Anterior cingulate metabolite levels, memory, and inhibitory control in abstinent men and women with alcohol use disorder

Abstract

Alcohol use disorder (AUD) has been shown to have harmful cognitive and physiological effects, including altered brain chemistry. Further, although men and women may differ in vulnerability to the neurobiological effects of AUD, the results of existing studies have been conflicting. We examined brain metabolite levels and cognitive functions in a cross-section of men with AUD (AUDm) and women with AUD (AUDw) to determine the degree of abnormalities after extended periods of abstinence (mean, 6 years) and to evaluate gender differences in neuropsychological and metabolite measures. Participants were 40 abstinent individuals with AUD (22 AUDw, 18 AUDm) and 50 age-equivalent non-AUD comparison participants (26 NCw, 24 NCm). Proton magnetic resonance spectroscopy (MRS) was employed at 3 Tesla to acquire metabolite spectra from the dorsal anterior cingulate cortex (dACC). Brain metabolites N-acetyl aspartate (NAA), choline (Cho), myo-Inositol (mI), and glutamate & glutamine (Glx) were examined relative to measures of memory and inhibitory control. Metabolite levels did not differ significantly between AUD and NC groups. Memory and inhibitory-control impairments were observed in the AUD group. There also were significant group-specific associations between metabolite ratios and measures of inhibitory control. There were no group-by-gender interactions for the four metabolite ratios. These findings demonstrate that brain metabolite levels in men and women with AUD, following long-term abstinence, do not differ from individuals without AUD. The data also provide preliminary evidence of sustained associations between metabolite levels and measures of inhibitory control, a functional domain important for curtailing harmful drinking.

Short Summary: Measures of cortical metabolites are similar in long-term abstinent men and women with and without alcohol use disorder (AUD). Associations between metabolite levels and inhibitory control measures in AUD suggested continued functional relevance of these metabolites in AUD even after sustained abstinence. Gender differences in metabolite levels were not observed.

Introduction

According to the 2019 National Survey on Drug Use and Health (NSDUH, 2020), 14.5 million people in the USA met the criteria for alcohol use disorder (AUD) in the prior year. AUD is characterized by an impaired ability to control drinking despite adverse personal, societal, or occupational consequences and has been shown to have long-lasting harmful physiological and neuropsychological effects, including associations with brain abnormalities. Cognitive abilities such as general intelligence and over-learned knowledge are preserved with AUD (Bates et al. 2013; Stavro et al. 2013), while impairments in other functions continue despite abstinence (Chanraud et al. 2007; Oscar-Berman et al. 2014; Jia et al. 2021). Among the most common and persistent cognitive domains of impairment are memory and inhibitory control (Stephan et al. 2017; Mullins-Sweatt et al. 2019). Structural and functional magnetic resonance imaging (MRI) studies have sought to offer biological explanations linking heavy alcohol consumption and AUD with brain abnormalities (Sullivan and Pfefferbaum 2005; Oscar-Berman and Marinkovic 2007; Buhler and Mann 2011; Zahr and Pfefferbaum 2017). Likewise, AUD-related alterations in brain chemistry, measured using proton (1H) magnetic resonance spectroscopy (MRS), have been used to detect and quantify the stability of metabolites that have important physiological functions for maintaining brain health and function (Behar et al. 1999; Mason et al. 2006; Mon et al. 2012; Meyerhoff et al. 2013).

The metabolite alterations widely reported in current alcohol drinkers and recently detoxified chronic heavy drinkers (Meyerhoff et al. 2013; Fritz et al., 2019) have focused on prominent, readily detectable, and reliably quantifiable brain metabolites: N-acetyl aspartate (NAA), a marker of neuronal integrity; choline (Cho), white matter (WM) integrity, membrane turnover, and inflammation; creatine (Cr), energy metabolism; myo-Inositol (mI), phospholipid metabolism and osmotic equilibrium; glutamine and glutamate (combined as Glx), neuronal activation and glucose metabolism (Govindaraju et al. 2000; Behar and Rothman 2001).

Despite promising evidence of neuropsychological improvements (Oscar-Berman et al. 2014) and the associated recovery of metabolite levels after short-term abstinence (Bendszus et al. 2001; Durazzo et al. 2013), it remains unclear whether these associations are maintained after longer lengths of abstinence. Therefore, a primary objective of this study was to employ MRS to examine NAA, mI, Cho, and Glx metabolites relative to measures of inhibitory control and memory in individuals with AUD who reported longer lengths of abstinence (mean, 6 years) than had previously been examined. It also is critical to consider possible differences in metabolite levels between men with AUD (AUDm) and women with AUD (AUDw). Although mounting evidence demonstrates an important role of gender in differentiating the impact of AUD on the brain (Nixon et al. 2014; Sawyer et al. 2018; Oscar-Berman et al. 2021), investigations of brain metabolite levels have largely been limited to men, or have included insufficient numbers of women to detect group differences (Verplaetse et al. 2021). Thus, gender differences were directly examined by comparing AUDm and AUDw on metabolite levels and on their associations with the length of abstinence and with neuropsychological measures of memory and inhibitory control.

Materials and Methods

Participant recruitment and inclusion

The study included 40 abstinent individuals with a history of AUD (22 AUDw, 18 AUDm) and 50 non-AUD age-matched comparison subjects (26 NCw, 24 NCm). Participants were recruited from the Boston area via newspaper and web-based advertisements and flyers. The study was approved by the participating Institutional Review Boards: Boston University School of Medicine (#H24686), VA Boston Healthcare System (#1017 and #1018), and Massachusetts General Hospital (#2000P001891). Participants provided written informed consent and were reimbursed $15 per hour for assessments, $25 per hour for scans, and $5 for travel expenses.

