Free-access intravenous alcohol self-administration in social drinkers and individuals with alcohol use disorder: Evaluation of relationships with phosphatidylethanol and self-reported alcohol consumption

Abstract

Background:

The free-access (FA) intravenous alcohol self-administration (IV-ASA) paradigm is an experimental approach that can identify modulators of alcohol consumption in humans. Moreover, the outcome measures of IV-ASA paradigms are associated with self-reported alcohol intake using the timeline follow-back method (TLFB). To evaluate how FA IV-ASA reflects drinking in real life, we examined the relationship between an objective marker of recent alcohol intake, phosphatidylethanol in blood (B-PEth), and TLFB and measures obtained during IV-ASA in individuals with alcohol use disorder (AUD) and social drinkers (SD). We also explored the associations between these measures and gut-brain peptides involved in AUD pathophysiology.

Methods:

Thirty-eight participants completed a laboratory session in which they self-administered alcohol intravenously. The safety limit was 200 mg%, and main outcomes were mean and peak breath alcohol concentrations (BrAC). Blood samples were drawn prior to IV-ASA and subjective alcohol effects were rated during the experiment.

Results:

The study sample comprised 24 SD and 14 participants with DSM-5 mild AUD. Although BrACs were not associated with B-PEth or TLFB in the full sample or AUD subgroup, there was an association with TLFB in SD. In both subgroups, BrACs were associated with alcohol craving but with differential timing. Total ghrelin levels were higher in AUD participants than in SD.

Conclusions:

No associations between B-PEth levels and achieved BrACs were observed in the mild AUD group, the SD group, or the full sample. The ability for FA IV-ASA to reflect recent drinking was confirmed only for TLFB in SD, whereas there were no associations within the smaller subsample of participants with mild AUD or in the full sample. Further studies that include a larger AUD sample are warranted. The association of BrACs with craving for alcohol suggests that the IV-ASA method may be useful for assessing interventions that target craving. This could be explored by using the FA IV-ASA model to evaluate the effects on craving of approved pharmacotherapies for AUD.

INTRODUCTION

Alcohol use disorder (AUD) is a chronically relapsing brain disorder that substantially contributes to the global burden of disease (Hasin et al., 2013; Koob & Volkow, 2010; Rehm et al., 2017; World Health Organization et al., 2018). The pathophysiology of AUD involves alterations in several different brain and peripheral signaling systems including the gut-brain peptides ghrelin, glucagon-like peptide 1 (GLP-1), and their associated receptor systems (Farokhnia et al., 2019; Jerlhag, 2019). Since the efficacies of available interventions are low, identifying new, improved interventions that reduce alcohol intake is warranted (Jonas et al., 2014; Maisel et al., 2013). Novel treatment concepts for AUD are today evaluated in expensive and time-consuming randomized controlled trials (RCT). Hence, the field of addiction medicine would benefit from a resource-effective, experimental approach that serves to indicate the ability of an intervention to reduce alcohol intake before the initiation of an RCT (Litten et al., 2016).

Intravenous self-administration of alcohol (IV-ASA) has been presented as a sensitive tool for detecting modulators of alcohol intake and thus may be suitable for evaluating interventions in man that should be tested in subsequent RCTs (Zimmermann et al., 2013). In this model, study participants self-administer alcohol intravenously (IV) by requesting standardized increments in breath alcohol concentration (BrAC) by pushing a button connected to software that drives an infusion pump (Zimmermann et al., 2013). In the free-access (FA) paradigm, alcohol is infused following each button press whereas the progressive ratio (PR) paradigm instead demands a successively increasing effort for the delivery of alcohol. Consequently, the gastrointestinal absorption phase of alcohol is circumvented in IV-ASA, thereby eliminating the largest source of variation in BrAC when alcohol is ingested. In addition, a physiologically-based pharmacokinetic (PBPK) model employed, which individualizes the infusion profiles, nearly eliminates the interindividual variability in incremental BrAC (Plawecki et al., 2012; Zimmermann et al., 2013). This precision allows for a lower number of participants to partake in the study. Adding to the reliability of the method, study participants reach high mean and peak BrACs, which are in line with self-reported drinking behavior (Cyders et al., 2020; Zimmermann et al., 2013). Several variables, for example, age, sex, personality traits, and heredity, have been shown to impact IV-ASA behavior (Cyders et al., 2016; Hendershot et al., 2016; Vaughan et al., 2019). Previous studies have presented a correlation between alcohol consumption during FA (Stangl et al., 2017) and PR (Bujarski et al., 2018) and recent drinking history assessed with the self-reported measures timeline follow-back (TLFB; Sobell & Sobell, 1992) and Alcohol Use Disorder Identification Test (AUDIT; Babor et al., 1989). However, it is yet to be established if an objective biomarker of recent alcohol intake would reflect and/or predict the outcome in IV-ASA. Phosphatidylethanol in blood (B-PEth) is a highly specific and sensitive biomarker of alcohol consumption, which reflects the average alcohol intake during the preceding 2 weeks both in participants with AUD and in social drinkers (SD; Isaksson et al., 2011; Kechagias et al., 2015; Walther et al., 2015). In a pharmacological RCT for AUD, B-PEth appeared superior to self-reported drinking as a measure of alcohol intake and for disclosing a medication effect, presumably due to a markedly reduced placebo response (de Bejczy et al., 2015; Walther et al., 2015). In order to further evaluate how the IV-ASA model may be implemented in research related to putative AUD interventions, it is of considerable interest to examine how accurately FA IV-ASA reflects real-life drinking behavior as measured by B-PEth in individuals with and without AUD.

The IV-ASA model can be used also for examining the pathophysiology of excessive alcohol intake and/or AUD. Indeed, previous studies have explored the relationship between neuropeptides suggested to be involved in AUD pathophysiology and IV-ASA outcome, where ghrelin administered exogenously increased achieved BrACs in PR IV-ASA (Farokhnia et al., 2018) whereas IV alcohol administration did not appear to alter ghrelin levels (Leggio et al., 2013). However, it is still unknown whether alcohol exposure during IV-ASA is associated with endogenous levels of gut-brain peptides, for example, ghrelin and GLP-1.

