and anterior cingulate cortices to the inferior frontal gyrus and supramarginal gyrus. In addition, AUD was associated with weaker connectivity of the anterior insula with the superior parietal lobule, the thalamus with the anterior insula, and the ACC with the thalamus, medial frontal gyrus, and orbitofrontal cortex. Recently, Bottino and colleagues (2025) conducted a systematic review of 39 studies of AUD using a resting-state functional connectivity relevance matrix that quantified the frequency of brain regions associated with significant results in the literature. Overall, AUD was positively associated with connectivity between the insula, ACC, and superior temporal gyrus, and between the inferior frontal gyrus and middle temporal pole. AUD was negatively associated with a connection between the superior temporal cortex and lingual gyrus, and between the middle temporal cortex and precuneus. The brain regions most associated with significant results in AUD were corticolimbic brain regions: the medial PFC, dorsolateral PFC, putamen, ACC (subgenual, pregenual, supracallosal), and insula. Overall, these findings suggest consistent differences in people with AUD across studies of brain structure and resting state functional connectivity, especially in corticolimbic brain regions relevant for anxiety processing and responding.

Task-based measures.

The most common tasks for determining the impact of alcohol on brain function relate to the three-stage models of AUD, which highlights the plasticity of brain networks continually adapting and shifting with continued alcohol use. The three-stage model outlines the stages of addiction: binge/intoxication, withdrawal/negative affect, and preoccupation/anticipation. More recently, the Addictions Neuroclinical Assessment, building on the National Institute of Mental Health’s Research Domain Criteria framework, has suggested measurable domains for each of the three stages: incentive salience, negative emotionality, and executive function (Cuthbert & Insel, 2013; Kwako et al., 2016, 2017). Binge/intoxication (incentive salience) is characterized by increased reactivity to alcohol-related cues and impaired reward learning—higher striatum and thalamus activation are reported and are associated with a decreased threshold for reward. Withdrawal/negative affect (negative emotionality) is characterized by negative affect and stress, resulting from the cessation of alcohol use—higher activation of the BNST, amygdala, nucleus accumbens, and hypothalamus are reported during alcohol withdrawal and abstinence, and are associated with a decreased threshold for stress responses. Preoccupation/anticipation (executive function) is characterized by alcohol craving in response to stress- and alcohol-related cues and deficits in executive functioning—higher insula activation is combined with lower medial PFC activation and these differences are associated with increased craving and decreased executive functioning.

The earliest task-based fMRI studies of AUD focused on assessing brain alterations to alcohol cues. These studies found higher activation of the striatum, dorsal ACC, and medial and lateral PFC to alcohol cues in individuals with AUD, compared with controls (Braus et al., 2001; George et al., 2001; Grüsser et al., 2004; Modell & Mountz, 1995). Additionally, these early studies found a relation between alcohol cue-induced activity and relapse in people receiving treatment, where higher activation of the striatum and PFC in response to alcohol cues predicted relapse three months later (Braus et al., 2001; Grüsser et al., 2004). A recent meta-analysis from Zeng and colleagues (2021) investigated 21 studies of alcohol cue reactivity in people with an AUD (n = 795), showing that people with AUD had higher cue-related activation of the insula and posterior cingulate cortex. The authors also investigated 17 studies comparing people with AUD with controls (n = 817, 56% AUD), revealing higher activation of the dorsal ACC, mid-cingulate cortex, and the ventromedial PFC in people with an AUD. In addition, in 14 studies of people receiving treatment for an AUD (n = 338), following treatment (relative to placebo or pre-treatment) there was higher activation of the precentral gyrus and lower activation of the rostral medial PFC, dorsolateral PFC, caudate, and insula.

Studies of negative emotionality have largely used emotional face tasks to probe emotion processing. Both acute and chronic alcohol use is associated with lesser activation of the amygdala, dorsal ACC, and hippocampus to both positive and negative emotional faces (Gilman et al., 2012; Marinkovic et al., 2009; Salloum et al., 2007). Further, chronic alcohol use impacts task-based functional connectivity when responding to emotional stimuli, such that individuals with AUD display less connectivity between the amygdala and dorsal ACC and BNST, and between the insula and ventromedial PFC, ventrolateral PFC, and medial orbitofrontal cortex (Kienast et al., 2013; O’Daly et al., 2012). To our knowledge, there are no meta-analyses that investigate studies of emotion processing in AUD. However, across the select studies reviewed, there are consistencies: lesser activation and connectivity of brain regions involved in emotion processing and regulation in AUDs.