Participants were interviewed to ensure they met the inclusion criteria. Participants were given the Computerized Diagnostic Interview Schedule (Robins et al. 2000), which provides psychiatric diagnoses according to criteria established by the American Psychiatric Association (DSM-IV) (APA 1994). Participants were excluded if English was not one of their first languages, if they were predominantly left-handed, or if they had any of the following: Korsakoff’s syndrome; HIV; cirrhosis; major head injury with loss of consciousness greater than 15 min unrelated to AUD; stroke; epilepsy or seizures unrelated to AUD; history of electroconvulsive or electro-shock therapy; major neurotoxicant exposure; psychotic disorders; bipolar II; Hamilton Rating Scale for Depression (HRSD) (Hamilton 1960) score over 18; or history of drug use once per week or more within the past 5 years, except for two cases: one AUD woman who had used marijuana regularly (but not within the past 3 months), and one NC man who had used marijuana regularly (but not within the past 2 years). AUD participants were excluded if they did not have positive criteria for alcohol abuse or dependence, if their duration of heavy drinking (DHD) was less than 5 years, or if they had consumed alcohol within four weeks prior to testing. Non-AUD control participants were excluded if they reported a DHD greater than 1 year. Heavy drinking refers to 21 or more drinks per week for at least 5 years (one drink equals 355 ml beer, 148 ml wine, or 44 ml hard liquor).

Participant assessment procedures

Neuropsychological measures of the domains of memory and inhibitory control were examined. Memory was assessed using the Working Memory subtest of the Wechsler Adult Intelligence Scale—Fourth Edition (WAIS-IV) and the Immediate Memory and Delayed Memory subtests of the Wechsler Memory Scale—Fourth Edition (WMS-IV) (Holdnack and Drozdick 2010). Inhibitory control was assessed using Dickman’s Impulsivity Inventory (Dickman 1990) and the Barratt Impulsiveness Scale (BIS-11) (Patton et al. 1995). Three Delis–Kaplan Executive Function System (D-KEFS) measures (Delis et al. 2001) were acquired: Trails Number Sequencing-2 (analogous to Trails A), Trails Number-Letter Switching-4 (analogous to Trails B), and Verbal Letter Fluency-1 (analogous to Controlled Oral Word Association Test, COWAT) (Lezak et al. 2004). For the D-KEFS, scaled scores were used, and higher scores indicate better performance.

Participants were administered the Alcohol Use Questionnaire (Cahalan et al. 1969), which yields measures reflecting length of sobriety (LOS; in years), DHD (in years), and a quantity frequency index [ounces of alcohol per day, roughly equivalent to daily drinks (DD)]. The LOS refers to the period between the MRI scan date and the last reported drink. The DHD represents the total number of years participants drank at least 21 drinks per week (an average of three drinks per day). The DD measure reflects the last 6 months (NC group) or during the 6 months preceding cessation of drinking (AUD group). Five AUD participants drank fewer than three drinks/day during the 6 months prior to cessation; thus, DD was obtained for the last 6 months of heavy drinking. All participants also completed the Alcohol Use Disorders Identification Test (AUDIT) (Babor et al. 1992). The questionnaire was modified to be past tense for the AUD group to assess the time during which they were drinking heavily.

MRI acquisition

MRI scans were obtained at Massachusetts General Hospital on a 3 Tesla Siemens MAGNETOM Trio Tim scanner with a 32-channel head coil (123.18 MHz). Once positioned, head placement was confirmed using three-plane scout images. Two T1 weighted multi-echo MP-RAGE series for volumetric analysis (one AUD man had only one series) were obtained with these parameters: TR = 2530 ms, TE = 1.79, 3.71, 5.63, 7.55 ms (RMS average used), flip angle = 7°, field of view = 256 mm, matrix = 256x256, slice thickness = 1 mm, 176 interleaved sagittal slices, with GRAPPA acceleration factor of 2.

Magnetic resonance spectroscopy

A 20 mm x 20 mm x 20 mm oblique voxel was prescribed in the anterior cingulate cortex (ACC) along the midline, with the inferior plane of the voxel parallel to the descending surface of the corpus callosum (Fig. 1). Voxel shimming, flip-angle, water-suppression, and frequency were automatically adjusted using Siemens software. Proton spectroscopy data were acquired using a Point-Resolved Echo Spectroscopy Sequence (PRESS) to acquire water-suppressed TE = 30 ms ACC spectra. Additional acquisition parameters included TR = 2 s, spectral bandwidth = 1.2 kHz, readout duration = 512 ms, NEX = 128, total scan duration = 4.3 min. Spectroscopic data processing was performed using in-house reconstruction code and LCModel (Provencher 1993). The 30 ms LCModel basis set utilized a GAMMA-simulated model based on the PRESS sequence. Five metabolite levels were assessed: Cho-containing compounds, Glx, mI, NAA, and total Cr (tCr, Cr + phosphocreatine). Calculating metabolite ratios relative to tCr as an internal standard, a method that has inherent limitations, is accepted in the field as a validated and reliable method for examining metabolites (Jansen et al. 2006). Furthermore, examination of tCr levels relative to the total proton signal did not yield between-group differences, thus, ratios of metabolites to tCr were calculated: Cho/tCr, Glx/tCr, mI/tCr, and NAA/tCr.

Figure 1.

MRS voxel placement and sample proton spectrum Sagittal image illustrating the placement of 20 mm x 20 mm x 20 mm single voxel in the ACC and a sample spectrum (black) and LCModel fit (red) from a study participant. Abbreviations: ACC, anterior cingulate cortex; tCr, creatine; Cho, choline; Glu, glutamate; Glx, glutamate, and glutamine; NAA, N-acetyl-aspartate; mI, myo-Inositol; ppm, parts per million.

Image segmentation analysis

The high-resolution image sets (T1-weighted) were segmented into gray matter (GM), WM, and cerebrospinal fluid (CSF) binary-tissue maps (FSL, Oxford, UK). Partial tissue percentages were extracted for the oblique ACC voxel for use in analyses.