In the present study, we examined the correlation between alcohol consumption during an FA IV-ASA paradigm and B-PEth in SD and participants with an on average mild AUD. Furthermore, we examined the interplay between subjectively rated acute alcohol effects, such as changes in craving, and alcohol consumption during IV-ASA. Finally, we explored the relationship between endogenous serum levels of gut-brain peptides and alcohol consumption during FA IV-ASA.

MATERIALS AND METHODS

Participants

Participants were recruited through local advertisements and respondents were prescreened by telephone to review inclusion and exclusion criteria and to inform about the experimental procedures. Fifty-four prescreened participants underwent a screening visit at the study site including a medical examination, obtaining blood samples and an electrocardiogram. Inclusion and exclusion criteria are listed in Table S1. A total number of 38 screened participants fulfilled all criteria and were enrolled for subsequent study. The time interval between screening and study session was no longer than 3 weeks. An a priori sample size calculation based on previously reported small to moderate associations between self-reported drinking history and FA IV-ASA BrACs (Stangl et al., 2017) indicated that a sample size ranging between n = 29 to 123 would yield a power of 0.80 at an alpha level of 0.05. For this pilot study, a sample size of 38 was considered to be sufficient.

Experimental design and procedures

A schematic overview of the experimental design is provided in Figure 1. On the screening day, participants reported to the study center at Clinical Trials Center (CTC), Sahlgrenska University Hospital at ~9 a.m. Written, informed consent was obtained after a full explanation of study procedures. The study protocol was approved by the Swedish Ethical Review Authority (2019–04120) and complied with the Declaration of Helsinki. A comprehensive medical and psychiatric examination was performed by a medical doctor in order to confirm inclusion and exclusion criteria (Table S1). Also at the screening visit, a TLFB interview (Sobell & Sobell, 1992) reviewing in detail the use of alcohol and, if applicable, tobacco during the preceding 4 weeks was performed. Alcohol consumption pattern and alcohol problems were assessed with AUDIT (Babor et al., 1989), as well as a structured clinical interview based on the DSM-5 AUD criteria. Blood and urine samples were obtained to screen for medical disorders, use of illicit substances and to rule out pregnancy in women. Participants were instructed to abstain from alcohol 24 h prior to the experiment and to ingest a standardized, light breakfast on the experimental day.

FIGURE 1.

Schematic overview of the experimental design. Volunteers were recruited through local advertisement and prescreened by telephone. A total number of 54 participants underwent a screening visit where 14 participants with AUD and 24 SD were eligible to partake in the study. The time that elapsed between screening and the IV-ASA experiment was 10.6 ± 5.5 days (mean ± SD) On the experimental day, blood samples were drawn for both phosphatidylethanol (B-PEth) and gut-brain peptides prior to exposure to alcohol. The experiment started with a priming period during which participants were instructed to press for intravenous IV alcohol four times consecutively. This was followed by the free-access phase where participants were free to press for alcohol or refrain from doing so, and instructed to achieve their preferred, pleasant level of alcohol effect as if they were drinking at home or at a social gathering. Simultaneously, subjective acute alcohol effects were rated at baseline, 25, 60, and 100 min.

On the day of the experiment, participants arrived at the study center at ~9 a.m. and provided a BrAC reading and a urine sample for pregnancy and drug testing. Next, an IV catheter was placed in the antecubital fossa of the nondominant arm, and baseline blood samples were obtained for subsequent analyses of B-PEth and gut-brain peptides. Participants were then instructed on how the Computer-assisted Alcohol Infusion System (CAIS) would lead them through the experiment; how to use the button to request an alcoholic “drink” and how to complete electronic forms assessing subjective effects of infused alcohol. Participants were informed that each button press resulted in an IV alcohol infusion lasting several minutes, during which it was not possible to press for more alcohol until the ordered “drink” was fully delivered. The experiment started with a priming period during which participants were instructed to press for alcohol four times consecutively. Each request raised BrAC by 7.5 mg% over 2.5 min, thereby achieving a priming alcohol exposure of 30 mg% in 10 min. During the following 15 min, no “drink” could be requested, and the infusion rate was programmed to decrease BrAC by 1 mg%/min, thus resulting in a BrAC of approximately 15 mg% at the end of the priming period. For the following 75 min, participants were free to press for alcohol or refrain from doing so, and instructed to achieve their preferred, pleasant level of alcohol effect as if they were drinking at a social gathering, but to avoid reaching a level of intoxication accompanied by aversive effects. This procedure is referred to as the FA (free-access) paradigm for IV-ASA, previously described by Zimmermann et al. (2013). Participants evaluated their subjective effects at baseline, 25, 60, and 100 min (end of experiment). In addition, a precise reading of BrAC was obtained every 15 min, using a highly accurate breathalyzer (Evidenzer Classic, Nanopuls, Sweden) provided by the Swedish National Forensic Center (NFC). Throughout the experiment, participants were allowed to watch television or listen to music but were instructed not to read or use their telephone, as that could alter their self-administration or serve as a distraction from the messages on the CAIS computer screen. Participants were not informed about the increase in BrAC per request or what BrAC level they had reached during the experiment. After 100 min, the experiment was terminated, the IV catheter was removed and participants were offered a full meal. They were free to leave the study center as soon as BrAC had fallen to 20 mg%, or to take a taxicab or be collected by a friend, provided that they were not visibly impaired by alcohol. Prior to the experiment, participants were led to believe that they had to remain at the study center for 3 h following the termination of the experiment regardless of intoxication level in order to discourage participants from infusing less alcohol with the intention of leaving earlier. All participants received a 50 USD gift card as a small compensation for partaking in the study.

