People who binge drink show neuroendocrine tolerance to alcohol cues that is associated with immediate and future drinking- results from a randomized clinical experiment

Abstract

Neuroendocrine tolerance to alcohol, i.e., a blunted cortisol response to alcohol, has been linked to Ventromedial Prefrontal Cortex (VmPFC) alcohol cue reactivity and relapse risk in severe Alcohol Use Disorders (AUDs), but its role in the development of AUDs is not clear. Recent work suggests that blunted cortisol responses to alcohol cues in individuals who engage in binge drinking (BD) may play a role in motivation to consume larger amounts of alcohol, but the link between this dysregulated endocrine response and BD’s neural responses to alcohol cues remains unclear. To examine this, two groups of participants were recruited based on their recent drinking history. Thirty-three BD and 31 non-binging, social drinkers (SD) were exposed to alcohol cues and water cues in two separate 7 T functional magnetic resonance imaging (fMRI) scans. Each scan was followed by the Alcohol Taste Test (ATT) of implicit motivation for alcohol and a post-experiment, one-month prospective measurement of their “real world” drinking behavior. During each scan session, blood plasma was collected repeatedly to examine the separate effects of alcohol cues and alcohol consumption on cortisol levels. Relative to water cues and SD, BD demonstrated blunted cortisol cue reactivity that was negatively associated with VmPFC cue reactivity. In BD, both blunted cortisol and greater VmPFC cue reactivity were related to immediate and future alcohol consumption in the month following the scans. Thus, neuroendocrine tolerance in BD may be associated with increased incentive salience of cues and contribute mechanistically to increased alcohol consumption seen in the development of AUDs.

Introduction

Binge drinking is a known risk factor for the development of Alcohol Use Disorders (AUDs), yet 22.4% of US adults engage in this behavior at least monthly [1]. A growing body of preclinical and clinical research suggests that one physiological mechanism by which binge drinking increases risk for the development of AUDs is neuroendocrine tolerance to alcohol [2, 3]. Neuroendocrine tolerance to alcohol has two components. First, repeated exposure to large quantities of alcohol results in increased basal hypothalamic-pituitary axis (HPA) activity [3, 4]. Second, blunted phasic HPA axis responses to alcohol consumption are associated with greater alcohol seeking behavior and consumption, with greater consumption required to obtain previous levels of HPA axis stimulation [5]. Initially, glucocorticoid release to alcohol is associated with dopaminergic transmission and reward valuation activation in the prefrontal cortex [6]. However, in AUDs, as a result of repeated exposure to alcohol-induced higher levels of central nervous system glucocorticoids, excitotoxic cascades eventually alter the structure of the prefrontal cortex [7] and blunt its functional response to alcohol consumption [2, 7–10]. For example, the efficacy of the Ventromedial Prefrontal Cortex (VmPFC) in inhibiting the HPA Axis is diminished [2, 11].

In the progression of AUDs, it is generally accepted that the prefrontal reward valuation activation that once occurred in response to alcohol consumption itself shifts and becomes more responsive to cues that predict alcohol consumption is imminent or possible (for review [12], for meta-analysis in AUD [13, 14], for theory [15]). This change in incentive salience is associated with a shift in peak dopamine responsivity from alcohol consumption to the people, places, and things associated with drinking [16]. Neural and physiological responsivity to these cues is a key factor in relapse in AUDs [17]. Specifically, we have previously shown that increased Ventromedial Prefrontal Cortex (VmPFC) responsivity to alcohol cues is associated with neuroendocrine tolerance and predictive of time to relapse in two independent samples and therefore may contribute to the course of AUDs [11, 18].