Statistical analyses

Statistical analyses were performed using R version 3.4.0 (R Core Team 2020). To examine group-by-gender interactions, we constructed separate general linear models using the lm function in R, predicting each measure of interest: the participant characteristics (Table 1a), the tissue contributions, and four metabolite ratios (Table 2). We visually confirmed that regression assumptions were satisfied (normality of outcome measures, homogeneity of variance), and we set thresholds for multicollinearity (Pearson correlations <0.5) and influence (Cook’s D < 1.0). For each model, we report findings from the analysis of covariance (ANCOVA) (using the car: Anova function with Type III sums of squares), followed by the results from the post hoc analyses (using the emmeans package) (Lenth 2016). We examined the interaction of group-by-gender to assess how the impact of AUD differed for men and women.

Table 1a.

Participant characteristics

	AUDm n = 18		AUDw n = 22		NCm n = 24		NCw n = 26
	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Age (years)	48.9	(9.3)	51.3	(11.4)	47.6	(10.7)	50.5	(15.7)
Education (years)	14.3	(1.7)	14.3	(1.5)	15.4	(2.6)	15.2	(2.3)
Depression score (HRSD)	2.8	(3.7)	3.8	(3.2)	1.3	(2.4)	2.0	(2.2)
DD (ounces ethanol/day)	11.8	(7.1)	9.8	(8.2)	0.2	(0.2)	0.3	(0.3)
DHD (years)	14.0	(7.6)	13.6	(5.9)	0.0	(0.0)	0.0	(0.0)
LOS (years)	4.4	(6.0)	7.3	(9.5)	N/A	N/A	N/A	N/A
WAIS
Full Scale IQ	107.2	(16.6)	100.9	(15.7)	111.5	(15.3)	108.6	(13.8)
Working Memory	108.8	(16.1)	102.0	(14.5)	114.5	(14.3)	104.4	(14.3)
WMS
Immediate Memory	102.8	(17.6)	105.5	(19.3)	109.4	(16.2)	112.5	(15.7)
Delayed Memory	105.6	(16.4)	108.3	(18.9)	113.8	(17.6)	114.2	(11.8)
D-KEFS
Trails Number Sequencing	11.0	(1.6)	10.3	(2.8)	10.3	(3.5)	12.0	(2.2)
Trails Number-Letter Switching	10.6	(2.4)	10.4	(3.3)	10.9	(2.6)	11.0	(3.0)
Verbal Letter Fluency	11.6	(4.2)	10.8	(3.0)	13.1	(3.4)	14.0	(3.4)
Dickman
Functional Impulsivity	5.9	(2.3)	4.7	(1.8)	4.6	(1.9)	5.1	(1.4)
Dysfunctional Impulsivity	4.9	(2.2)	5.1	(2.1)	3.8	(1.7)	4.0	(0.9)
BIS-11	31.5	(7.9)	31.4	(7.6)	26.2	(5.5)	27.7	(6.8)

Abbreviations: HRSD = Hamilton Rating Scale for Depression; DD = Daily Drinks; DHD = Duration of Heavy Drinking; LOS = Length of Sobriety; WAIS = Wechsler Adult Intelligence Scale; WMS = Wechsler Memory Scale; D-KEFS = Delis-Kaplan Executive Function System; BIS-11 = Barratt Impulsivity Scale

Table 2.

Metabolite ratios (normalized to creatine levels), and tissue contributions

	AUDm n = 18		AUDw n = 22		NCm n = 24		NCw n = 26
	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Cho/tCr	0.25	0.03	0.23	0.03	0.25	0.03	0.24	0.03
Glx/tCr	0.98	0.14	1.05	0.12	1.02	0.15	1.04	0.14
mI/tCr	0.80	0.12	0.76	0.11	0.78	0.10	0.77	0.10
NAA/tCr	1.07	0.17	1.08	0.14	1.09	0.18	1.09	0.16
GM	57.77	6.29	59.27	6.84	59.19	7.55	59.49	8.49
WM	26.65	4.92	26.24	5.61	26.56	5.87	27.34	7.11
CSF	15.81	4.62	15.24	5.59	13.01	5.16	12.60	4.18
	NC –AUD 95% CI	Women – Men 95% CI	NCm –AUDm 95% CI	NCw –AUDw 95% CI	AUDw - AUDm 95% CI	NCw –NCm 95% CI
Cho/tCr	[−0.01, 0.01]	[−0.02, 0.00]	[−0.02, 0.02]	[−0.02, 0.02]	[−0.03, 0.00]	[−0.03, 0.00]
Glx/tCr	[−0.05, 0.06]	[−0.02, 0.10]	[−0.06, 0.13]	[−0.09, 0.06]	[−0.02, 0.15]	[−0.06, 0.10]
mI/tCr	[−0.05, 0.04]	[−0.07, 0.02]	[−0.09, 0.05]	[−0.05, 0.07]	[−0.11, 0.03]	[−0.07, 0.05]
NAA/tCr	[−0.05, 0.08]	[−0.06, 0.07]	[−0.10, 0.12]	[−0.07, 0.10]	[−0.10, 0.10]	[−0.09, 0.10]
GM	[−2.29, 3.80]	[−2.27, 3.88]	[−2.91, 5.74]	[−4.23, 4.68]	[−2.72, 5.70]	[−4.26, 4.86]
WM	[−1.92, 3.00]	[−2.25, 2.72]	[−3.46, 3.28]	[−2.59, 4.80]	[−3.79, 2.96]	[−2.92, 4.48]
CSF	[−4.77, −0.63]	[−2.52, 1.72]	[−5.86, 0.27]	[−5.57, 0.28]	[−3.83, 2.70]	[−3.11, 2.28]

The minus signs in the header row indicate subtraction. The upper set of numbers provides means and standard deviations, and the lower numbers represent 95% confidence intervals (CI). Abbreviations: NAA = N-acetyl aspartate; Cho = choline-containing compounds; mI = myo-Inositol; Glx = glutamate + glutamine; tCr = total creatine; GM = gray matter; WM = white matter; CSF = cerebrospinal fluid. Bolded numbers indicate significance at P < 0.05.