Methods and equipment for ethanol (EtOH) infusion

The alcohol infusate was prepared in the lab by mixing NaCl 0.9% with 95% EtOH (Alkohol Konzentrat 95% Braun, Germany) to give a final concentration of 6% (v/v). The infusion was delivered by a dual infusion pump (Alaris Imed Gemini PC-2TX; DiaMedical USA) and controlled by a computer equipped with the CAIS software and its incorporated PBPK model. The computer was also attached to double screens; one informing the participant on whether it was possible to press for alcohol, and the other, hidden from the study participant, that presented estimated BrAC with a time resolution of 30 s and infusion rates. Prior to the IV-ASA session, the participant’s age, sex, height, and weight were entered into the CAIS software, which transformed these data into PBPK parameters that individualized the model of alcohol distribution and elimination (O’Connor et al., 2000; Plawecki et al., 2007). The CAIS software calculated the infusion rates and controlled the pump to deliver the participant’s chosen BrAC profile. While the alcohol reward, that is, incremental BrAC exposure, was delivered, the button was deactivated. Once every 3 s in the background, the CAIS software calculated the future time course of BrAC as if a new drink had been ordered. If the BrAC was anticipated to reach the safety limit of 200 mg%, the drink button was deactivated until a request would not exceed that constraint. In parallel, as a further safety measure, readings of the actual BrAC as measured by a highly accurate breathalyzer were entered into the CAIS software every 15 min in order to adjust the infusion rate profile in real time. Readings from the breathalyzer, showing the BrAC (mg/L air) were converted to blood alcohol concentration (mg%, i.e., mg/100 mL blood) by applying the factor 230. This method has previously been shown to give a precise estimate for alcohol administered IV (Jones et al., 1997).

Measures

Alcohol use severity measures

A semi-structured clinical interview led by a medical doctor reviewed whether participants met the criteria for AUD, as specified in DSM-5 (American Psychiatric Association, 2013). Study participants were defined to have a current AUD if they met two or more out of 11 diagnostic criteria. A 30-day TLFB assessed drinking quantity and frequency (Sobell & Sobell, 1992). Hazardous drinking on a weekly basis was defined as consuming >14 standard drinks of alcohol (12 g) per week for men and >9 standard drinks of alcohol per week for women, consistent with guidelines from The Public Health Agency of Sweden. Participants also completed the AUDIT (Babor et al., 1989).

Nicotine use severity measures

The degree of nicotine dependence was assessed using Fagerström Test of Nicotine Dependence (FTND) and a 30-day TLFB assessed quantity and frequency of tobacco use (Heatherton et al., 1991).

Subjective responses to infused alcohol measures

Stimulation and sedation in response to alcohol were assessed with the Biphasic Alcohol Effects Scale (BAES; Martin et al., 1993). Also, the effect of drug (alcohol) in terms of (I) experiencing a drug effect, (II) liking the effects, (III) feeling a “high” and (IV) wanting more, were assessed with the Drug Effects Questionnaire (DEQ; Morean et al., 2013). Craving for alcohol was assessed using a visual analogue scale where participants rated their craving for alcohol from 1 to 10 on a sliding scale.

Outcome measures

FA IV-ASA outcome variables were the mean BrAC, defined as the arithmetic mean of all BrAC values and the peak BrAC throughout the FA phase as estimated by CAIS. Binge drinking was defined as reaching a peak BrAC ≥80 mg%. Depending on the number of button presses for alcohol (1 to 6 or 7 to 12) during the first 30 min of the FA phase, participants were classified as low or high responders, respectively, as previously described (Stangl et al., 2017). Correlations between self-administration variables and B-PEth sampled on the day of the experiment constituted the primary analysis of the outcome measures. The secondary objectives were examined using correlations between the following variables: subjectively rated effects of infused alcohol as measured by BAES, DEQ; changes in alcohol craving as rated on a VAS-scale; recent alcohol and tobacco use as reported in TLFB; gut-brain peptide levels in serum collected on the day of the experiment.

Blood samples and biochemical analyses

B-PEth (16:0/18:1) and gut-brain peptides, that is, total ghrelin and active GLP-1 without the addition of enzyme inhibitors, were collected in blood and serum on the day of the experiment. A subset of blood samples from each participant was sent to an external laboratory, Klinisk Kemi Malmö/Lund, Lund University Hospital, for rapid analysis of B-PEth using liquid chromatography–mass spectrometry as previously described (Walther et al., 2015). The remainder of blood and serum samples were first stored at −80°C and at a later stage collected for biochemical detection and quantification of gut-brain peptides at an in-house laboratory. In brief, a fluorescence bio-Plex Pro^™ Assay (171A7001M, Bio-Rad) was used to analyze serum samples in duplicates according to manufacturer instructions. The fluorescence intensity for each well was detected using a Bio-Plex 200 systems (Bio-Rad). Sample concentrations were determined from a standard curve using a five-parameter logistic nonlinear regression model. Mean intra-assay %CV is summarized in Table S2. There were no data under the minimum detection level of assays for ghrelin (3 pg/mL), but for GLP-1 (12 pg/mL), which conceivably stem from fast degradation in serum mediated by dipeptidyl peptidase-4 (DPP-4; Mentlein, 2009). B-PEth, mean, and peak BrAC did not differ between participants with detectable or invalid values of GLP-1 (Table S3).

Statistical analysis

Parametric (Pearson correlation) versus nonparametric tests (Spearman correlation and Mann–Whitney test) were applied for parametric versus nonparametric data, respectively. Distribution in frequencies for nominal variables was assessed using the Fischer’s exact test. When applicable (i.e., for analyses of subjective alcohol effects), a Bonferroni–Holm correction was performed to control the family-wise error rate. Data were analyzed in IBM SPSS Statistics version 27 (IBM Corp) and graphs were thereafter plotted in GraphPad Prism version 9 (GraphPad Software).