Our recent studies also suggest that individuals who engage in binge drinking (BD) display increased basal cortisol levels and blunted HPA axis response to alcohol, which are both linked to increased craving, greater alcohol consumption in the laboratory, and blunted orbitofrontal PFC reactivity to alcohol consumption, i.e., neural reward tolerance [19, 20]. These results build on other clinical studies that have reported that BD show both a blunted cortisol response to a fixed, moderate or high oral alcohol administration compared to social drinkers who do not binge (SD [21–23]); and blunted prefrontal and striatal responses to acute alcohol [24, 25]. Increased alcohol cue reactivity, indicative of increased incentive salience, has also been shown in BD [26–28]. However, it is unclear if BD show neuroendocrine tolerance to cues and if neuroendocrine tolerance to cues is related to increased incentive salience of cues, i.e., greater VmPFC activation. Furthermore, it is unknown how these neuroendocrine and neural responses to cues relate to drinking behavior in BD outside the laboratory.

To examine how neuroendocrine and neural responses to alcohol cues may be associated with immediate and future alcohol consumption in BD, thereby increasing risk of the development of AUDs, we recruited beer drinking, non-smoking men and women ages 21–45 (N = 64, equal sex), 33 BD/ 31 SD, for two functional magnetic imaging (fMRI) cue reactivity sessions. The goal was to determine how immediate alcohol consumption and “real world” drinking behavior in a prospective one-month follow-up would be associated with neuroendocrine tolerance and VmPFC cue reactivity in at-risk BD (NCT04412824). Water cues were used as an active control and sessions were counter balanced and randomized among participants. We hypothesized that BD would show both signs of neuroendocrine tolerance, i.e., elevated baseline levels of cortisol relative to SD and blunted cortisol responses to alcohol cues, which would be negatively associated with VmPFC cue reactivity, and with both immediate and future alcohol consumption.

Materials and methods

Participants

At intake, 87 beer drinking, non-smoking men and women ages 21–45 were categorized as non-binging Social Drinkers (SD; <7 standard drinks/week for women or 14 standard drinks/week for males, with no occasions of binge drinking) or as people who engage in Binge Drinking (BD; binges of ≥4 drinks in a 2 h time span and ≥8 standard drinks/week for women or ≥5 drinks for men in a 2 h time span and ≥15 standard drinks/week for men) [29] (see Supplemental Fig. 1 for CONSORT Diagram). There was a requirement of at least 3 binges per month in the last 3 months for BD, as indicated by the Timeline Followback [30] and Cahalan Quantity and Frequency Variability Index [31] interviews. Current DSM-V psychiatric disorders, including current alcohol use disorders, and any prescription medications were exclusionary (except hormonal birth control). While 87 participants completed baseline assessments, 8 dropped out due to COVID-19 quarantine, 2 were excluded because of DSM-V substance abuse disorder diagnoses, 2 were not approved for participation in the MRI due to medical implants, and 1 did not consume enough alcohol in the past year (less than 6 drinks). Additionally, 5 participants who completed the MRI but did not have complete imaging data and/or blood collection were removed. Finally, 5 participants were removed prior to analysis because they did not complete at least 85% of the ecological momentary assessment surveys. The final sample size of N = 64 for cortisol analyses was determined a priori based on a previous study published by the first author demonstrating neuroendocrine tolerant responses to alcohol cues in BD [19]. However, the final sample size was N = 47 for all neuroimaging analyses, due to excessive head movement in 17 participants. Group comparisons for demographic, drinking, and psychological variables involved chi-square tests of frequency or independent t-tests based on mean, standard deviation, and number of participants per group. Groups were equivalent in participants’ sex, race, years of education, family history of AUD as measured by the Family Tree Questionnaire [32], years of regular drinking, and number of drinking days in the past month; groups were also not different on current subjective stress levels, depression and anxiety, impulsivity and number of childhood and lifetime traumatic events (Table 1).

Table 1.

Participant demographic, drinking, and psychological characteristics.

