The study protocol was approved by the local ethics committee (Charité Universitätsmedizin Berlin; EA1/245/11), and all patients or their families/legal guardians gave written consent before participating in the study. A subsample of this proton magnetic resonance spectroscopy (1H-MRS) data has already been published together with 18F-fallypride positron emission tomography (PET) data [11]. This study was a component of a large, multi-site study examining the neurobiological and cognitive alterations associated with the onset and persistence of AUD (LeAD study; Project 5, The Role of Dopaminergic and Glutamatergic Neurotransmission for Dysfunctional Learning in Alcohol Use Disorders, https://gepris.dfg.de/gepris/projekt/209293518, clinical trial number: NCT02094196).

The study sample consisted of 96 participants divided into three groups: 31 LR individuals, 38 HR individuals, and 27 recently detoxified AD patients. The three groups were comparable in terms of age, handedness, and smoking status (see Table 1 (Ref. [12, 13, 14]) for detailed demographic information). Although we tried to match groups carefully, there was a trend towards a significant difference in the smoking status between groups which is why we included smoking as a covariate in our analyses. LR control participants were recruited from the general population, while AD patients were enrolled from various inpatient psychiatric clinics in Berlin. HR individuals were also recruited from the community and classified as such based on an Alcohol Use Disorders Identification Test (AUDIT) score exceeding eight [15]. To ensure diagnostic accuracy, all participants underwent a Composite International Diagnostic Interview (CIDI) to exclude alcohol dependence according to the Diagnostic and Statistical Manual of Mental Disorders 4th edition (DSM-IV) criteria in HR and LR groups and to rule out any history of other substance use disorders [16, 17].

All AD patients completed an inpatient detoxification program prior to study enrollment. Participants were diagnosed with AD according to both the International Statistical Classification of Diseases and Related Health Problems 10th revision (ICD-10) and DSM-IV criteria by a qualified clinician and had a minimum of three years of AD history. To minimize pharmacological influences, AD subjects were medication-free for at least four half-lives of any previously prescribed psychotropic drug and exhibited mild withdrawal symptoms (Clinical Institute Withdrawal Assessment for Alcohol (CIWA) - Score 3) at the time of magnetic resonance imaging (MRI) scanning. All subjects were scanned at one time point and at rest. Exclusion criteria encompassed past or present substance dependence beyond alcohol or nicotine (verified through urine screening), major psychiatric disorders, neurological conditions, and any contraindications for MRI.

MRI data were acquired using a 3 Tesla Verio scanner (Siemens Healthcare, Erlangen, Germany) equipped with a 32-channel receive-only head coil. Following a localization scan, high-resolution anatomical images were obtained using a three-dimensional T1-weighted magnetization-prepared rapid acquisition with gradient echo (MPRAGE) sequence. Image acquisition parameters included a repetition time (TR) of 2.3 ms, an echo time (TE) of 3.03 ms, a flip angle of 9°, a 256 $\times$ 256 $\times$ 192 matrix, and a voxel size of 1 $\times$ 1 $\times$ 1 mm³. Subsequently, a 25 $\times$ 35 $\times$ 20 mm³ voxel of interest (VOI) was positioned over the anterior cingulate cortex (ACC) for subsequent magnetic resonance spectroscopy (MRS). Prior to MRS data acquisition, first and second-order shims were optimized using the “fast automatic shimming technique by mapping along projections” (FAST(EST)MAP) method [18]. Subsequently, water signal linewidth was assessed, ensuring a value not exceeding 7 Hz in all cases. Finally, spectra were collected using the spin-echo full intensity-acquired localized (SPECIAL) technique [19]. The employed sequence combines Image-Selected In Vivo Spectroscopy (ISIS) with a slice-selective spin-echo technique [20]. Alternate scans utilize a slice-selective adiabatic inversion pulse. This was performed to achieve localization perpendicular to the spin-echo plane. Further, a subsequent spin-echo sequence, employing asymmetric 90° and 180° slice-selective refocusing pulses, localizes the signal within two dimensions. Signal subtraction between odd and even scans are performed. This method enables the acquisition of very short echo times, owing to the use of a single refocusing pulse, facilitating the detection of numerous metabolites and precise glutamate quantification [21, 22]. Water suppression was achieved using the variable power radio frequency pulses with optimized relaxation delays (VAPOR) technique [23]. Additionally, six outer volume suppression slices were positioned around the spectroscopic voxel to minimize signal contamination from surrounding tissues. A total of 256 scans were acquired for each metabolite spectrum, employing a TR of 3 seconds and a TE of 8.5 milliseconds. Immediately following this, a water-unsuppressed spectrum was recorded with eight averages. Water linewidth was calculated by fitting the water spectrum after shimming to 6.2 $\pm$ 0.4 Hertz. Metabolite levels were determined using linear combination (LC) model [24] with a simulated basis set of 20 metabolites, utilizing the unsuppressed water scan for eddy current correction. Mean Cramer-Rao lower bounds (CRLBs) for GABA and glutamate quantification were determined to be 16.0% and 4.4%, respectively. Metabolite signal intensities were corrected for relaxation effects using T1 and T2 relaxation times established at 3T [25, 26]. To account for cerebrospinal fluid (CSF) contamination within the voxel, Magnetization-Prepared Rapid Acquisition Gradient Echo (MPRAGE) images were segmented using statistical parametric mapping (SPM8) [27]. Voxel pixels were classified as CSF, gray matter, or white matter based on probability values generated by SPM8. See Fig. 1 for the voxel position and voxel size.