Studies of executive function in AUD use a variety of cognitive tasks related to set/goal shifting, working memory, and processing speed. Quaglieri and colleagues (2020) conducted a meta-analysis of 13 neuroimaging studies investigating decision making (monetary incentive delay tasks, delay discounting tasks, and Go/NoGo tasks) in AUD (n = 654, 47% AUD). During decision making, people with AUD exhibited higher activation of the putamen, dorsomedial PFC, and precuneus and lower activation of dorsolateral PFC, dorsal ACC, mid-cingulate cortex, and putamen, compared with controls. Cao et al. (2023) performed a meta-analysis of 13 studies investigating response inhibition comparing people with AUD with controls (n = 584, 43% AUD). People with AUD exhibited higher activation of the medial orbitofrontal cortex, dorsal ACC, paracingulate gyrus, striatum, and postcentral gyrus. People with AUD also exhibited lower activation of the dorsolateral PFC, inferior parietal gyrus, middle temporal gyrus, and superior temporal gyrus.

Chronic alcohol use results in widespread and pervasive changes to the brain, spanning fundamental regions involved in memory and executive function, to regions necessary for emotional processing and regulation. Notably, impacts of alcohol are consistently reported across imaging modalities (structural, resting-state, task-based).

Overlapping neural network in anxiety and alcohol use

Anxiety and alcohol are associated with distinct, yet overlapping, brain networks (Figure 1). Overlap in the neural networks associated with anxiety and alcohol use is of particular interest and concern. First, the overlap is of interest considering the prominent longitudinal data that suggests that having an anxiety disorder confers risk for developing an AUD. Second, the overlap is of concern considering that anxiety disorders, stress processing, and the ability to regulate stress may impact an individual’s ability to remain abstinent. Anxiety and alcohol use shape brain function in similar ways, suggesting an important lens through which to understand the development, maintenance, and treatment of AUDs.

Figure 1. Overlapping brain networks in anxiety and alcohol use disorders.

Top panel (“Anxiety network”) displays the canonical regions associated with anxiety and anxiety disorders. Middle panel displays the three stages of AUDs with regions associated with each stage highlighted; regions unshared with the anxiety network are gray or solid color. Bottom table shows the overlapping regions and highlights research reviewed in this paper determining how anxiety impacts brain activation and connectivity in each stage of AUD. Regions are shown at approximate anatomical location for illustrative purposes. The only non-shared region is the striatum (binge/intoxication and preoccupation/anticipation); however, note that the striatum is frequently associated with anxiety disorders but is not in the canonical anxiety network. PFC = prefrontal cortex; ACC = anterior cingulate cortex.

III. Role of anxiety and alcohol use disorder on brain function

Few studies have directly investigated the specific impact of anxiety on AUD. However, a related area—the relation between stress and AUD—has received more attention. Stress and anxiety are similar physiological responses that differ in their scope and definition: anxiety refers to the physiological response related to unknown or uncertain threats, whereas stress is a physiological experience that can result from anxiety or any short- or long-term exposure to real or perceived threat (i.e., stressor) (Daviu et al., 2019). Both anxiety and stress result in altered responses to emotional and cognitively demanding stimuli and tasks, leading to similar impacts on brain function. Thus, we will also cover literature outlining the impact of stress on different components of AUD, especially in cases where the literature on the intersection of anxiety and AUD is scant. The intersection of anxiety and alcohol is on the forefront of alcohol research, resulting in a limited amount of research that focuses on their intersection.