RESULTS

Description of the study sample

The study sample comprised 38 adults (mean ± SD age = 46.6 ± 14.5 years, 13% female, 37% AUD) that presented a mean B-PEth of 0.20 ± 0.20 μmol/L, a mean BrAC of 66.1 ± 28.4 mg% and a peak BrAC of 105.3 ± 52.3 mg% during the self-administration period (Table 1). Individual BrAC trajectories are plotted in Figure S1. The total sample reported drinking on average 11.6 ± 7.0 days out of the previous 30, with an average of 4.0 ± 1.7 drinks reported per drinking day, as measured by TLFB. Further, eight out of 38 participants presented a weekly drinking pattern consistent with hazardous drinking (Table 1). Upon stratification of participants based on DSM-5 criteria met for AUD, the total AUDIT score and the frequency of a weekly hazardous drinking pattern or B-PEth levels exceeding 0.3 μmol/L, indicating high recent alcohol intake, were higher among AUD as compared to SD study participants (Table 1). However, no significant differences in absolute values for B-PEth, TLFB scores, or mean and peak BrAC between the groups were revealed (Table 1). Consequently, data were primarily analyzed for the full sample and thereafter for each subgroup in the subsequent analyses.

TABLE 1.

Demographics and alcohol-drinking characterization, stratified across the presence of AUD^a.

	AUD (n = 14)	SD (n = 24)	Comparison	Total (n = 38)
Female	1	4		5
Nicotine use	4	7		11
	Mean (SD)	Mean (SD)	t(df), pb; u, pc or pd	Mean (SD)
Age (years)	46.6 (14.5)	45.6 (12.6)	0.23(36), 0.821	46.0 (13.1)
AUDITe	11.9 (5.0)	6.1 (2.1)	54, <0.001 ***	8.2 (4.5)
DSM-5f score	3.4(1.5)	0.2 (0.4)	0.00, <0.001 ***	1.4 (1.8)
FTNDg	3.5 (3.6)	3.3 (3.0)	−0.14(9), 0.891	3.2 (3.1)
Recent drinking history: 30-day TLFBh
Total drinks	62.8 (48.8)	36.3 (16.2)	120.5, 0.152	46.1 (34.2)
Drinking days	12.7 (8.3)	11.0 (6.2)	0.74(36), 0.464	11.6 (7.0)
Drinks per drinking day	4.7 (2.0)	3.6 (1.4)	1.86(36), 0.071	4.0 (1.7)
Heavy drinking days	6.7 (7.4)	2.8 (2.1)	118.5, 0.135	4.2 (5.0)
Hazardous weekly drinking frequencyi (n)	6/14	2/24	0.035 *	8/38
IV-ASAj
Mean BrACk (mg%)	62.8 (27.8)	68.0 (29.1)	−0.54(36), 0.595	66.1 (28.4)
Peak BrAC (mg%)	102.3 (55.7)	107.0 (51.4)	−0.26(36), 0.793	105.3 (52.3)
Binge drinking frequencyl (n)	8/14	16/24	0.729	24/38
B-PEthm (μmol/L)	0.29 (0.28)	0.14 (0.10)	114.5, 0.203	0.20 (0.20)
B-PEth >0.3 frequencyn (n)	6/14	2/24	0.035 *	8/38
Ghrelino (pg/mL)	346.3 (190.9)	216.6 (132.2)	59, 0.036 *	265.7 (166.8)
GLP-lp (pg/mL)	90.7 (60.4)	91 (67.8)	55.5, 0.654	93.7 (63.6)

^aAlcohol use disorder.

^bAn independent samples t-test for equality of means was performed for variables that were normally distributed and with equal standard deviations.

^cA Mann–Whitney U test for equality of means was performed if assumptions above were violated.

^dDistribution in frequencies was assessed using a Fischer's exact test.

^eAlcohol use disorders identification test.

^fDiagnostic and statistical manual of mental disorders 5.

^gFagerström test for nicotine dependence.

^hTimeline follow-back.

ⁱDistribution of hazardous weekly drinking, defined as consuming >14 (men)/9 (women) drinks per week.

^jIntravenous alcohol self-administration.

^kBreath alcohol concentration.

^lDistribution of binge drinking episode defined as exceeding a peak BrAC >80 mg% during IV-ASA.

^mPhosphatidylethanol in blood.

ⁿDistribution of B-PEth levels exceeding 0.3 μmol/L, indicating a high recent alcohol intake.

^oTotal ghrelin levels in serum.

^pGlucagon-like peptide 1 level in serum.

^*p < 0.05

^***p < 0.001.

IV-ASA does not correlate to B-PEth

The mean (Figure 2A) and peak (Figure 2B) BrAC during free-access alcohol self-administration did not correlate to B-PEth in the full sample. The post hoc power for identifying a moderate association (0.4) at an alpha of 0.05 in this sample was estimated to be 0.67. Further, in subsequent analyses for study participants with or without on average mild AUD, no associations between BrACs and B-PEth were revealed (Figure 2C,D). When dividing participants into groups based on B-PEth levels indicative of low to moderate (≤0.3 μmol/L) or high (>0.3 μmol/L) recent alcohol intake, no differences in achieved mean (Figure S2A) or peak (Figure S2B) BrACs were identified in the full sample or in the subgroups SD and mild AUD (Figure S2C–F). Moreover, B-PEth levels did not differ between groups depending on a laboratory binge-level exposure (Figure S3; Table S4) or responder status (low or high; Table S8) during the experimental session.

FIGURE 2.

Intravenous alcohol self-administration (IV-ASA) measures did not correlate to blood phosphatidylethanol (B-PEth). (A) The mean and (B) peak BrAC did not correlate to B-PEth in the full sample. In subsequent analyses for participants with or without AUD, no associations between (C) mean and (D) peak BrACs and B-PEth were revealed. A total number of six tests were conducted.

Associations between IV-ASA and TLFB

In the full sample, no associations were identified between mean and peak BrAC achieved during the experimental session and the self-reported, preceding total number of drinks consumed (Figure 3A,B), drinking days (Figure 3C,D), drinks per drinking day (Figure 3E,F) or heavy drinking days (Figure 3G,H).

FIGURE 3.