Demographic variables	Social drinkers (N = 31)	Binge drinkers (N = 33)	Cohen’s d/ Chi-square
Sex
Female	14 (45%)	16 (48%)
Male	17 (55%)	17 (52%)
Race and ethnicity
Black/African American	5 (16%)	1 (3%)
Caucasian	23 (74%)	32 (97%)
Asian American	3 (10%)	0 (0%)
Hispanic	6 (19%)	6 (18%)
Years of education	17 (2.5)	16 (1.6)
Age*	28 (7)	24 (4.5)	0.68
Drinking variables
Number of AUD first degree relatives	0.26 (0.6)	0.24 (0.6)
Years of regular drinking	7.5 (6.9)	5.2 (5)
Drinking days in past month	9.9 (7.4)	14.6 (6.4)
Total amount consumed in past month*	23.5 (19.2)	78.3 (63.7)	1.16
Cahalan QFVI usual number of drinks*	2.5 (0.93)	5.1 (2.8)	1.25
Cahalan QFVI max number of drinks*	5 (1.8)	9.2 (2.8)	1.78
Lifetime mild alcohol use disorder*	7 (22.6%)	18 (54.5%)	6.86
Alcohol use disorders identification test (AUDIT)*	4.6 (1.8)	11 (4)	2.06
Psychological variables
Perceived stress scale (PSS)	31.6 (4.4)	31.4 (4.3)
Beck depression inventory (BDI)	5.6 (4.9)	5.8 (5.8)
Barret impulsiviness scale (BIS)	68.9 (8.4)	71 (5.7)
Childhood trauma questionnaire (CTQ)	63.4 (4.6)	65.2 (4.6)
State-trait anxiety index (STAI)- TRAIT	30.6 (3.8)	29.7 (2.8)
State-trait anxiety index (STAI)- STATE	60.4 (11)	62 (5.4)

*denotes significantly different at p<0.05. When not a percentage, numbers in parentheses () indicates standard deviation.

Within person and between participants experimental design (Fig. 1)

Fig. 1. Between subject and repeated measures study design (NCT04412824).

Two groups of participants, SD and BD, categorized on the basis of NIAAA criteria [42], were randomly assigned to view alcohol pictures during one fMRI scan and water pictures during another scan on a different day. An Alcohol Taste Test followed each scan [19]. Before, during, and after the scans, blood samples were taken from participants for measurement of cortisol levels. After completion of the two scans, participants answered questions on drinking behavior via a smartphone app for 30 days.

In counter-balanced and randomized order, participants underwent the alcohol (ALC) and water (H20) cue fMRI sessions using a crossover design. Randomization of condition order was performed by the first author using a random number sequence generator which is freely available on the internet. Prior to each experimental scan session, participants were required to be drug free as tested by a urine drug screen for cannabis, benzodiazepines, opiates, and stimulants, and alcohol-free using a Draeger Alcotest 6820 Breathalyzer test (Lubeck, Germany). All scan sessions occurred at 2:00 PM to control for any effects of the diurnal cycle of cortisol. At the start of each scan session, a registered nurse or nurse practitioner inserted an in-dwelling intravenous catheter into the participant’s non-dominant arm to allow for repeated blood measurements. A baseline blood sample was drawn 55 min later to reduce the effects of needle insertion on baseline blood cortisol levels. Participants were placed in the 7 Tesla Siemens MAGNETOM MRI and underwent three 10 min Blood Oxygen Level Dependent (BOLD) functional visual cue runs, accompanied by blood draws for measurement of the peak cortisol response to cues. This crossover design allowed us to isolate the effects of alcohol cues on neural activity and thus the ALC-H2O contrast was utilized to assess the alcohol cue effect in all responses.

