Thirty-eight ALC (including eight females) between the ages of 25 and 60 years were referred by the VA Medical Center Substance Abuse Day Hospital and the Kaiser Permanente Chemical Dependence Recovery Program in San Francisco. In this naturalistic study design, 13 of these ALC had received daily oral GBP for at least 1 week before study (GBP+), whereas 25 ALC had not received GBP (GBP−) from their treatment provider at any time before study. The duration of GBP treatment before study and the daily dose were estimated from information obtained from treatment providers and corroborated by patient self-report. Given the nature of the data obtained and the fact that this was not a prospective treatment trial, only the daily dose data allowed some quantitative analyses; the GBP duration data were sufficiently incomplete for most patients to allow computing reliable quantitative estimates beyond the statement that all patients received GBP for at least 1 full week before study. Sixteen CON similar in age (48.7 ± 12.3 years) to our ALC group were recruited from the local community and studied contemporaneously with the ALC participants.

All study participants provided written informed consent prior to study per the Declaration of Helsinki and underwent procedures approved by the Institutional Review Board of the University of California San Francisco and the San Francisco VA Medical Center. Primary inclusion criteria for ALC participants were fluency in English, DSM-IV diagnosis of alcohol dependence or abuse at the time of enrollment, consumption of >150 standard alcohol-containing drinks (i.e., 13.6 g of ethanol/drink) per month for at least 8 years prior to enrollment for men, or consumption of >80 drinks/month for at least 6 years prior to enrollment for women. Exclusion criteria for ALC participants were history of the following: dependence on any substance other than alcohol or nicotine in the 5 years prior to enrollment, any intravenous drug use in the 5 years prior to study, current opioid agonist therapy, intrinsic cerebral masses or vascular malformations, HIV/AIDS, cerebrovascular accident, myocardial infarction, uncontrolled chronic hypertension, type-1 diabetes or insulin dependent conditions, moderate or severe COPD, non-alcohol related seizures, significant exposure to known neurotoxins, demyelinating and neurodegenerative diseases, documented Wernicke–Korsakoff syndrome, alcohol-induced persisting dementia, any penetrating head trauma, and closed head injury with loss of consciousness >10 min. Exclusion criteria also included history of schizophrenia-spectrum and bipolar disorders, PTSD, obsessive-compulsive disorder, and panic disorder. Hepatitis C, type-2 diabetes, hypertension, and unipolar mood disorder (major depression and/or substance-induced mood disorder) were permitted in the ALC cohort given their high prevalence in alcohol use disorders (AUDs) (25, 26). For ALC, inclusion/exclusion criteria were verified via electronic medical records, when available.

The ALC participants were involved in outpatient treatment at the time of assessment and none tested positive for alcohol or illicit substances (cannabinoids, cocaine, amphetamines, opioids, or PCP) before any of the assessments. Participants could be scanned only after their Clinical Institute Withdrawal Assessment for Alcohol Revised (CIWA-Ar) scores were below 10 (mild alcohol withdrawal is defined by a CIWA-Ar score less than or equal to 15). GBP− and GBP+ groups had a similar average duration of abstinence when studied. All CON participants were fluent in English, had no history of any DSM-IV Axis I disorder or biomedical condition that may have adversely affected their neurobiology, and they had not endorsed averaging >50 standard alcohol-containing drinks during any month over lifetime.

All ALC participants completed the Structured Clinical Interview for DSM-IV Axis I Disorders, Version 2.0 and CON were administered the accompanying screening module. Semi-structured interviews for lifetime alcohol consumption [Lifetime Drinking History (LDH)] (27, 28) and substance use (in-house questionnaire assessing substance type and quantity and frequency of use) (29) were administered to all participants. From the LDH, average number of alcoholic drinks per month over 1 year prior to enrollment, average number of drinks per month over lifetime, duration of heavy drinking (here operationalized as drinking >100 drinks per month in males, >80 drinks per month in females) were calculated. Premorbid verbal intelligence was estimated with the American National Adult Reading Test. Participants also completed standardized questionnaires assessing depressive (Beck Depression Inventory) (30) and anxiety symptoms (State-Trait Anxiety Inventory) (31), self-reported impulsivity (Barratt Impulsivity Scale) (32), and nicotine dependence via the Fagerstrom Tolerance Test for Nicotine Dependence (33). Table 1 shows demographics, behavioral measures, alcohol, and cigarette consumption variables for all GBP+ and GBP− participants. A standard neurocognitive battery was administered, test specific Z-scores were calculated based on corresponding normative data, cognitive domains were formed from specific neurocognitive tasks as previously described, and summary scores were calculated for each cognitive domain and global cognition (34).