Resting-state neural networks

Two approaches to investigating resting-state neural networks are (1) including measurements of anxiety in studies of AUD and (2) including measurements of alcohol use in studies of anxiety. There are limited reports that explore resting-state connectivity, either in the context of anxiety in AUD and AUD in anxiety. Our group recently investigated the impact of anxiety and sex on resting-state connectivity during early alcohol abstinence (n = 40, 50% AUD) (Flook et al., 2023); anxiety was measured by several anxiety measures and combined to create a composite score. We focused on the BNST, given its prominent role in anxiety and the negative affect stage of AUD and examined a putative BNST network which included the BNST, amygdala, hypothalamus, anterior hippocampus, anterior insula, and ventromedial PFC. The early abstinence group had weaker BNST-hypothalamus connectivity compared with controls. Additionally, in the early abstinence group, higher anxiety symptoms were correlated with stronger BNST-amygdala and BNST-hypothalamus connectivity. Recently, Patel and colleagues (2025) investigated resting-state brain connectivity of the salience network in adults with and without AUD and anxiety (n = 264, 25% AUD, 25% AUD and anxiety). Individuals with AUD and anxiety had weaker connectivity between the supramarginal gyrus and PFC, compared with controls. Individuals with AUD with and without co-occurring anxiety had stronger connectivity between the left and right supramarginal gyrus, compared with controls. These limited data suggest that people with AUD in early abstinence have alterations in brain regions at the intersection of alcohol and anxiety networks.

Tasks probing general emotion processing

Cognition and reward.

Bach and colleagues (2024) investigated stress induction on brain responses to alcohol cues in people with AUD (n = 91, 100% AUD). People with AUD who underwent psychosocial stress (vs. control condition) exhibited higher insula activation to alcohol cues. Tenekedjieva and colleagues (2024) investigated the impact of different transdiagnostic factors on brain function in three cognition and reward tasks in people with AUD (n = 86, 73% men). Participants completed tasks probing reward (monetary incentive delay task), craving (cue reactivity task), and cognition (Go/NoGo task). A factor analysis of the Depression Anxiety and Stress Scale-42 and Posttraumatic Stress Disorder Scale-5 revealed four factors: trauma distress, negative affect, hyperarousal, and somatic anxiety—thus, three out of four symptom factors described anxiety (negative affect, hyperarousal, and somatic anxiety). During the reward task, all anxiety factors were associated with lower posterior parietal cortex activation. There were also some factor specific associations during the reward task: (1) higher negative affect was associated with decreased caudate and putamen activation, (2) higher hyperarousal was associated with decreased caudate activation, and (3) higher somatic anxiety was associated with decreased putamen activation. During the craving task, negative affect and hyperarousal were associated with lower dorsolateral PFC activation, negative affect was associated lower ACC activation, and hyperarousal was associated with lower posterior cingulate cortex activation. Finally, during the cognition task, negative affect was associated with higher dorsolateral PFC activation. Thus, anxiety was associated with alterations in multiple brain regions during reward, craving, and cognition, suggesting that anxiety impacts many underlying components of AUD.

Emotion processing.

Kienast et al. (2013) investigated the impact of anxiety on brain activation to aversive (vs. neutral) images in men recently abstinent from an AUD and controls (n = 24, 46% AUD); anxiety was measured using the State-Trait Anxiety Inventory (Spielberger et al., 1983). When viewing aversive images, men with AUD had lower activation of the posterior cingulate cortex, cuneus/precuneus, and caudate, compared with controls. Anxiety was not associated with differences in brain activation. Men with AUD had higher trait anxiety and stronger amygdala-dorsal ACC connectivity, compared with controls, but there was no correlation between anxiety and brain connectivity. MacIlvance and colleagues (2020) investigated brain responses during an emotional face-matching task in three groups (n = 232): AUD and co-occurring anxiety (39%), AUD without co-occurring anxiety (17%), and controls (44%). The co-occurring AUD and anxiety group had higher face-matching accuracy and stronger temporoparietal junction and supplementary motor area activation (neutral faces), compared with controls. The AUD without anxiety group exhibited higher supramarginal gyrus activation (fearful faces > control). When comparing the AUD groups with and without co-occurring anxiety, the anxiety group exhibited lower cerebellum and inferior frontal gyrus activation (fearful > neutral).

These data suggest that there are complex associations between AUD with and without a co-occurring anxiety disorder, and these associations impact fundamental brain regions related to stimuli processing.

Tasks probing anxiety-related brain networks

Stress.