After each MRI scan, participants underwent a post-scan alcohol taste test (ATT) which allowed a specific assessment of post-scan alcohol motivation to isolate the effect alcohol versus water cues (ALC-H2O) on implicit motivation to consume alcohol. The 10 min ATT involved presenting the participant with 3 mugs of alcoholic beers (total of 1440 ml, equivalent to 4 cans of beer) and instructing them to taste the beers to assess if they are the “same or different” kind of beer. They were also instructed to “drink as much as they need to” to make that determination and that they would be paid $10 if they were correct. Participants received beer with a 4.2% alcohol concentration and the 3 mugs of beers presented were always the same as each other. Therefore, a participant could take a small sip from each beer glass and be able to make their determination. Notably, participants often choose to consume more than a sip of each beer, and the amount consumed serves as a behavioral index of alcohol motivation [19, 33]. A final blood draw occurred 30 min after completion of the ATT to measure peak cortisol response to alcohol consumption. It is important to note that in the afternoon, cortisol levels are naturally dropping, as per the diurnal cycle [34]. Thus, if cortisol release is seen in response to alcohol cues or consumption, this normal reduction in levels is diminished (the difference between responses to alcohol and water cues is closer to zero). If cortisol is not released, or in fact is suppressed, the normal afternoon reduction is amplified. Finally, after completion of the two scans, participants answered questions about drinking behavior via a smartphone app for 30 days.

Neuroimaging procedures and analysis

Each scan consisted of 3 functional blocks with two types of visual stimuli (i.e., neutral pictures and pictures of water cues or pictures of alcohol cues). Each run lasted 10:24 and included a 90 s fixation cross, 3:18 of neutral stimuli presentation (33 images shown for 5 s each in randomized order per block with a one second fixation displayed between images), and 6:36 of alcohol or water cues (66 images shown for 5 s each in randomized order per block with a one second fixation displayed between images; see Supplemental Methods for details on cue selection and evaluation). All visual stimuli were used only once. Scanning occurred in a 7 T Siemens MAGNETOM MRI system equipped with a standard 32 channel head coil, using the T₁ magnetization-prepared rapid gradient-echo (MPRAGE) sequence for structural scanning and an echo planar (T₂^*) sequence was used to collect functional images (see Supplemental Methods for detailed scan parameters).

Results included in this manuscript come from preprocessing performed using FMRIPREP version 20.2.0 [35] [RRID:SCR_016216], a Nipype [36] [RRID:SCR_002502] based tool. General linear modeling (GLM) was used for first-level analyses (e.g., individual-level) on each voxel in the entire brain volume using FSL FEAT (FMRI Expert Analysis Tool) Version 6.00, part of FSL [37] (FMRIB’s Software Library, www.fmrib.ox.ac.uk/fsl; see Supplemental Methods for details of preprocessing performed).

For higher-level (e.g., group level) data analysis, linear mixed effects modeling using FSL FEAT was implemented with a 2 (average of runs for each session: Alcohol-Neutral, Water-Neutral) x 2 (group: BD, SD) design while covarying for age and sex. Session and run were treated as within-person fixed-effect factors, group as a between-person factor, and participant as a random factor. A voxel wise z = 3.1 (p < 0.001) and cluster wise correction of p < 0.05 was applied to the ALC-H2O contrast. All associations presented are with each participants’ ALC-H2O percent signal change value for each significant cluster.

Ecological momentary assessment

The MetricWire app (metricwire.com; Waterloo, Ontario, Canada), which can be used on Android or Apple devices, was used to collect a waking survey, 2 random prompt surveys, and an end of day survey from participants for a period of 30 days. Participants reported on the number of drinks since last survey during the waking survey, random prompts, and the end of day surveys. Participants were provided with training and visuals regarding what a standard unit of alcohol is in beer, wine, and liquor. Waking survey number of drinks reported were assigned to the previous day and added to the values from the random prompts and end of day to survey to get a total number of drinks per day. Average response rate for the entire sample was 85% completion of surveys. In total, 6792 surveys were completed (an average of 106 out of 120 per participant), of which 8% (538) represented unique drinking occasions. All participants reported at least one drinking occasion in the month following the scans. The average number of drinking days in the 30 day follow up was 8.4 ± 4.1 and was not different between groups, t(62) = 1.041, p = 0.3. In contrast, the average drinks per drinking day was 5.1 ±  1.2 for BD and 1.42 ± 1.25 for SD, which was significantly different between groups, t(62) = 2.1, p = 0.037. Total drinks consumed over the 30 days therefore also differed between groups, (BD 78.3 ± 63.7 ;SD 23.5 ± 19.2; t(62) = 4.6, p < 0.0001) and was used in association analyses.