Five categories of regression analyses were conducted, as detailed below. For group and gender differences, (1) participant characteristics, (2) brain tissue, GM, WM, and CSF, and (3) metabolite ratio levels were examined. We also measured relationships between metabolite levels, (4) drinking history, and (5) neuropsychological measures.

(1) Group and gender differences for participant characteristics were assessed with a model that included the interaction of group and gender (for example, specified as ‘age ~ group*gender’).

(2) Group and gender differences for tissue contributions were assessed with a model that included the interaction of group and gender.

(3) Group and gender differences for metabolite ratios were assessed with a model that included the interaction of group and gender, with GM and Education as covariates (for example, specified as ‘choline_ratio ~ group*gender + GM + education’). Age was highly correlated with GM (r = 0.5), so it was not included as a covariate. Bayes factors were calculated for these same linear models against intercept-only models and additionally calculated for linear models omitting gender for a total of eight Bayes factors.

(4) Relationships of metabolite levels to the three measures of drinking history (DHD, DD, and LOS) were assessed in the AUD group. For each of the four metabolites, a model was constructed to assess the effects of drinking history measures and interactions with gender (for example, specified as ‘choline_ratio ~ DHD*gender + DD*gender + LOS*gender + GM + education’).

(5) Relationships between metabolites and neuropsychological measures were assessed. For each of the four metabolites, separate models were constructed for each of the ten measures: WAIS-IV Full Scale IQ, WAIS-IV Working Memory, WMS-IV Immediate Memory, WMS-IV Delayed Memory, D-KEFS Trails Number Sequencing-2, D-KEFS Trails Number-Letter Switching-4, D-KEFS Verbal Letter Fluency-1, Dickman Functional Impulsivity, Dickman Dysfunctional Impulsivity, and BIS-11. Each model was specified to examine interactions of the neuropsychological measure with group and gender while accounting for GM tissue contribution (for example, ‘choline_ratio ~ bis*alcoholism*gender + GM’) for a total of 40 models. Age and education were not included as covariates because normative scaled scores were used. For all analyses, significant predictors and interactions indicated by ANCOVAs were followed by post hoc comparisons using the emmeans::emtrends function. This function allows slopes within the model to be compared (for example, how relationships of NAA/tCr with BIS differ for the AUD vs. NC groups).

In addition to regressions, correlational analyses were conducted to examine how drinking history related to memory and inhibitory control measures in the AUD group, using the ppcor::pcor. Test function to calculate partial correlation values for relationships between metabolite ratios and neuropsychological measures for each group separately, accounting for GM tissue contribution.

Results

Participant characteristics

Participant characteristics are summarized in Table 1a. The AUD group had a mean LOS of 6.0 years (range 0.1 to 32.8), a mean DD of 10.7 drinks (range 2.3 to 34.8), and a mean DHD of 13.8 drinks (5.0 to 32.0). Analyses of group (AUD vs. NC) and gender revealed no significant main effects or interactions with age (NC mean = 49.1 years, range = 26.6 to 77.0 years; AUD mean = 50.2 years, range = 28.2 to 73.3 years). There was a significant main effect of group for education (F_1,86 = 4.40, P = 0.04), with the education of the NC group (mean = 15.3 years) being higher than the AUD group (mean = 14.3 years), but there was no significant main effect of gender or group-by-gender interaction. HRSD scores were significantly higher, 0.4 to 3.0 points (95% CI) in the AUD group (mean = 3.3) compared with the NC group (mean = 1.6) (F_1,86 = 7.56, P = 0.007). Analyses of gender effects for drinking history variables in the AUD group revealed no significant main effects of gender.

Analyses of memory measures demonstrated that women had lower scores on the WAIS Working Memory Scale than men (F_1,86 = 7.36, P = 0.008) and that the AUD group had lower scores on the WMS Delayed Memory than the NC group (F_1,86 = 4.21, P = 0.04) (Table 1b). Analyses of inhibitory control measures showed a group-by-gender interaction for D-KEFS Number Sequencing-2 (F_1,86 = 4.75, P = 0.03), with NCw exhibiting higher scores than AUDw, while NCm scored significantly lower than AUDm. The AUD group had lower D-KEFS Verbal Letter Fluency-1 scores than the NC group (F_1,86 = 10.06, P = 0.002). There was a significant group-by-gender interaction for Dickman Functional Impulsivity (F_1,86 = 4.56, P = 0.04); the AUDm had a higher score than NCm, but AUDw had significantly lower scores than NCw. The AUD group also demonstrated significantly higher scores than the NC group on the Dickman Dysfunctional Impulsivity Scale (F_1,86 = 8.59, P = 0.004) and on the BIS-11 (F_1,86 = 9.33, P = 0.003).

Table 1b.