Intravenous alcohol self-administration (IV-ASA) measures were not associated with self-reported drinking history in the full sample but for drinks per drinking day in SD. In the full sample that combined participants with mild AUD and SD, no associations were identified between mean and peak BrAC achieved during the experimental session and the self-reported, preceding (A, B) total number of drinks consumed, (C, D) drinking days, (E, F) drinks per drinking day or (G, H) heavy drinking days. The number of drinks per drinking day was correlated to peak BrAC in the SD but not in the AUD group (F) whereas no other associations were revealed upon subgroup analyses. A total number of 24 tests were conducted, *p < 0.05.

However, in subsequent subanalyses, the number of drinks per drinking day correlated to peak BrAC in the SD but not in the AUD group (Figure 3F), whereas no other associations were revealed (Figure S4). Moreover, participants that revealed a hazardous weekly drinking pattern as assessed by TLFB also presented a B-PEth indicative of high recent alcohol intake and its frequency was higher in the mild AUD subgroup. Nonetheless, BrACs did not differ between these subgroups (Figure S2). Finally, self-reported recent drinking history was fairly to strongly correlated with B-PEth when assessed in the full study sample (Figure S5A–D). In subanalyses for groups based on AUD status, total drinks (Figure S5E) were strongly to moderately correlated to B-PEth in both AUD and SD groups, whereas drinking days (Figure S5F), drinks per drinking day (Figure S5G) and heavy drinking days (Figure S5H) were strongly correlated to B-PEth in the AUD but not in the SD group. AUDIT scores, which were higher for AUD as compared to SD participants, correlated to B-PEth (Figure S6) whereas no associations were evident between AUDIT total score or subscores and IV-ASA measures (Table S5).

Associations between IV-ASA and subjective effects of alcohol

Stimulatory and sedative effects of alcohol, as rated by BAES obtained prior to and after the first and second half of the FA self-administration period, did not correlate with mean or peak BrAC during the FA period, neither in the full sample nor in the mild AUD or SD subgroups (Table 2). The lack of association was also evident upon stratification of data depending on whether subjective ratings were obtained when the BrAC limb was ascending or descending BrAC (Tables S6 and S7). Subjective effects of infused alcohol were also rated using the DEQ (Table 3). A Bonferroni–Holm correction was applied to control for family-wise errors following multiple comparisons. At timepoint 25 min, that is, the timepoint between the alcohol priming period and the beginning of the FA period, and at timepoint 60 min, near halfway of the FA period, no correlations between mean or peak BrAC and a reported experience of feeling an effect of alcohol, liking the tentative effects, feeling a “high” or wanting more alcohol were observed (Table 3). At timepoint 100 min, directly after the FA period was terminated and marking the end of the experiment, experiencing an alcohol “high” was significantly correlated to mean and peak BrACs achieved in the total sample and SD participants (Table 3). By contrast, in participants with AUD, none of the DEQ items rated at timepoint 100 min, correlated with the achieved BrAC (Table 3). Moreover, upon stratification based on responder status during the early phase of IV-ASA, no associations were identified (Tables S8, S9, and S10).

TABLE 2.

Correlation between subjectively rated stimulatory and sedative effects of alcohol as measured by BAES^a versus mean and peak BrAC^b during the self-administration period.

	AUDc, n = 14	SD, n = 24	Total, n = 38	AUD, n = 14	SD, n = 24	Total, n = 38
	Spearman correlation coefficient BAES versus mean BrACd			Spearman correlation coefficient BAES versus peak BrAC
BAES stim.e after priming timepoint 25 min	r = 0.11 p = 0.707	r = −0.02 p = 0.924	r = −0.01 p = 0.963	r = 0.18 p = 0.545	r = −0.02 p = 0.923	r = 0.04 p = 0.816
BAES stim. timepoint 60min	r = 0.28 p = 0.338	r = 0.28 p = 0.181	r = 0.26 p = 0.116	r = 0.35 p = 0.218	r = 0.26 p = 0.217	r = 0.28 p = 0.084
BAES stim. after experiment timepoint 100min	r = 0.17 p = 0.565	r = 0.25 p = 0.230	r = 0.20 p = 0.224	r = 0.24 p = 0.402	r = 0.28 p = 0.194	r = 0.25 p = 0.126
BAES sed.f after priming timepoint 25 min	r = −0.23 p = 0.423	r = −0.21 p = 0.333	r = −0.27 p = 0.099	r = −0.26 p = 0.371	r = −0.17 p = 0.432	r = −0.24 p = 0.153
BAES sed. timepoint 60min	r = −0.12 p = 0.685	r = −0.13 p = 0.532	r = −0.17 p = 0.299	r = −0.20 p = 0.480	r = −0.11 p = 0.601	r = −0.16 p = 0.343
BAES sed. after experiment timepoint 100min	r = −0.04 p = 0.902	r = −0.07 p = 0.737	r = −0.07 p = 0.680	r = −0.12 p = 0.685	r = −0.06 p = 0.772	r = −0.07 p = 0.663

Note: A total number of 36 tests were conducted.

^aBiphasic alcohol effects scale.

^bBreath alcohol concentration.

^cAlcohol use disorder.

^dBreath alcohol concentration.

^eStimulatory alcohol effects.

^fSedative alcohol effects.

TABLE 3.

Correlation between subjectively rated effects of alcohol as measured by DEQ^a versus mean and peak BrAC^b during the self-administration period.