Cortisol assays

Plasma samples (4 ml each) from each timepoint were processed in duplicates and analyzed via the Elecsys Cortisol II immunoassay at the Biomarker Core Laboratory at Penn State University (State College, PA). All associations presented are with each participants’ baseline corrected ALC- H2O cue response.

Statistical analysis

Baseline group differences and group differences by condition and timepoint for outcome variables were examined in JASP (Version 0.17.2) [38] using ANCOVA, repeated measures ANCOVA (age and sex were included as covariates in all analyses). A p-value of 0.05 or less was considered significant. Associations between cortisol responses, ATT alcohol consumption, and future drinking were examined using multiple linear regression with group and sex as dummy coded variables and age as a continuous covariate. If group was a significant predictor (i.e., moderator) of the relationships and/or if a priori hypotheses existed regarding relationships among the cortisol cue response, VmPFC activation, ATT alcohol consumption, or future drinking, the data was split by groups and linear regression was performed to determine the relationship between experimental outcomes and future drinking within each group.

Study approval

The protocol was approved by the Auburn University Human Subjects Research Institutional Review Board. All participants provided informed, written consent before participation. All data was collected between June 2020 and November 2021.

Results

Participants

Sixty-four participants completed 2 MRI scanning sessions with viable blood samples, in addition to completing at least 85% of surveys during the 30 days follow up phase. At baseline, BD showed significantly higher AUDIT scores (t(62) = 8.16, p < 0.0001), total amount of alcohol consumed in the past month (t(62) = 4.6, p < 0.0001), and both usual (t(62) = 4.9,p < 0.001) and maximum (t(62) = 7.09,p < 0.0001) number of drinks per drinking episode (see Table 1 for means and standard deviations). Seventeen participants were excluded from analysis due to excessive head movement in the scanner (greater than 1.5 mm in any direction). 11 BD and 6 SD were excluded from neuroimaging analyses due to excessive motion; this did not differ as a function of group (χ² 9(64) = 1.6, p = 0.21). Therefore, N = 47 participants were used for fMRI analyses.

Baseline cortisol

Unlike in our previous study [20], we did not find baseline differences in cortisol levels between BD and SD, F (1,60) = 0.293, p = 0.59 (Fig. 2a). SD demonstrated unusually high levels of baseline cortisol in this study, an effect that may attributable to COVID-19 stress [39, 40], as data was collected between June 2020 and November 2021. There was also no effect of condition on baseline cortisol levels, F (1,60) = 0.29, p = 0.6.

Fig. 2. Neuroendocrine responses to alcohol cues and consumption.

a Groups did not differ in baseline cortisol levels, contrary to prior findings. b BD consumed significantly greater alcohol than SD across conditions and specifically after exposure to alcohol cues. c BD show blunted cortisol cue reactivity relative to SD, but no difference in cortisol responses to alcohol consumption. d Blunted cortisol reactivity in BD was associated with greater alcohol intake in the Alcohol Tate Test. All associations were calculated with the cortisol response to alcohol cues after controlling for the response to water cues. * denotes significantly different at p < 0.05.

Alcohol taste test

A group by condition interaction was seen in the 2 × 2 repeated measures ANCOVA for alcohol consumption in the ATT, (F (1,58) = 4.9, p = 0.03, η² = 0.024; Fig. 2b), such that BD drank significantly more alcohol in the alcohol cue condition relative to the water cue condition, while the amount of alcohol consumed by SD did not differ by condition. There were also significant effects of group, such that BD consumed more alcohol than SD F(1, 58) = 10.7, p = 0.002, η² = 0.13, and condition, such that more alcohol was consumed in the alcohol cue condition relative to the water cue condition F(1,58 = 8.5, p = 0.005, η² = 0.02). The amount of alcohol consumed in the ATT was significantly related to the amount of alcohol consumed in the month after the scan across all participants, F(1,58) = 5.16, p = 0.03, R² = 0.11. However, group significantly moderated this association [F(1,61) = 15.7, p < 0.0001, R² = 0.25], which was significant in BD, F(1,32) = 6.43, p = 0.018, R² = 0.21, but not for the SD group alone, F(1,30) = 1.05, p = 0.32.