Participant characteristics

	NC - AUD		Women - Men		NCm - AUDm		NCw - AUDw		AUDw - AUDm		NCw - NCm
Age (years)	[−6.1, 3.9]		[−2.3, 7.7]		[−7.5, 5.0]		[−8.7, 7.1]		[−4.2, 9.0]		[−4.7, 10.5]
Education (years)	[0.1, 1.8]		[−1.1, 0.7]		[−0.3, 2.4]		[−0.3, 2.0]		[−1.1, 1.0]		[−1.6, 1.2]
Depression score (HRSD)	[−3.0, −0.4]		[−0.4, 2.1]		[−3.6, 0.5]		[−3.5, −0.2]		[−1.3, 3.3]		[−0.6, 2.0]
DD (oz ethanol/day)	[−13.0, −8.0]		[−3.6, 2.6]		[−15.2, −8.1]		[−13.2, −5.9]		[−6.9, 2.9]		[−0.1, 0.2]
DHD (years)	[−15.9, −11.7]		[−3.2, 3.7]		[−17.8, −10.2]		[−16.2, −11.0]		[−4.9, 4.1]		N/A
LOS (years)	N/A		[−0.7, 5.4]		N/A		N/A		[−2.1, 7.9]		N/A
WAIS
Full Scale IQ	[−0.3, 12.8]		[−11.1, 1.9]		[−5.8, 14.5]		[−1.0, 16.4]		[−16.7, 4.1]		[−11.2, 5.4]
Working Memory	[−2.2, 10.6]		[−15.0, −2.6]		[−4.1, 15.4]		[−6.0, 10.8]		[−16.8, 3.1]		[−18.3, −2.0]
WMS
Immediate Memory	[−0.6, 14.0]		[−4.5, 9.9]		[−4.2, 17.4]		[−3.5, 17.3]		[−9.1, 14.5]		[−6.1, 12.1]
Delayed Memory	[0.0, 13.9]		[−5.7, 8.2]		[−2.5, 18.9]		[−3.5, 15.4]		[−8.6, 14.0]		[−8.3, 9.1]
D-KEFS
Trails Number Sequencing	[−0.6, 1.7]		[−0.5, 1.8]		[−2.4, 0.9]		[0.2, 3.2]		[−2.1, 0.7]		[0.1, 3.5]
Trails Number-Letter Switching	[−0.7, 1.7]		[−1.2, 1.1]		[−1.2, 1.9]		[−1.2, 2.4]		[−1.9, 1.7]		[−1.5, 1.7]
Verbal Letter Fluency	[0.9, 3.9]		[−1.5, 1.6]		[−1.0, 4.0]		[1.3, 5.1]		[−3.3, 1.6]		[−1.1, 2.8]
Dickman
Functional Impulsivity	[−1.2, 0.4]		[−1.0, 0.6]		[−2.6, 0.1]		[−0.5, 1.3]		[−2.5, 0.2]		[−0.5, 1.4]
Dysfunctional Impulsivity	[−1.9, −0.3]		[−0.6, 1.0]		[−2.4, 0.2]		[−2.1, −0.1]		[−1.3, 1.5]		[−0.6, 1.0]
BIS-11	[−7.4, −1.5]		[−2.1, 3.9]		[−9.7, −0.9]		[−8.0, 0.5]		[−5.1, 4.9]		[−2.0, 5.0]

The minus signs in the header row indicate subtraction. Abbreviations: HRSD = Hamilton Rating Scale for Depression; DD = Daily Drinks (ounces ethanol/day); DHD = Duration of Heavy Drinking (years); LOS = Length of Sobriety (years); WAIS = Wechsler Adult Intelligence Scale; WMS = Wechsler Memory Scale; D-KEFS = Delis-Kaplan Executive Function System; BIS-11 = Barratt Impulsivity Scale

The upper set of numbers provides means and standard deviations, and the lower numbers represent confidence intervals of the mean differences (CI). Abbreviations: HRSD = Hamilton Rating Scale for Depression; DD = Daily Drinks (ounces ethanol/day); DHD = Duration of Heavy Drinking (years); LOS = Length of Sobriety (years); WAIS = Wechsler Adult Intelligence Scale; WMS = Wechsler Memory Scale; D-KEFS = Delis-Kaplan Executive Function System; BIS-11 = Barratt Impulsivity Scale. Bolded numbers indicate significance at P < 0.05.

Partial correlational analyses of memory and inhibitory control measures relative to drinking history in the AUD group demonstrated a significant positive correlation between DHD and Dickman Dysfunctional Impulsivity scores (r₄₀ = 0.40, P = 0.01). Additionally, LOS in the AUD group was positively related to age (r₄₀ = 0.50, P = 0.001) and Education (r₄₀ = 0.33, P = 0.04) and negatively related to DD (r₄₀ = −0.31, P = 0.05). LOS also was positively associated with WAIS Working Memory (r₄₀ = 0.33, P = 0.04) and Immediate Memory (r₄₀ = 0.39, P = 0.01).

Metabolite ratios and tissue contributions

Analyses of group and gender effects for tissue contributions revealed no significant main effects or interactions for GM or WM. However, CSF levels were higher in the AUD group relative to the NC group (Table 2).

The ratios of tCr/total proton signal did not differ significantly between groups, AUDm = 0.20 ± 0.03, AUDw = 0.20  ± 0.01; NCm = 0.20 ± 0.01; NCw = 0.20 ± 0.02 (Table 2).

Metabolite ratios for the AUD and NC groups also were not statistically different (Fig. 2). The main effect of gender for the Cho/tCr ratio was significant (F_1,84 = 5.04, P = 0.03), with higher Cho/tCr in men than in women. No other significant main effects of group or gender or group-by-gender interactions were observed (Table 2). Bayes factors were calculated against an intercept-only model. Group-by-gender interaction effects yielded Bayes factors of 0.020 ± 3.36%, 0.043 ± 3.31%, 0.011 ± 2.99%, and 0.002 ± 2.58%, for Cho/tCr, Glx/tCr, mI/tCr, and NAA/tCr, respectively. Group effect models yielded Bayes factors of 0.032 ± 1.11%, 0.223 ± 1.48%, 0.093 ± 1.13%, and 0.029 ± 4.35%.

Figure 2.

Metabolite levels Abbreviations: AUD = Alcohol Use Disorder; NC = Non-AUD Control; NAA = N-acetyl aspartate; Cho = choline-containing compounds; mI = myo-Inositol; Glx = glutamate; tCr = total creatine.

Associations between metabolite ratios and drinking history

Regression models constructed to identify associations between the four metabolite levels and the drinking history variables revealed no significant relationships or interactions with gender. Scatterplots for each metabolite level relative to the LOS are provided in the supplementary material.