MRI data were acquired on a 4 T MedSpec system using an 8-channel transmit-receive head coil and a Siemens Trio console (Siemens, Erlangen, Germany). A magnetization prepared rapid gradient (TR/TE/TI = 2,300/3/950 ms, 7° flip angle, 1 mm × 1 mm × 1 mm resolution) and a turbo spin-echo (TR/TE = 8,400/70 ms, 150° flip angle, 0.9 mm × 0.9 mm × 3 mm resolution) sequence were used to acquire 3-D sagittal T1-weighted and 2-D axial T2-weighted anatomical images, respectively. The images were then displayed on the scanner console for the prescription of up to five different MRS volumes-of-interest (VOIs) acquired in two scanning sessions (for typical VOI locations see Figure 1). After VOI-specific 3-D shimming and optimization of water suppression, J-edited GABA spectra were acquired from the POC (size: 40 mm × 20 mm × 20 mm) using a modified MEGA PRESS sequence (TR/TE = 2,000/71, editing frequencies at 1.9 and 7.5 ppm, 12.5 min acquisition time) (35), followed by the acquisition of a water spectrum from the same VOI. This was then followed by the acquisition of a stimulated echo acquisition mode (STEAM) sequence (TR/TE/TM = 2,000/12/10 ms, 2.5 min acquisition time) from the same VOI location. Additional STEAM metabolite and water spectra were acquired from the ACC (35 mm × 25 mm × 20 mm) and the right DLPFC (40 mm × 20 mm × 20 mm), with VOIs placed to maximize the inclusion of as much gray matter (GM) tissue as possible. For this analysis, GABA spectra from the ACC and DLPFC were available from all GBP− patients but only from two GBP+ patients, so that potential GABA differences in these VOIs could not be analyzed statistically.

In a second scan of all GBP+ patients generally within two days of the first scan, we further acquired metabolite and water STEAM spectra from the right basal ganglia (40 mm × 25 mm × 20 mm, including the head of the caudate, lenticular nucleus, and anterior thalamus) and the left frontal WM (25 mm × 15 mm × 20 mm), placed to include as much frontal WM tissue as possible (see Figure 1). Metabolite concentrations from these two subcortical VOIs were compared to the corresponding spectra available from a representative subgroup of 15 GBP− patients with characteristics like those of the entire GBP− group; these subcortical spectra had been obtained in the previous year on the same scanner, with the same software version and with identical experimental parameters. The five brain VOIs were chosen because of our group’s longstanding interest in regional metabolite concentration changes during abstinence from alcohol [for review see Meyerhoff (13)].

All metabolite and water MRS data were processed using established routines of a combination of in-house Matlab programs and IDL (Research Systems, Inc., Boulder, CO, USA) with SITOOLS (14, 36), yielding peak areas for total NAA, Glu, total Cr, total Cho, mI, and water for all five VOIs and additional GABA signal from cortical VOIs. At 4 T magnetic field strength, the Glu signal can be measured separately from the glutamine signal, and in our implementation of the GABA editing method, the measured GABA signal has some contribution from co-edited macromolecules. The T1 images obtained at the time of MRS were segmented into GM, WM, and cerebrospinal fluid (CSF) (37) and yielded estimates of the tissue fraction and CSF contributions to each VOI; these measures did not differ between GBP− and GBP+ for any of the analyzed VOIs. The individual spectral and segmentation information was then merged to calculate absolute molar concentrations [in institutional units (i.u.), i.e., without making assumptions about metabolite relaxation times] for metabolites in all VOIs, with details of these procedures described earlier (14).

Demographic and clinical variables were compared between GBP+ and GBP− groups with t-tests and Fisher’s exact test, where indicated. We previously reported extensively on metabolite group differences between ALC (here called GBP−) and CON (14–16); consistent with those reports, in this new cohort, both NAA (p < 0.04) and Cr (p < 0.09) were lower in all three cortical VOIs of the entire ALC cohort of this report vs. CON. In this report, we focus on simple pairwise comparisons of region-specific metabolite levels between GBP+ and GBP−. Any such significant differences were then further explored by pairwise comparisons to CON. Because of our specific a priori hypotheses for group comparisons (cortical GABA elevated, Glu unchanged), an uncorrected p < 0.05 was considered statistically significant. In exploratory analyses, Pearson correlations were calculated for associations of behavioral and GBP dose data with spectral outcome measures and corresponding partial correlations were computed where indicated.