Fig. 1.

Proton magnetic resonance spectroscopy (1H-MRS) voxel position of the voxel of interest in the anterior cingulate cortex (ACC) with a voxel size of 25 $\mathrm{\times }$ 35 $\mathrm{\times }$ 20 mm³.

Statistical analyses were carried out using the Statistical Package for the Social Sciences (SPSS Statistics Version 29.0.0, IBM, Armonk, NY, USA). From the total sample size of 96 subjects GABA levels in the ACC were available for 92 subjects. Further, outliers were visually identified and consecutively excluded. We excluded two outliers in the LR, one in the HR and one in the AD group with a resulting sample size of 88 (LR = 28, HR = 35, AD = 25). After the exclusion of outliers’ normal distribution was achieved and statistically confirmed with the Kolmogorov-Smirnov Test.

We included smoking status as a covariate into the analyses as it has been shown that smoking seems to influence cortical GABA levels in HC as well as AD [28] and our groups differed marginally significant in their smoking status (see descriptive statistics, Table 1). We performed the group comparison of GABA levels in the ACC with an analysis of covariance (ANCOVA) and consecutive post hoc tests. To analyze the relationship between the alcohol use pattern and GABA levels, planned contrast weights were investigated for a (negative) quadratic (LR $>$ HR AD) relationship between GABA and group. The negative quadratic effect would support our hypothesis of a decrease of GABA levels in HR subjects and an alteration of GABA levels in abstinent AD subjects. Correlation with clinical scales such as abstinence duration, Obsessive-Compulsive Drinking Scale (OCDS), Alcohol Dependence Scale (ADS), or Hospital Anxiety and Depression Scale (HADS) were conducted using the Pearson correlation coefficient.

There was a significant effect of group (LR/HR/AD) on GABA levels in the ACC F(2, 84) = 4.08, p = 0.020. Further, we observed a significant quadratic trend F(2, 85) = 4.99, p = 0.030 throughout the sample, indicating a distribution of GABA levels as follows: LR $>$ HR AD. Post hoc tests (Tukey’s Honestly Significant Difference (Tukey HSD)) revealed significantly higher GABA levels in AD compared to HR subjects (p = 0.006), but neither between AD and LR subjects (p = 0.066), nor between LR and HR subjects (p = 0.372). Boxplots of GABA levels in the ACC in each group are shown in Fig. 2.

Fig. 2.

Boxplots of gamma-aminobutyric acid (GABA) concentration (in mmol/L) within the anterior cingulate cortex (ACC) were generated for three groups: low-risk healthy controls (LR), individuals at high risk of alcohol use disorder (HR), and recently detoxified alcohol-dependent patients (AD). *a significant group difference was observed (F(2,85) = 4.14, p = 0.019), as well as a significant quadratic trend (F(2,85) = 4.99, p = 0.028), indicating a GABA concentration distribution of LR $>$ HR AD.

Abstinence duration in AD (mean: 36.2 days, minimum: 8 days, maximum: 95 days, standard deviation: 20.1) was not correlated with GABA levels in the ACC Pearson correlation: r = –0.165, p = 0.431. Furthermore, GABA levels in the ACC were not correlated with craving symptoms (OCDS), symptom severity of alcoholism (ADS), depressive or anxiety symptoms (HADS) across the entire sample or within either the HR or AD groups.