Associations between metabolite ratios and neuropsychological measures

Regression models were constructed to identify how relationships between metabolite ratios and neuropsychological scores differed by group. We considered these relationships to be exploratory but potentially important. Therefore, in order to reduce the possibility of missing valid relationships (and the probability of Type II errors), we did not correct for multiple comparisons. There were four associations with slopes that were significant within a group and also were significantly different from the corresponding (non-significant) relationship in the other group (Table 3). The AUD group had stronger relationships (steeper slopes) than the NC group for three of the four measures, as follows. (1) For Dickman Functional Impulsivity, the AUD group had a significantly more positive slope in relation to Cho/tCr (t₈₁ = 2.2, P = 0.03). (2) For Dickman Dysfunctional Impulsivity, the AUD group had a significantly more negative slope in relation to mI/tCr (t₈₁ = −2.4, P = 0.02). (3) For the BIS-11, the AUD group had a significantly more positive slope in relation to NAA/tCr than the NC group (t₈₁ = 2.9, P = 0.005). For the fourth measure, the NC group demonstrated stronger associations than the AUD group between D-KEFS Trails Number Sequencing scores and Glx/tCr levels (t₈₁ = −2.0, P = 0.05). That is, a positive relationship was identified for the NC group but not for the AUD group, wherein higher scores were associated with higher levels of Glx/tCr. Additionally, compared with women, men had a stronger association between scores on the Dickman Functional Impulsivity and mI/tCr ratios (t₈₁ = 2.3, P = 0.03), wherein men showed higher Functional Impulsivity scores in conjunction with higher mI/tCr. All five of these results were considered preliminary and, therefore, provisional.

Table 3.

Metabolite ratios: interactions of group with neuropsychological measures

Metabolite ratio	Measures of inhibitory control	Group interaction F(1, 81)	AUD 95% CI	r	NC 95% CI	r
Glx/tCr	D-KEFS Trails Number Sequencing-2	6.9	[−0.03, 0.01]	−0.3	[0.00, 0.03]	0.3
Cho/tCr	Dickman Functional Impulsivity	4.2	[0.00, 0.01]	0.4	[−0.01, 0.00]	−0.1
mI/tCr	Dickman Dysfunctional Impulsivity	6.5	[−0.03, 0.00]	−0.3	[−0.01, 0.04]	0.2
NAA/tCr	BIS-11	7.0	[0.00, 0.01]	0.4	[−0.01, 0.00]	−0.2
Metabolite Ratio	Measures of Inhibitory Control	Gender Interaction F(1,81)	Men 95% CI	r	Women 95% CI	r
mI/tCr	Dickman Functional Impulsivity	5.1	[0.01, 0.04]	0.4	[−0.03, 0.01]	−0.1

The r values represent partial correlations that account for gray matter tissue contribution, using the ppcor::pcor.test function. Abbreviations: NAA = N-acetyl aspartate; Cho = choline-containing compounds; mI = myo-Inositol; Glx = glutamate + glutamine; tCr = total creatine; GM = gray matter; WM = white matter; CSF = cerebrospinal fluid. Bolded numbers indicate significance at P < 0.05. No results were significant at the 40 comparisons, Bonferroni-adjusted threshold of P < 0.00125 (Fig. 3 illustrates these relationships).

Discussion

The main results of this project are 3-fold. First, by demonstrating that proton metabolites measured in the ACC of long-term abstinent men and women with AUD are similar to those of individuals without AUD, our results extend findings from prior studies that demonstrated recovery of metabolite levels after shorter-term periods of detoxification and abstinence. Second, we describe significant relationships between drinking history and neuropsychological measures. Third, associations between metabolite levels and measures of inhibitory control in individuals with AUD were observed, suggesting the possibility of continued functional relevance of these metabolites in AUD even after sustained abstinence. Despite the importance of our findings, it should be noted that, as with other MRS research, our results do not provide a cohesive story that ties them together and that the explanations rely on select features of each metabolite. We anticipate that subsequent research may allow for more theory-driven explorations.

Metabolite levels

The equivalence in metabolite levels observed between AUD and NC groups (within 10% and with small Bayes factor values) is an encouraging sign for current or recently detoxified heavy drinkers (Table 2 and Fig. 2). It suggests that alterations in metabolite levels, which are thought to be markers of acute brain injury (Meyerhoff et al. 2013), are reversible and may recover completely and permanently, even when some volumetric or functional brain differences persist. This finding is consistent with previous MRS studies documenting the normalization of brain metabolites after shorter periods of abstinence. This AUD cohort had been abstinent for a long duration (mean 6 years); thus, these results contribute to the characterization of the timeline of metabolite recovery by suggesting that improvements in metabolite levels seen in recently detoxified heavy drinkers persist with long-term sobriety.

Indeed, the AUD group had impairments in measures of memory and in inhibitory control, despite there being no substantial differences from the NC group in metabolite levels. This finding supports the idea that the memory and cognitive control deficits seen in long-term abstinent AUD individuals are more likely to stem from persistent structural brain abnormalities (O'neill et al. 2001; Monnig et al. 2013) or from pre-existing personality factors (Ruiz et al. 2013; Mullins-Sweatt et al. 2019) than they are to stem from the acute disruption of brain chemistry associated with active alcohol consumption or withdrawal. For example, poor memory (Finn and Hall 2004) and impulsivity (Stephan et al. 2017) can be risks for heavy drinking.

There have been few prior studies that have included sufficient sample sizes of AUDw to permit the examination of gender differences in metabolites measured by proton spectroscopy. The identification of gender differences is a crucial component of precision medicine and will contribute to the understanding of how addiction differs among people depending on individual factors like age and gender. Armed with that knowledge, clinicians can better tailor treatment and prevention strategies. In this study, the proportion of women (48 women out of 90 participants) was substantially larger than in prior MRS studies examining metabolite levels in the context of AUD. While the results indicated similar levels of NAA/tCr, mI/tCr, and Gx/tCr in men and women, Cho/tCr was significantly higher in men (by ~6% or 0.4 SD). Cho/tCr is known to be associated with WM integrity, but we found no relationship with LOS or with our behavioral measures, suggesting this gender difference may reflect differences in brain composition.