The GBP+ group was not significantly different from the GBP− group on sex distribution, average age, predicted premorbid intelligence, depressive and anxiety symptoms, self-reported impulsivity, common measures of lifetime alcohol consumption, and duration of abstinence from alcohol at the time of MR scan. The groups also did not differ significantly on cognitive domain summary scores (executive function, cognitive efficiency, processing speed, working memory, audioverbal learning and memory, visuospatial learning and memory, visuospatial skills, fine motor skills; all p > 0.29) or global cognition (see Table 1). The proportion of cigarette smokers was approximately equal between the GBP− (68%) and GBP+ groups (54%) and cigarette pack years as well as duration of lifetime smoking did not differ between the groups (all p > 0.25). However, the GBP+ had a lower Fagerstrom score than the GBP− group (p = 0.05), indicating low vs. moderate nicotine dependence, respectively.

At the time of the MR scan, GBP+ patients had received at least 1 week of oral GBP in three divided doses each day at a median daily dose of 1,200 mg (range: 800–3,600 mg, with 85% receiving between 800 and 1,800 mg; mean ± SD: 1,600 ± 900 mg). The large range of doses reflects different practices across treatment centers, different individual tolerance to the medication, and generally missing guidelines for the pharmacological treatment of alcohol dependence during early abstinence.

We compared cortical data from 13 GBP+ and 25 GBP− patients (unless otherwise specified). In the POC, none of the measured absolute metabolite concentrations (GABA, Glu, NAA, Cr, Cho, and mI) differed between GBP+ and GBP− (all p > 0.30). In the ACC, Glu, NAA, Cr, and Cho were equivalent between GBP+ and GBP− (all p > 0.20), while mI tended to be lower in GBP+ (p = 0.08). In the DLPFC, Glu, NAA, Cr, Cho, and mI were not different between GBP+ and GBP− (all p > 0.32). GABA concentrations from the ACC and DLPFC were only obtained in two GBP+ participants each and their values were within the range of the values obtained in the 25 GBP− patients.

In the basal ganglia, similar to our findings in the ACC, the Glu, NAA, Cr, and Cho concentrations did not differ between 13 GBP+ and 15 GBP− patients (all p > 0.30), while mI tended to be lower in GBP+ (p = 0.08).

In the frontal WM, Glu was 15% lower in the 13 GBP+ vs. 15 GBP− patients (2.23 ± 0.39 vs. 2.61 ± 0.36, p = 0.017; Cohen’s d effect size = 0.99), while NAA, Cr, Cho, and mI were statistically equivalent between the groups (all p > 0.12). The lower frontal WM Glu in GBP+ remained significant after covarying for Fagerstrom total score, indicating that the Glu difference was independent of the severity of nicotine dependence. The lower frontal WM Glu of the GBP+ group showed a moderately strong, but non-significant correlation, with higher daily GBP dose (univariate r = −0.47, p = 0.12; partial r = −0.58, p = 0.13, with abstinence duration as covariate); by contrast, none of the cortical GABA and Glu concentrations correlated even weakly with GBP dose. The frontal WM Glu in GBP+ was not associated with any measure of lifetime alcohol consumption, smoking severity, or age.

In group comparisons of basal ganglia and frontal WM metabolite levels available from a subset of 10 of the 16 CON participants, frontal WM Glu was significantly higher in GBP− vs. CON (2.61 ± 0.36 vs. 2.20 ± 0.46, p = 0.026), whereas it was statistically equivalent between GBP+ and CON (2.23 ± 0.39 vs. 2.20 ± 0.46, p = 0.85). The combined group of ALC participants showed no such Glu difference compared to the subset of CON (p > 0.18). Furthermore, the GBP+ and GBP− groups did not differ from CON on any of the other regional metabolite concentrations.

All findings remained essentially the same when re-analyzing the data without the small number of women included in this U.S. Veterans sample. Figure 2 is a bar graph of Glu concentrations (i.u.) observed in the different VOIs across CON, GBP−, and GBP+ groups.

Within the ALC group (GBP+ and GBP− combined), all correlations of frontal WM Glu with the summary scores for cognitive domains and global cognition were positive (all r = 0.044–0.525), indicating that higher frontal WM Glu was associated with better cognitive function. This relationship was observed for both groups; also, recall that the GBP groups did not differ significantly on cognitive performance. The strongest and most significant correlations were observed with visuospatial memory (r = 0.525, p = 0.004) and visuospatial learning (r = 0.456, p = 0.015), while the corresponding correlations with cognitive efficiency and global cognition were at the trend level (r > 0.380, p < 0.075). None of the associations of frontal WM Glu with measures of self-reported impulsivity and anxiety was significant.