	AUDc, n = 14	SD, n = 24	Total, n = 38	AUD, n = 14	SD, n = 24	Total, n = 38
	Spearman correlation coefficient DEQ vs mean BrAC			Spearman correlation coefficient DEQ vs peak BrAC
DEQ FEELd, after priming timepoint 25 min	r = −0.40 p = 0.160	r = −0.06 p = 0.800	r = −0.17 p = 0.323	r = −0.33 p = 0.252	r = −0.08 p = 0.727	r = −0.17 p = 0.332
DEQ FEEL timepoint 60min	r = −0.13 p = 0.653	r = 0.40 p = 0.052	r = 0.20 p = 0.226	r = −0.20 p = 0.502	r = 0.29 p = 0.173	r = 0.09 p = 0.583
DEQ FEEL after experiment timepoint 100min	r = 0.18 p = 0.547	r = 0.50 p = 0.013	r = 0.41 p = 0.011	r = 0.20 p = 0.502	r = 0.41 p = 0.049	r = 0.35 p = 0.031
DEQ LIKEe after priming timepoint 25 min	r = −0.25 p = 0.391	r = −0.07 p = 0.750	r = −0.15 p = 0.266	r = −0.25 p = 0.391	r = −0.11 p = 0.606	r = −0.16 p = 0.359
DEQ LIKE timepoint 60min	r = −0.02 p = 0.958	r = 0.28 p = 0.191	r = 0.18 p = 0.268	r = 0.02 p = 0.935	r = 0.22 p = 0.304	r = 0.16 p = 0.341
DEQ LIKE after experiment timepoint 100min	r = 0.21 p = 0.469	r = 0.47 p = 0.020	r = 0.36 p = 0.026	r = 0.22 p = 0.441	r = 0.42 p = 0.039	r = 0.39 p = 0.038
DEQ HIGHf after priming timepoint 25 min	r = −0.08 p = 0.776	r = −0.05 p = 0.822	r = −0.09 p = 0.591	r = −0.01 p = 0.970	r = −0.09 p = 0.695	r = −0.10 p = 0.558
DEQ HIGH timepoint 60min	r = 0.24 p = 0.409	r = 0.55 p = 0.005	r = 0.42 p = 0.008	r = 0.20 p = 0.503	r = 0.48 p = 0.018	r = 0.36 p = 0.026
DEQ HIGH after experiment timepoint 100min	r = 0.54 p = 0.046	r = 0.79 p < 0.0001 *	r = 0.67 p < 0.0001 *	r = 0.44 p = 0.111	r = 0.73 p < 0.0001 *	r = 0.60 p < 0.0001 *
DEQ MOREg after priming timepoint 25 min	r = −0.14 p = 0.631	r = 0.31 p = 0.142	r = 0.12 p = 0.471	r = −0.08 p =0.776	r = 0.27 p = 0.204	r = 0.11 p = 0.507
DEQ MORE timepoint 60min	r = 0.09 p = 0.759	r = 0.60 p = 0.002	r = 0.45 p = 0.005	r = 0.17 p = 0.563	r = 0.56 p = 0.004	r = 0.44 p = 0.006
DEQ MORE after experiment timepoint 100min	r = 0.21 p = 0.464	r = 0.44 p = 0.033	r = 0.37 p = 0.021	r = 0.25 p = 0.383	r = 0.43 p = 0.038	r = 0.38 p = 0.019

A total number of 72 tests were conducted, * significant after controlling for multiple comparisons, criterion alpha 0.00072.

^aDrug effects questionnaire.

^bBreath alcohol concentration.

^cAlcohol use disorder.

^dFeeling an alcohol effect.

^eLiking the alcohol effect.

^fFeeling high from the alcohol consumed.

^gWanting to consume more alcohol.

Finally, the mean and peak BrACs achieved correlated with changes in self-reported craving for alcohol throughout the FA period (Figure 4A–D). In the SD group, this association was evident in the first half of the FA period (Figure 4E,F) whereas in the mild AUD group, the mean and peak BrACs achieved instead correlated with changes in alcohol craving score during the second half of the FA period (Figure 4G,H).

FIGURE 4.

The mean and peak BrAC were associated with changes in alcohol craving with differential timing for participants with AUD and SD. The mean and peak BrACs achieved correlated with changes in self-reported craving for alcohol during the (A, B) first (25 to 60 min) and (C, D) second half (60 to 100 min) of the free-access (FA) period in the full sample. (E) Mean and (F) peak BrACs were correlated with changes in craving for alcohol during the first half of the FA period in the SD, but not in the AUD subgroup. During the second half of the FA period, (G) mean and (H) peak BrACs were instead correlated with changes in alcohol craving score in the AUD but not in the SD group. A total number of 12 tests were conducted, *p < 0.05, **p < 0.01, ***p < 0.001.

Associations between IV-ASA and gut-brain peptides

Ghrelin serum levels were significantly higher among study participants with AUD, as compared to SD controls (Figure S7A), whereas GLP-1 serum levels did not differ between the two groups (Figure S7B). As depicted in Figure S7C–H, there were no significant relationships between baseline serum levels of the gut-brain peptides and baseline B-PEth, or with mean and peak BrACs achieved during FA IV-ASA. However, a nonsignificant trend for a small to moderate positive correlation between ghrelin levels and B-PEth was observed (Figure S7C).

DISCUSSION

This is to our knowledge the first study to implement an objective marker for recent alcohol intake, combined with a high safety ceiling for achieved BrACs (200 mg%), in order to examine how FA IV-ASA reflects recent alcohol consumption in SD and individuals with AUD. Mean and peak BrACs during FA IV-ASA correlated to changes in craving for alcohol but not to recent drinking history assessed either by TLFB or by B-PEth. Interestingly, in subanalyses for SD and mild AUD participants, changes in alcohol craving were associated with IV-ASA measures at distinct time intervals in AUD and SD groups, respectively. Also, a subset of subjectively rated hedonic alcohol effects and recent alcohol intake, but not the recent drinking marker B-PEth, correlated with BrACs in SD, whereas no associations were identified in AUD participants. Total ghrelin levels were higher in AUD participants but did not correlate to IV-ASA parameters in either group.

Study sample

The current study included both SD and a smaller sample of AUD participants where the latter group, by definition, presented higher, albeit modest, DSM-5 symptom counts indicating an on average mild AUD. Moreover, AUDIT scores, the frequency of a hazardous weekly drinking pattern, and the frequency of B-PEth levels indicative of high or excessive alcohol intake were also higher in AUD participants as compared to SD. In contrast to what was expected, recent alcohol consumption as measured by TLFB, absolute B-PEth levels, and achieved BrACs during IV-ASA did not significantly differ between the two groups. This lack of differences is likely explained by the small sample size, combined with the distinct heterogeneity for these measures in the AUD sample.