Lee and colleagues (2023) investigated the impact of social stress on brain activation in adults with AUD and high trait anxiety (n = 39) and whether pexacerfont, a stress system antagonist, would impact brain activation. Anxiety was measured with the State- Trait Anxiety Inventory-Trait version (cut off score = 40). Participants completed the Trier Social Stress Task, then watched video recordings of themselves and other participants completing the social stress task while in the scanner. Adults with AUD and high trait anxiety had increased activation of the insula, ACC, ventrolateral PFC, and dorsolateral PFC when viewing themselves complete the social stress task; there was no effect of pexacerfont.

In addition to social stress, several studies have used two well-validated tasks that induce stress by either developing individualized stress scripts (e.g., retelling a recently stressful event the participant experienced) or showing stressful images. Seo and colleagues (2013) investigated brain activation to individualized stress vs. neutral script (n = 60, 50% AUD). Compared with controls, individuals with AUD in acute abstinence exhibited higher activation during stress imagery (vs. neutral) in the dorsomedial PFC, ventromedial PFC, dorsal ACC, posterior cingulate cortex, hippocampus, amygdala, caudate, temporal lobe, and cerebellum. Blaine and colleagues (2020) investigated brain activation to stress images in two samples. The first sample included individuals with an AUD during acute abstinence (2–4 weeks) and controls (n = 87, 51% AUD). Compared to controls, individuals with AUD showed lower activation during stress (vs. neutral) images in the dorsomedial PFC, ventromedial PFC, rostral ACC, and striatum. The second sample included individuals in AUD treatment (n = 69). Shorter abstinence length predicted lower activation during stress (vs. neutral) images in the dorsomedial PFC, ventromedial PFC, rostral ACC, insula, hippocampus, amygdala, and striatum. Goldfarb and colleagues (2022) investigated whole-brain connectivity networks during a stressful images task between light and risky drinkers (n = 104, 51% risky drinkers). While both groups exhibited similar networks, risky drinkers exhibited a network with stronger connections with the dorsolateral PFC, dorsal ACC, insula, primary visual and motor cortices, superior temporal gyrus, cerebellum, and ventral tegmental area. Radoman and colleagues (2024) investigated brain activation during a task with three conditions: alcohol images, stress images, and neutral images in treatment-seeking individuals with AUD (n = 77). Individuals with AUD showed higher activation of the orbitofrontal cortex, amygdala, hippocampus, dorsal caudate, and putamen during stress (vs. neutral) images. Individuals with AUD also showed lesser activation of the dorsomedial PFC, dorsolateral PFC, ventromedial PFC, ventrolateral PFC, and ventral caudate during stress (vs. neutral) images. Finally, Seo and colleagues (2024) investigated brain activation during the stressful images task in light and risky drinkers (n = 48, 54% risky drinkers). Risky drinkers exhibited lower activation during stress images in the ventromedial PFC, dorsal ACC, motor cortex, insula, striatum, and temporal gyrus, compared with the light drinkers.

Across these data, there is consistent recruitment of key stress and anxiety brain regions: the dorsomedial PFC, ventromedial PFC, dorsal ACC, insula, amygdala, and striatum. However, direction of effects differs across task type (script vs. imagery) and sample (current drinkers, risky drinkers, or abstinent), suggesting a complex relation between responding to stress and stage of AUD.

Uncertainty and anticipation.

Wilcox and colleagues (2020) investigated brain responses to threat and the association with anxiety in adults with an AUD (n = 52). Participants completed an anticipatory anxiety task that varied the threat of a heat stimulus, where brain activation was measured during early, middle, and late phases of the task. Overall, people with AUD reported higher anxiety during high (vs. low) threat stimuli. During the early phase of threat, people with AUD exhibited lower activation of the subgenual ACC/ventromedial PFC (> low threat). During the middle phase of threat, people with AUD also exhibited lower activation of the posterior cingulate cortex, dorsolateral PFC, middle and superior frontal gyri, and precuneus (vs. low threat).

Gorka and colleagues (2020) used the “N-P-U” task to study people with lifetime AUD and controls (n = 95, 38% AUD). The AUD group displayed greater activation of the insula and dorsal ACC to unpredictable threat, compared with controls. Later, Gorka et al. (2023) investigated whether brain activation during the “N-P-U” task predicted later drinking habits in youth at high-risk for AUD (n = 91). Greater activation in the anterior insula and dorsal ACC to unpredictable threat at baseline predicted higher binge drinking one year later.