Two significant group-by-gender interactions were identified: one for Dickman Functional Impulsivity and one for Trails Number Sequencing. For Dickman Functional Impulsivity, no significant simple main effects were observed. The NCw had higher scores than AUDw on Trails Number Sequencing, whereas the patterns for men were not significant. While the direction of effect (NCw > NCm) may be counterintuitive, these findings could suggest a gender-specific difference in processing speed in women with a history of AUD, which is consistent with previous reports (Fama et al. 2019).

Relationships between drinking history and neuropsychological measures

There was a significant positive correlation between DHD and Dickman Dysfunctional Impulsivity scores, indicating that people with longer heavy drinking histories have more dysfunctional behavior patterns. Additionally, LOS was positively related to age and education, and negatively related to DD, indicating that individuals in our sample who had maintained their sobriety longest were older (perhaps a direct function of the years required to have a longer LOS), better educated, and drank less when they were actively drinking. LOS also was positively associated with WAIS Working Memory and with WMS Immediate Memory, suggesting recovery of neuropsychological functions with abstinence or perhaps a protective effect of memory skills in maintaining sobriety. These data provide preliminary evidence of associations between metabolite ratios and scores of the three impulsivity measures that are significantly more pronounced in the AUD group than in the NC group and one measure that was more pronounced in the NC group (Fig. 3).

Figure 3.

Relationships between metabolite ratios and neuropsychological measures Abbreviations: AUD = Alcohol Use Disorder; NC = Non-AUD Control; NAA = N-acetyl-aspartate; Cho = choline-containing compounds; mI = myo-inositol; Glx = glutamate + glutamine; tCr = total creatine.

Relationships between metabolite levels and neuropsychological measures

Of the 40 models that evaluated relationships between the four metabolites and 10 neuropsychological measures of IQ, memory, and inhibitory control, four significant interactions with the AUD group were identified, along with one significant gender interaction (Table 3 and Fig. 3). Because corrections for multiple comparisons were not applied, we consider these results to be provisional. (For additional discussion, see Limitations.) These patterns could reflect changing associations with long-term abstinence and could benefit from further investigation, especially if they can provide insights regarding resistance to or risk for relapse. For three of the four significant interactions of metabolites with group, the relationships were significant for the AUD group but not for the NC group, and the interaction indicated the relationships were significantly stronger for the AUD group than the NC group, although it was not clear if the relationships were different in sign, or only in magnitude. For the fourth, the relationship was significant for the NC group but not the AUD group.

NAA levels were related to higher impulsivity on the BIS-11 in the AUD group, suggesting a relationship between ACC neuronal integrity and the well-documented impulsive behavior of people with a history of heavy drinking (Mitchell et al. 2005). The AUD group’s scores on the BIS-11 could indicate that a higher level of impulsivity supports better ACC function, as measured by NAA, a marker of neuronal integrity. That the relationship between NAA levels and BIS-11 scores is smaller for the NC group further suggests that this relationship either is related to the pathological development of AUD or is created as a consequence of heavy drinking.

A similar interaction was observed for Dickman Functional Impulsivity: higher levels in the AUD group were associated with higher Cho/tCr ratios, suggesting a relationship between spontaneous behavior patterns and ACC membrane turnover in this population. As with NAA and scores on the BIS-11, the more steeply positive relationship for the AUD group indicates a stronger connection of a cellular measure with greater impulsivity than the connection observed for the NC group. This suggests impulsivity, a common characteristic of AUD (Crews and Boettiger 2009), could be connected to membrane turnover in the ACC for the AUD group but not the NC group, although a causal mechanism remains opaque.

Dickman Dysfunctional Impulsivity was associated with lower mI/tCr in the AUD group, which might suggest less ACC membrane turnover or inflammation despite more behavioral impairment in this population. The association of ACC mI/tCr with dysfunctional impulsivity is not surprising because the ACC is involved in functions known to be impaired in AUD, such as decision-making and deployment of cognitive control (Holroyd and Yeung 2012) in humans, and variability in mI in rodents has been connected to impulsivity (Jupp et al. 2020). Additionally, mI has a complex basis in pharmacological markers by controlling cellular processes and generating calcium signals required for the formation of memory in neurons (Pattij and Vanderschuren 2020).

For Trails Number Sequencing, the relationship to Glx/tCr was stronger in the NC group than in the AUD group. Trails Number Sequencing is sensitive to measures of visual scanning, graphomotor speed, and visuomotor processing speed (Llinas-Regla et al. 2017), skills required for normal inhibitory control functions. Since Glx/tCr ratios have been associated with tissue health, it follows that lower scores on a measure that influences control functions could be lower in the ACC. The weaker association between Trails Number Sequencing and Glx/tCr in the AUD group may point to other as yet unmeasured forms of dysfunction in the ACC.

There was also a significant gender interaction for functional impulsivity predicting mI/tCr, wherein the relationship was more pronounced in men than women (Fig. 3), although we consider this finding tentative. Interestingly, Jupp et al. (2020) reported that diminished levels of mI in frontal brain regions of male rats mediated a specific form of impulsivity linked to vulnerability for addiction, but results for female rats were not reported. Perhaps hormone levels, which are modulated by stress in the progression of AUD in men and women (Peltier et al. 2019), might contribute differently to impulsivity between the two sexes. It could be speculated that higher levels of mI/tCr in the ACC might interact with hormones in men to produce higher impulsivity.

Limitations

A limitation of this study was that 40 regression models were conducted to evaluate relationships between four metabolites and 10 neuropsychological measures, which could be considered a single ‘family’ of analyses. Using the Bonferroni-correction approach (threshold 0.05/40 = 0.00125), in order to correct for the family of analyses, of the five results, significant at P < 0.05, none was significant at P < 0.00125. In this regard, these five results from this study should be considered provisional.