Cortisol cue reactivity and immediate consumption

Confirming past results, a repeated measures ANCOVA showed a condition by group effect of cues such that BD showed blunted cortisol release to alcohol cues relative to water cues and relative to SD, F(1,58) = 4.1, p = 0.048(Fig. 2c). The relationship between cortisol levels and alcohol consumption in the ATT differed by group at a trend level, F(1,58) = 4.82, p = 0.04. In BD, greater blunting in cortisol cue reactivity was negatively related to alcohol consumption in Alcohol Taste Test, F(1,32) = 7.03, p = 0.01, R² = 0.14 (Fig. 2d), but this was not true for SD, F(1,29) = 0.7, p = 0.41.

Neural alcohol cue reactivity and immediate consumption

Contrary to our hypotheses, there were no group differences in BOLD response to alcohol cues after controlling for the effects of water cues after applying a voxel threshold of p < 0.001 and a cluster threshold of p < 0.05. All participants showed increased response in the left Ventral Anterior Cingulate (Brodmann Area [BA] 24; Montreal Neurological Institute [MNI] coordinates xyz −3,−16,39), right Dorsomedial Prefrontal Cortex (BA 9; MNI xyz 0,41,30), right Lateral Orbitofrontal Cortex (Brodmann area 47; MNI coordinates xyz 32,39,−9), left Ventromedial Prefrontal Cortex (BA 10; −29,57,15; Fig. 3a), left Angular Gyrus (BA 39; −52,−72,25), left Frontal Eyefields (BA 8; −17,35,39), left Ventral Posterior Cingulate Cortex (BA 23; −9,−50,30; Fig. 3a), and left Dorsal Anterior Cingulate Cortex (BA 32; −1,49,−11; Table 2). Across all participants VmPFC activation was positively associated with alcohol consumption in the immediate post-scan ATT, F(1,46) = 4.0, p = 0.049. R² = 0.09. While groups did not differ in the relationship between VmPFC and ATT alcohol consumption, F (1, 46) = 0.56, p = 0.46, VmPFC activation was significantly related to immediate alcohol consumption in the ATT in each group: BD F(1,23) = 5.67, p = 0.026, R² = 0.198, SD F(1,23) = 4.94, p = 0.045, R² = 0.27.

Fig. 3. Neural and endocrine cue and consumption reactivity vs. future drinking behavior in the “real world”.

a All participants showed greater Ventromedial Prefrontal Cortex activation in response to alcohol cues relative to water cues and this neural cue reactivity did not differ between the groups. Contrast shown is (alcohol-neutral)- (water-neutral) in a whole brain analysis corrected at a voxel threshold p < 0.001 and cluster correction of p < 0.05 (Z = 3.1). MNI coordinates of slice shown are at x = −8, y = 51, z = −10. b However, neural cue reactivity (in response to alcohol cues after correcting for response to water cues), was differentially related to cortisol cue reactivity (in response to alcohol cues after correcting for response to water cues), such that as VmPFC cue reactivity (x axis) increased cortisol responses (y axis) became more blunted in BD, but not SD. c Greater alcohol consumption in the month following the scans (y axis) was negatively associated with cortisol cue reactivity in BD, such that those with blunted cortisol responsivity consumed more. d BD’s increased VmPFC cue reactivity (x axis) was also associated with greater alcohol consumption in the month after scans (y axis). All associations were calculated with the cortisol and/or VmPFC response to alcohol cues after controlling for the response to water cues. * denotes significantly different at p < 0.05.

Table 2.

Significant clusters in alcohol- water contrast in all participants, whole brain p<0.001, cluster p<0.05.