Our group recently investigated brain activation and BNST functional connectivity during an unpredictable threat task in early abstinence from an AUD and controls (n = 40, 50% AUD) (Zabik et al., 2024). In the early abstinence group, higher anxiety was correlated with higher BNST activation during unpredictable threat, compared with controls. There was an interaction with sex; men in early abstinence with higher anxiety exhibited higher activation in the insula, dorsal ACC, and ostral ACC during unpredictable threat, compared with control men. For BNST functional connectivity during unpredictable threat, adults in early abstinence with higher anxiety displayed stronger BNST connectivity with the amygdala, ventromedial PFC, and dorsomedial PFC. Alcohol abstinent men also showed stronger BNST-ventromedial PFC connectivity than control men. Women in early abstinence had weaker BNST-ventromedial PFC and BNST-thalamus connectivity than control women. Thus, both sex and anxiety impacted brain responses to unpredictable threat in early abstinence.

Several conclusions can be drawn from these data. First, studies of anxiety using unpredictable threat tasks are a recent development in AUD research and offer a new insight into brain processes associated with chronic alcohol use. Second, emerging evidence suggests there are consistent alterations in anxiety brain networks including heightened insula and dorsal ACC activation. Third, sex differences are seen in AUDs and affect anxiety-related brain networks.

Conclusion

In summary, findings from task-based fMRI studies converge on alterations in the BNST, insula, dorsal ACC, and medial and lateral PFC subregions (dorsomedial, ventromedial, dorsolateral). The BNST is the hub of anxiety in the brain and is impacted by chronic alcohol use, causing hyperactivation during alcohol abstinence (Avery et al., 2016). The insula is largely responsible for processing internal information—like anxiety—and is altered by chronic alcohol use (Uddin et al., 2017). The dorsal ACC also has numerous unique roles relevant to anxiety and alcohol use; the dorsal ACC coordinates responses to threatening and anticipatory information, monitors for errors and rewards in the environment, and regulates motor control related to threat and reward (Bush et al., 2002; Knutson et al., 2000; Mesulam, 1990; Murtha et al., 1996). Finally, the medial and lateral PFC are the primary integrators of emotional and cognitive information, where they balance reward processing and decision making to guide behavior (Lindquist et al., 2012). Together, these regions represent critical hubs for emotional states and behaviors resulting from them—ultimately making them susceptible to anxiety and alcohol use. Of note, while the studies reviewed included people with current AUD, with lifetime AUD, and recently abstinent, no simple differences emerge in brain activation across these groups. The direction of effects was largely task-dependent: lower activation was observed during cognitive and emotion processing tasks, and higher activation was observed for alcohol cue and unpredictable threat tasks.

IV. Anxiety in the treatment of alcohol use disorder

Thus far, we discussed the impact anxiety has on brain measures in AUD. These findings suggest that anxiety plays a significant role in acute and chronic alcohol use, as well as abstinence. Importantly, anxiety shapes the course of AUD beyond development—the following sections will review how anxiety impacts alcohol treatment and the need for anxiety-focused treatments in AUD.

Anxiety as a barrier to recovery

In the early 1990’s, Johnston and colleagues (1991) characterized AUD treatment differences between people with and without a co-occurring anxiety disorder (n = 33, 52% anxiety). During the first 21 days of abstinence, men with AUD and co-occurring anxiety had more severe withdrawal symptoms. A follow-up study in the same sample revealed that the AUD with anxiety group also had significantly elevated levels of self-reported anxiety across the first 21 days of abstinence (Thevos et al., 1991).

Severe withdrawal symptoms and anxiety may also impact recovery from an AUD. Kushner et al. (2005) studied adults with AUD, with or without a co-occurring anxiety disorder, at treatment entry and 120 days after treatment onset (n = 53, 55% anxiety). Adults with an AUD and co-occurring anxiety disorder had more severe withdrawal symptoms, higher number of drinks consumed 90 days after abstinence, fewer days to first drink, and fewer days to first binge-drinking occasion.