Associations between IV-ASA and craving for alcohol

Alcohol craving constitutes one of the hallmarks of AUD and may also be targeted for its treatment (Koob & Volkow, 2016). In the present study, changes in craving scores correlated with achieved BrACs, as also evident in a previous PR IV-ASA study in heavy drinkers (Bujarski et al., 2018). However, upon stratification by AUD status, this association was more pronounced and manifested itself in the first half of the FA phase in SD, as compared to the second half of the experiment in mild AUD participants. Assuming that AUD participants reach higher BrACs when drinking outside the lab, it follows that BrAC levels achieved later in the experiment, which were on average higher, may better resemble habitual drinking and may thus explain the present results. Accordingly, it could be argued that IV-ASA may be employed for screening interventions tailored to interfere with alcohol craving. Next step to further explore this application would be to evaluate the effects of approved medications to treat AUD that target alcohol craving, on achieved BrACs and subjectively rated craving for alcohol in the IV-ASA model (Kranzler & Soyka, 2018).

Associations between IV-ASA and hedonic alcohol effects

Subjective ratings of feeling “high” from the alcohol infused at the end of the FA period were associated with mean and peak BrACs. In subanalyses for the SD group, this association persisted. This is partly consistent with findings in a previous FA IV-ASA study in SD where perceptions of rewarding effects were associated with alcohol infused following priming and during the first part of the FA period (Stangl et al., 2017). By contrast, in mild AUD participants, no relationship could be identified between ratings of hedonic alcohol effects and BrACs. This relationship has not, to our knowledge, been previously examined for the FA paradigm in an AUD population. However, in a study that used the PR paradigm in AUD participants, neither DEQ nor BAES measures were correlated to mean BrACs (Farokhnia et al., 2018). Moreover, the present study could not identify any relationship between the stimulatory or sedative effects of the alcohol consumed and the BrACs achieved during IV-ASA. This absence was also evident when adjusting for the timing when ratings were obtained, that is, during an ascending or descending slope of the alcohol rewards, recognized to be closely linked to respective biphasic alcohol effects (Hendler et al., 2013). The overall lack of subjective responsiveness in relation to BrACs in the AUD group may possibly be related to a higher degree of tolerance to the pharmacological effects of alcohol. A previous IV-ASA study in SD has indeed shown that participants who more frequently press for alcohol during the first 30 min, referred to as high responders, also present greater hedonic responses to the alcohol infused (Stangl et al., 2017). Hence, the early phase of IV-ASA has been suggested to be more sensitive to the rewarding and motivational properties of alcohol, plausibly due to less influence by acute sensitization or acute tolerance to alcohol (Stangl et al., 2017). Nonetheless, besides higher BrACs for the total session among high responders, no such differential effects could be identified in the present study sample. The fact that changes in alcohol craving, but not hedonic alcohol effects, were associated with IV-ASA BrACs in the AUD subgroup, is also partly in line with the incentive sensitization theory, as conceptualized by Berridge and Robinson (2016), which posits brain “wanting” systems to be hyper-reactive to alcohol cues whereas alcohol “liking” may be reduced in AUD.

Associations between IV-ASA and self-reported drinking history

In order to further explore whether alcohol consumption in the FA IV-ASA paradigm reflects real-life drinking, the relationship between BrACs and recent alcohol intake measured by TLFB was examined. Contrary to what was hypothesized, BrACs were not correlated with TLFB measures in the full study sample. However, in line with previous FA IV-ASA studies, peak BrACs were correlated with TLFB measures in SD (Junger et al., 2016; Stangl et al., 2017). This association was not evident in our participants with AUD, a rather small and heterogenous study sample presenting an on average mild AUD. These AUD subsample results are consistent with a previous study employing the FA paradigm in which TLFB did not predict BrACs in heavy drinkers although in a study design where 61% of participants reached the BrAC safety limit (Vatsalya et al., 2015).

IV-ASA does not correlate to an objective alcohol marker of recent drinking

Notably, the present study could not identify any relationship between FA IV-ASA-derived outcome measures and recent alcohol intake as measured by B-PEth. The lack of association persisted when examined separately in the SD and mild AUD groups. However, the study design allowed for several weeks to elapse between the telephone interview, screening visit, and experimental session, respectively, which may explain the lack of an association, as stigmatized behavior including excessive drinking may have been reduced following screening. For example, three out of 14 participants with AUD, self-administered to a peak BrAC of less than 50 mg%, which was unexpected. It could also be argued that B-PEth, despite its high sensitivity and specificity as an alcohol marker, might have been less suited for indicating recent alcohol intake in the present AUD sample that mostly encompassed mild AUD (nine out of 14 participants; Isaksson et al., 2011). B-PEth reflects alcohol intake for the preceding 2 weeks and less frequent heavy drinking episodes may, in theory, pass unnoticed. Nonetheless, B-PEth levels were in line with previous studies tightly correlated with self-reported drinking (Isaksson et al., 2011). Also, upon stratification based on either B-PEth intervals applied in the clinic to indicate heavy recent alcohol consumption or self-reported hazardous weekly drinking pattern, no differences in achieved BrACs could be observed. Thus, based on the findings from the current study, B-PEth does not appear to identify participants that achieve high BrACs in the IV-ASA, FA paradigm. Since our FA IV-ASA measures in AUD participants reflected neither biomarker nor self-reported measurements of recent alcohol intake, the ability of the method to replicate real-life drinking in the AUD group may be limited. However, a definitive conclusion warrants a larger study sample including more participants with moderate and severe AUD, and careful examination of participant motivation.