Region	BA	Max Z	Voxels	x	y	z
Left ventral anterior cingulate	24	4.19	190	−3.3	−16.2	39.1
Right dorsomedial prefrontal cortex	9	5.58	237	0.1	41.1	29.5
Right lateral orbitofrontal cortex	47	4.92	347	31.7	38.5	−8.9
Left ventromedial prefrontal cortex	10	4.76	532	−29.0	56.5	15.1
Left angular gyrus	39	5.86	742	−52.0	−71.8	24.7
Left frontal eyefields	8	5.16	1615	−17.0	34.5	39.1
Left ventral posterior cingulate cortex	23	6.02	2944	−9.3	−50.4	29.5
Left dorsal anterior cingulate cortex	32	5.64	3635	−0.4	48.8	−11.3

BA Brodmann area, Max Z Z statistic for peak voxel in cluster, x,y,z MNI coordinates.

Associations among neural and endocrine cue reactivity and future drinking

The amount of alcohol consumed in the ATT after alcohol cue exposure, relative to water cue exposure, was significantly related to the number of drinks consumed in the month following experimental appointments across all participants F (1,56) = 5.16, p = 0.03, R² = 0.11. This effect, however, was moderated by group, F(1,45) = 6.1, p = 0.017. Specifically the effect was driven by a strong relationship in BD, (1,27) = 6.43, p = 0.018, R² = 0.21. Conversely, the relationship was not significant when considering the SD alone, F(1,29) = 1.05, p = 0.32. Group also moderated the relationship between VmPFC activation and cortisol release, F(1,45) = 4.12, p = 0.046. A significant inverse relationship between VmPFC activation and cortisol release was seen in BD, F(1,21) = 5.3, p = 0.03, R² = 0.18, but not in SD, F(1,24) = 0.08, p = 0.78 (Fig. 3b). Cortisol cue reactivity was differentially related future alcohol consumption in the month after the scan between the two groups, F(1,58) = 4.1, p = 0.048. A negative relationship was seen between cortisol reactivity and future alcohol consumption in BD, F(1,31) = 4.23, p = 0.048, R² = 0.12, but not SD F(1,29) = 0.49, p = 0.49. Finally, VmPFC cue reactivity was also differentially related to future alcohol consumption in the two groups, F(1,45) = 8.08, p = 0.007. VmPFC cue reactivity was positively related to alcohol consumption in the following month for BD, F(1,21) = 10.76, p = 0.003, R² = 0.31, but not for SD F(1,24) = 0.7, p = 0.41 (Fig. 3d).

Discussion

Current findings from this crossover neuroimaging experiment demonstrate neuroendocrine tolerance to alcohol cues in BD that is linked not only to greater VmPFC cue reactivity, but also immediate alcohol consumption in the lab and future alcohol consumption in the real world over the next month. To our knowledge, this is the first study to link neuroendocrine tolerance in BD to cues to prefrontal cue reactivity, an accepted marker of incentive salience that has repeatedly been associated with measures of AUD severity and AUD treatment outcomes. In addition, this is the first study to demonstrate the ecological validity of laboratory measures of neuroendocrine tolerance as an indicator of real world drinking behavior.

Neuroendocrine tolerance is a known phenomenon in moderate-severe AUDs and its contribution to prefrontal dysfunction has been demonstrated in preclinical studies [41]. We have previously presented a model by which it may also contribute to the development of AUDs. Specifically, we posit that initial cortisol release in response to alcohol consumption is higher in BD and contributes to the greater stimulation BD receive from drinking compared to SD [2, 21, 22]. However, with repeated bouts of binge drinking, the HPA axis becomes dysregulated, resulting in higher tonic levels of cortisol and blunted cortisol responses to alcohol consumption. We hypothesize that BD then must consume greater amounts of alcohol to get the cortisol release and associated stimulation previously derived from lesser amounts of alcohol. In support of this model, we have previously shown increased basal levels of cortisol and neuroendocrine tolerance to cues in BD, both of which were associated with craving and drinking behavior in the laboratory [19]. We have also recently demonstrated the role of prefrontal neural tolerance to self-motivated alcohol administration in BD [20]. In the current study, we now show that laboratory measures of neuroendocrine tolerance are related to real world drinking behavior. In combination with previous studies, then, we have provided evidence that neuroendocrine tolerance may contribute mechanistically to the increased alcohol consumption that underlies the development of AUDs. Future studies could demonstrate this mechanistic function by pharmacologically manipulating both tonic cortisol levels and phasic cortisol responses to alcohol cues and consumption in the lab. If pharmacological manipulation is shown to be possible, neuroendocrine tolerance may represent a biological mechanism that can be intervened upon in the prevention of escalation of drinking behaviors and the associated development of AUDs.