However, these findings may be important, and eliminating them would increase the chance of missing valid relationships (a False Negative or Type II error). For context, sample sizes of 20 participants per group are sufficient to identify two-way interaction effects of 0.6 SD or larger with 80% power (Cohen 1988) considered 0.5 as ‘medium’ and 0.9 as ‘large’ effects), and the sample sizes we used exceeded those indicated by the power analyses. We interpret each result separately since we know of no overarching explanatory model pending further research. Nonetheless, the data make a crucial addition to the literature, as the sample examined represents a unique cross-section of participants that includes a sufficiently large cohort of women, which allows a novel examination of gender differences.

For Null Hypothesis Significance Tests involving P-values, we were cautious about the interpretation of P-values above 0.05; as such, we used confidence intervals (CIs) of mean differences intervals to illustrate the degree to which the groups were likely to differ. The sample size contributes to the width of this CI, as does sample variability, which in turn is influenced by the study design. Further, we calculated Bayes factors for group differences in the four metabolite levels, which supports similarities between the groups.

Another limitation of the present research is that, due to the cross-sectional nature of the study, we are unable to determine longitudinal changes over the course of long-term abstinence. However, our data are useful when interpreted in conjunction with existing longitudinal datasets of metabolite levels and drinking behaviors in shorter-term abstinence (Meyerhoff et al. 2013; Fritz et al., 2019) and provide important insight into what sustained abstinence means for ACC neurochemistry and neuropsychological outcomes. Moreover, further longitudinal studies can help to establish causal links between the long-term duration of abstinence and metabolite ratios and to elucidate relationships among metabolite levels and neuropsychological measures across groups.

It should also be noted that the potential impact of tobacco use on the metabolite ratios was not examined, which is relevant given that smoking can contribute to alterations seen in metabolite ratios (Durazzo et al. 2004; Gazdzinski et al. 2008) and can have meaningful implications for neuroimaging results (Luhar et al. 2013). There are opposing schools of thought regarding the quantification strategy for determining proton metabolite levels (Choi and Kreis 2021; Mandal 2012; Near et al. 2021; Öz et al. 2021; Tomiyasu and Harada 2022; Wilson et al. 2019). While recent recommendations suggest using unsuppressed water to calculate absolute metabolite concentrations (Abe et al. 2013; Jensen et al. 2005; Mon et al. 2012), T2 of unsuppressed water has been shown to differ in heavy alcohol-using groups (Silveri et al. 2014), reducing the feasibility of this quantification strategy in the current report. Using tCr as a normalizing denominator, while a common practice, can limit the interpretability of results (Tunc-Skarka et al. 2015; Zahr et al. 2016; Zahr and Pfefferbaum 2017); however, there was a lack of significant differences in the ratio of tCr/total proton signal between groups, which is consistent with prior studies (Kirkland et al. 2022; Marinkovic et al. 2022; Meyerhoff 2014; Silveri et al. 2014). The selection of the most ideal quantification strategy remains a work-in-progress (Choi and Kreis 2021). It is critical to guide decisions based on the existing body of research within a particular domain of research, such as in AUD (Kirkland et al. 2022), and particularly in light of subpopulations where data are scarce, i.e. women in long term recovery from AUD.

The same inclusion threshold of 21 drinks per week was used for the men and for the women, and because the bodies of men and women differ, the acute impact of an equal number of drinks for women may be greater than for men. For example, the DHHS guideline suggests moderate drinking levels of fewer than three drinks per day for men and fewer than two per day for women (US Department of Agriculture and US Department of Health and Human Services, 2020). Therefore, our minimum cutoff of 21 drinks per week for 5 years could reflect a selection bias wherein women who drank fewer drinks were not included in our sample. However, our final cohort generally drank more heavily. The mean DD scores were  ~10 for women and  ~12 for men, suggesting similar levels of severity, and only four AUD participants had DD levels below four (two men and two women).

Our sample had a heterogeneous LOS, which prevented us from specifying time points corresponding to the percentage of inferred changes we reported in the Results. Alcohol consumption and abstinence characteristics of our AUD cohort are consistent with estimates reported in the national population (World Health Organization 2018), thereby improving the generalizability of our results.

Conclusion

In conclusion, these findings demonstrate that brain metabolite levels in men and women with AUD, following long-term abstinence, do not differ from individuals without AUD. The data also provide evidence of associations between metabolite levels and measures of inhibitory control in the population of men and women with AUD in long-term abstinence. Additionally, there were no substantial differences in metabolite levels between men and women with AUD, suggesting that if metabolites are normalized to non-AUD levels, the current data do not support the notion that women have longer term impacts related to AUD on neurochemistry.

Acknowledgments

The authors thank Zoe Gravitz, Gordon Harris, Steve Lehar, Pooja Parikh Mehra, Diane Merritt, Susan Mosher Ruiz, and Maria Valmas for assistance with recruitment, assessment, data analyses, neuroimaging, or manuscript preparation. We also wish to acknowledge the Athinoula A. Martinos Center of Massachusetts General Hospital for imaging resources (1S10RR023401, 1S10RR019307, and 1S10RR023043). Finally, we would like to acknowledge the role of the research participants in making this study possible. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, the US Department of Veterans Affairs, or the US Government.

 

Conflict of Interest: The authors declare no financial or non-financial competing interests.

Funding

This work was supported by funds from the National Institute on Alcohol Abuse and Alcoholism (NIAAA) grants F31AA025824 to E.O., R01AA007112 and K05AA00219 to M.O.B., and K24AA025977, R01AA018153, and K01AA014651 to M.S., and from the US Department of Veterans Affairs Clinical Science Research and Development grant I01CX000326 to M.O.B.

Data availability

The data that support the findings of this study are available on request from the corresponding author, MMS. The data are not publicly available due to their containing identifiable personal information protected by HIPAA.