Anxiety’s impact on alcohol use disorder treatments

Milivojevic et al. (2020) investigated whether prazosin treatment for AUD, as well as lifetime anxiety diagnosis, would impact acute measurements of anxiety and stress. People with AUD received prazosin or placebo and completed a stress-induced imagery task to capture acute anxiety, stress-related alcohol cravings, and cortisol levels (n = 40). People with AUD who had an anxiety disorder exhibited higher anxiety ratings across the task, with no effect of medication on those ratings. A treatment x anxiety diagnosis interaction was detected on cortisol measures: adults with AUD without anxiety who received placebo had higher cortisol responses to stress- and alcohol-related cues, with no cortisol changes in the prazosin group. These findings suggest that a lifetime diagnosis of anxiety may not impact prazosin treatment.

A meta-analysis from Agabio et al. (2021) explored the impact of anxiety on randomized control trials of baclofen treatment (GABA agonist) for AUD. For studies that included participants with clinically significant levels of anxiety, baclofen resulted in more days abstinent than in the placebo group. For studies that included participants with sub-clinical levels of anxiety, no difference was detected between baclofen and placebo on days abstinent. Thus, baclofen was more effective for people with significant anxiety.

Anxiety-based interventions for alcohol use disorder

Schadé et al. (2005) investigated whether additional anxiety treatment (cognitive behavioral therapy and medication) would improve outcomes for adults with AUD and co-occurring anxiety disorders (n = 96). Participants who received dual AUD and anxiety treatment had a significant reduction in anxiety levels, but no differences in relapse rates, compared with the AUD treatment-only group. Kushner et al. (2009) used a similar clinical trial design where participants either received AUD treatment as usual or AUD treatment and cognitive behavioral therapy for panic disorder (n = 54). The combination of AUD treatment and cognitive behavioral therapy was associated with a decrease in anxiety/panic symptoms and AUD symptoms, compared with the AUD treatment as usual group. Kushner et al. (2013) investigated whether mindfulness-based treatment (vs. cognitive behavioral therapy) would decrease anxiety symptoms and motivations to drink in adults with AUD and co-occurring anxiety disorder (n = 247). Cognitive behavioral therapy was associated with lower relapse rates, compared with the mindfulness-based therapy group.

Conclusions

Anxiety continues to shape AUD, even when individuals are abstinent, by impacting the effectiveness of treatments. These data highlight the importance of measuring anxiety in people seeking treatment to determine who may benefit more from treatments that target anxiety. While many novel treatments are being tested in clinical trials, it is worth noting that many individuals with co-occurring anxiety disorders are largely excluded from early-phase trials (Blanco et al., 2008; Hoertel et al., 2014). Thus, the studies reviewed represent a small sample of focused effort determining how to manage the intersection of anxiety and AUDs.

V. Conclusions and future directions

In this review, we described how anxiety shapes the development and maintenance of alcohol use. By highlighting the intersection of anxiety and AUD through longitudinal and brain imaging studies, we hope to underscore the critical importance and cyclical relation of anxiety and AUDs. By highlighting how anxiety shapes alcohol use, we hope to provide researchers and clinicians with important insights to inspire future research and clinical care—ultimately easing the shame and burden of alcohol use for individuals with AUDs and their families.

Below, we provide suggestions for future research and potential avenues for intervention.

1. What can we do to halt the progress from anxiety to alcohol use? Anxiety disorders develop during critical years of adolescence. Early identification and treatment could diminish the association between anxiety and alcohol use, as well as introduce positive coping mechanisms for young adults, therefore preventing development of alcohol use.

2. How can clinicians and researchers identify anxiety? Measuring anxiety during alcohol abstinence will be a key avenue for gathering epidemiological data on the course of alcohol use beyond developmental milestones. Formal assessments of anxiety and anxiety screening in primary care will be necessary to understand the course and shaping of alcohol use and treatment.

3. How can we examine the negative affect component of AUD in the clinic and lab? An important knowledge gap in AUD is understanding how anxiety impacts each part of the addiction cycle. Researchers are beginning to collect and analyze data in the three domains proposed by the Addictions Neuroclinical Assessment (Kwako et al., 2016, 2017). State anxiety measures before and after tasks will be a simple—yet effective—way to gauge how anxiety may shape different behaviors and brain processes in AUD. Additionally, studying interventions that target anxiety symptoms will be critical for people with co-occurring anxiety disorders. Alcohol craving is a focal point of treatment and has been successful; we believe the next step is to identify ways to treat anxiety in AUD. Negative affect is a chronic, lingering symptom that can drive craving and impulsive alcohol use.