Further methodological aspects

Moreover, it could be argued that the FA paradigm applied in the present study may be inferior to the PR paradigm, in terms of modeling alcohol consumption behavior in AUD participants. The FA paradigm has previously been linked to alcohol liking whereas the PR paradigm, which demands a successively increasing effort for the delivery of alcohol, has been associated with alcohol wanting (Cyders et al., 2021). According to the three-stage theory of addiction, alcohol liking is indicative of the binge/intoxication phase, whereas alcohol wanting represents the later occurring preoccupation/anticipation phase more consistent with an AUD (Cyders et al., 2021; Koob & Volkow, 2010). Thus, self-administration under the PR paradigm, reflective of alcohol wanting, may be preferable to use in AUD participants. Indeed, a study employing the PR paradigm and a longer self-administration phase in a mixed AUD and SD population identified that alcohol use severity predicted BrACs, lending some external validity to the IV-ASA model (Bujarski et al., 2018). Moreover, since changes in alcohol craving correlated with BrAC in AUD participants later in the present experiments, when BrACs were on average higher, it is also plausible that the duration of the FA phase was too short for studying the AUD population.

Several other limitations to the current pilot study should also be pointed out. Besides the analyses for subjective alcohol effects, which entailed a rather extensive number of statistical tests, correction for family-wise errors was not conducted. Furthermore, the AUD sample size in the present pilot study was small, included few females, and contained highly variable but on average modest scores for alcohol use severity and recent drinking history measures. Consequently, B-PEth and TLFB measures substantially overlapped with the SD group and altogether limited data interpretation for the AUD sample per se, and for comparisons between the two groups. Furthermore, the present study did not entail adjustments for factors besides recent drinking history that have been shown to impact FA IV-ASA outcomes, for example, personality traits and family drinking history (Gowin et al., 2017; Sloan et al., 2020; Vaughan et al., 2019; Zimmermann et al., 2009). Further studies examining the association between alcohol biomarkers and IV-ASA measures in AUD populations are therefore warranted.

Future directions

Although the FA IV-ASA model appears to variably reflect real-life drinking, as measured by TLFB or B-PEth, it might still possess a predictive validity in terms of screening new treatments for AUD. Therefore, next step would be to evaluate compounds with a documented effect on AUD, that is, naltrexone, nalmefene, or acamprosate in an FA model, which has recently been explored for naltrexone using the PR model in nondependent subjects (Spreer et al., 2023). Notably, treatment with varenicline, which reduces drinking in AUD independent (de Bejczy et al., 2015; Litten et al., 2013) or dependent (McKee et al., 2009) of smoking status, did not reduce BrACs in a recent FA IV-ASA study on heavy drinkers, which, however, may be explained by the relatively low safety ceiling (120 mg%) employed (Vatsalya et al., 2015). Indeed, the present study demonstrated higher BrACs than have been reported previously (Stangl et al., 2017; Vatsalya et al., 2015), which presumably was enabled by the high safety ceiling (200 mg%), here shown to be safe and well-tolerated. Hence, it is suggested that a ceiling as high as 200 mg% may hereafter be implemented in IV-ASA studies in AUD participants, enabling a more accurate reflection of BrACs achieved in real life.

Associations between IV-ASA and gut-brain peptides

Finally, serum levels of total ghrelin and GLP-1, two gut-brain peptides with associated receptor systems implicated in modulating alcohol intake in humans, were compared in order to explore whether similar effects were present in the FA IV-ASA model (Farokhnia et al., 2019; Shevchouk et al., 2021). In this study, ghrelin levels were higher among AUD participants, a finding that overall corroborates previous studies where ghrelin was elevated in a sample of individuals with AUD and in alcohol consumers, as compared to abstainers in a SD population (Farokhnia et al., 2021; Kraus et al., 2005). Further, a nonsignificant trend for an association between ghrelin and B-PEth was noted, whereas previous associations between self-reported drinking and ghrelin levels have been inconsistent (Farokhnia et al., 2021; Wittekind et al., 2018). Notably, a previous study has shown that exogenously administered ghrelin increases craving for PR IV self-administration of alcohol (Farokhnia et al., 2018; Leggio et al., 2014). In the present study, ghrelin levels were not associated with BrACs, suggesting that endogenous levels versus exogenous delivery of the peptide may exert differential effects on alcohol intake. As investigated for the first time, to the best of our knowledge, GLP-1 levels did not differ between AUD and SD participants and were not correlated to B-PEth or BrACs. In a retrospective analysis of FA IV-ASA data, allele variations of the GLP-1 receptor influenced achieved BrACs (Suchankova et al., 2015). In light of our study, one may hypothesize that the receptor rather than the peptide itself impacts IV-ASA outcomes. It should be noted that the present study did not entail the addition of DPP-4 inhibitors to serum, which decelerates degradation from active GLP-1 to its metabolites. Nonetheless, serum levels of active GLP-1 could indeed be quantified in the majority of samples that are conceivably equally exposed to degradation and therefore would not preclude the interpretation of the correlational data.

CONCLUSION

In conclusion, these preliminary findings suggest that an FA IV-ASA paradigm employing the highest limit for BrACs so far reported does not reflect recent drinking as estimated by B-PEth, an objective marker of recent alcohol intake, in a small sample of AUD participants, in SD, or in a combined sample. However, in SD, but not in the mild AUD group, the self-reported number of drinks per drinking day correlated to peak BrACs, whereas no other TLFB-derived measure was associated with BrACs. Finally, the paradigm appears to identify alcohol-induced craving with differential timing for AUD and SD participants. Further studies on associations between alcohol biomarkers and IV-ASA behavior that encompass a larger AUD study sample are needed. Nevertheless, variable reports on associations with real-life drinking do not exclude a tentative predictive validity for the IV-ASA method in terms of identifying pharmacological treatments for AUD. In order to further explore the role of IV-ASA as a translational tool to screen and prioritize new interventions to treat AUD, future studies that evaluate the effects of approved treatments in the IV-ASA model, preferably aimed at alcohol craving, are warranted.

Supplementary Material

SUPPORTING INFORMATION

Additional supporting information can be found online in the Supporting Information section at the end of this article.