However, it should be noted that unlike in our previous studies, BD did not show elevated basal cortisol levels relative to SD in this study. This was most likely an effect of data collection occurring during the COVID-19 pandemic[39, 40]. BD showed basal levels similar to our previous reports [20], but SD showed elevated resting cortisol levels not seen in our previous studies. Moreover, in the current study SD did not show HPA axis tolerance to alcohol cues, i.e., they demonstrated a robust cortisol response to the alcohol cues. Combined, these results suggest healthy HPA axis functioning in SD in response to both current events (COVID-19) and to alcohol cues, an effect not seen in BD.

Another surprising result contrary to our hypothesis was that BD did not show greater VmPFC cue reactivity than SD. There are many potential factors that may have contributed to this result. First, the loss of 17 participants (11 BD; 6 SD) from neuroimaging analyses due to excessive head movement may have reduced our statistical power, i.e., ability to detect group differences. Moreover, we controlled for not only the effects of neutral cues, but also the appetitive effects of water cues to isolate neural activation related to alcohol cues specifically. Previous studies have often used only neutral or only appetitive controls when examining cue reactivity. Finally, the potential enhancement effect of the COVID-19 pandemic on the incentive salience of alcohol cues in SD may have contributed to our lack of group differences. It is important to note, however, that it is possible BD and SD do not, in fact, have differences in VmPFC cue reactivity. It is possible that prior to AUD diagnosis, binge drinking behavior is not associated with a large enough increase in incentive salience for alcohol cues to alter neurobiological function and therefore VmPFC activation.

Nevertheless, the current study has many strengths, including the similarity of groups on demographic and psychological variables, use of multiple modalities (i.e., neuroimaging of prefrontal alcohol cue reactivity, neuroendocrine measurement of cortisol in blood plasma, and ecological momentary assessment of drinking behavior in the real world) and linking of laboratory measurements to real world drinking behavior. Nevertheless, this study is limited by the small sample size and thus the inability to explore sex-related differences in neuroendocrine response to alcohol cues. Additionally, although we carefully obtained drinking histories using multiple measures prior to SD and BD classification, no biochemical verification of current alcohol use was performed. Moreover, the SD were statistically older than the BD, but the ranges overlapped significantly. Future studies should therefore be powered to examine sex differences and rule out any age effects. Larger sample sizes would also allow for loss of participants due to excessive head movement by participants in the scanner and likely show group differences in VmPFC cue reactivity. Finally, future data collection should aim to replicate these findings outside of a pandemic to confirm not only healthy HPA axis functioning in controls but also that the findings are replicated when considering normative drinking behavior (i.e., outside of lockdowns). These limitations notwithstanding, this study provides converging evidence for the validity of neuroendocrine tolerance as a mechanism that contributes to increasing alcohol consumption in the development of AUDs.

Author contributions

SKB designed the study, obtained funding, and supervised all aspects of data collection, analysis, and wrote the manuscript. CR, BC, and LC collected the data. RM, JLR, EBA, and EDC provided necessary assistance in setting up data collection and in data analysis, in addition to interpretation of results. EDC performed crucial analyses and edited the paper. All authors read and approved the final manuscript.

Funding

Funding for this work was provided by National Institutes of Health grant R00 AA025401 (SKB).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41386-023-01735-9.