MRSI and DTI: a multimodal approach for improved detection of white matter abnormalities in alcohol and nicotine dependence

Abstract

Our previous proton magnetic resonance spectroscopic imaging (¹H MRSI) studies showed that the frontal lobe white matter (WM) in smoking recovering alcoholics (sRA) had lower concentrations of N-acetylaspartate (NAA), a marker for neuron viability, compared to both nonsmoking recovering alcoholics (nsRA) and a control group of nonsmoking light drinkers (nsLD). Using diffusion tensor imaging (DTI) in a similar population, we found lower fractional anistropy (FA), a microstructural measure of WM fiber integrity, in regions of specific fiber bundles within frontal WM of recovering alcoholics compared to light drinkers. In this study, we hypothesized that in these regions of lower FA, NAA concentrations in the alcoholic groups are lower than in non-alcoholic controls. We hypothesized further that sRA have lower regional NAA concentrations than nsRA. We retrospectively analyzed existing ¹H MRSI data by quantitating metabolite concentrations from voxels that corresponded to previously identified WM regions of lower FA, and from a control region of normal FA in alcoholics. We found significant NAA concentration differences between groups in regions of abnormal FA. In particular, sRA had significantly lower NAA concentration than nsLD, but in no region was NAA significantly lower in nsRA than nsLD. Furthermore, no NAA group differences were detected in a frontal WM region of normal FA. These results indicate regionally localized NAA loss within the frontal WM, and specifically NAA loss in regions of low FA. Compared to our previous lobar analyses, DTI-guided MRSI analysis allows the selective evaluation of small WM regions with microstructural injury, thereby increasing statistical power to detect relevant pathology and group differences. DTI-guided MRSI analyses promise to contribute to a better understanding of brain injury in alcohol and nicotine dependence and, by extension, perhaps in other neurodegenerative diseases as well.

INTRODUCTION

It is well documented that alcohol use disorders (AUD) are associated with functionally significant in vivo morphological and biochemical abnormalities, as well as changes to brain white matter microstructure (1,2). Furthermore, chronic cigarette smoking is prevalent in those with AUD (3–5), and recent studies indicate that chronic cigarette smoking in non-alcoholic populations also adversely affects many of the same brain regions that are compromised by alcohol dependence (6–10). In vivo proton magnetic resonance spectroscopic imaging (¹H MRSI) allows for noninvasive and simultaneous measurement of various endogenous metabolites from the neocortical gray matter (GM), lobar white matter (WM), basal ganglia, and cerebellum. Concentrations of N-acetylaspartate (NAA), choline-containing compounds (Cho), creatine-containing compounds (Cr), and myo-inositol (m-Ino) can be obtained in a regionally selective manner (11,12). NAA is found almost exclusively in adult neurons and is virtually absent in glial cells (13,14). Thus, NAA serves as a surrogate measure of neuron viability and reduced NAA concentrations in AUD may reflect neuronal loss, atrophy of axons or dendrites, or abnormal neuronal metabolism, including mitochondrial dysfunction (1,12,15). Phosphatidylcholine, a major constituent of myelin and cell membranes, is broken down into glycerophosphocholine and phosphocholine, which are the primary choline-containing metabolites giving rise to the Cho resonance (11,16). Another important cell membrane component is m-Ino, which has been characterized as an astrocyte marker (17) and an osmolyte (18). The MR-detectable Cr corresponds to the summed concentrations of intracellular creatine and phosphocreatine, both of which are involved in the bioenergetics of neuronal and glial tissue (19).

Our previous ¹H MRSI studies showed that concurrent alcohol dependence and chronic cigarette smoking are associated with greater abnormalities of brain metabolite concentrations in major lobes and subcortical brain regions than chronic alcohol or cigarette use alone, both at 1 week and 1 month of abstinence from alcohol (20,21). The frontal WM, the largest volume of contiguous WM in the human brain, appears to be preferentially affected during alcohol dependence (1). Traditional clinical structural images of this region may appear rather homogeneous and reveal no gross pathology in AUD. Our ¹H MRSI studies revealed significantly lower NAA concentration in frontal WM of recently detoxified smoking recovering alcohol dependent individuals (sRA) compared to both nonsmoking recovering alcohol dependent participants (nsRA) and nonsmoking light drinking controls (nsLD) (20).

Magnetic resonance diffusion tensor imaging (DTI) measures the water diffusion in the brain tissue, as it is constrained by bundles of tightly packed axons, myelin sheaths, and cytoskeletal structure in white matter. DTI metrics include fractional anisotropy (FA), an index of directional water movement that is low in cerebral spinal fluid (isotropic diffusion) and relatively high along the main axes of WM fibers (anisotropic diffusion) (22). Lower FA observed in the corpus callosum and throughout the WM of the centrum semiovale of alcohol dependent cohorts suggests demyelination of fiber tracts and/or axonal degeneration (23). Using tract based spatial statistics of 1.5 T DTI data, we found that FA in regions of frontal and cortico-striatal WM tracts was lower in 11 1-week-abstinent alcoholics compared to ten non-alcoholic healthy controls (24). Specifically, in anterior WM, we found lower FA in the regions of the superior corona radiata (CR) adjacent to the lateral ventricles, the internal capsule, and the corpus callosum.

Here, we report on retrospective analyses of an existing larger ¹H MRSI data set that included volunteers previously described (21) who also participated in the DTI study described above (24). The overall goal of these analyses was to evaluate whether the select brain regions that demonstrated microstructural injury on DTI also show corresponding metabolite abnormalities, to further describe the nature of these abnormalities, and to test if these smaller regions provide greater sensitivity in measurements of the effects of chronic drinking and smoking on the brain compared to our conventional MRSI data obtained from the entire WM of the frontal lobe. To achieve this goal, we extracted spectra with signals for NAA, Cho, m-Ino and Cr from our existing ¹H MRSI data, which spatially corresponded to WM regions that demonstrated abnormal FA in our DTI study. Metabolite spectra were also extracted from the same regions in nsLD. We tested the hypotheses that (1) in smaller white matter regions of relatively low FA in the alcohol dependent group, NAA concentrations in both sRA and nsRA are lower than in nsLD and (2) in these small white matter regions, sRA have lower NAA concentrations than nsRA.

METHODS

Participants

For this retrospective analysis, we used ¹H MRSI data that had been collected for 26 nsLD (47.9 ± 6.6 years of age; one female), 26 sRA (50.2 ± 8.6; one female), and 22 nsRA (49.3 ± 9.0 years; one female) at approximately 1 month of abstinence from alcohol. All groups were equivalent on age. Primary inclusion criteria were current DSM-IV diagnosis of alcohol dependence or abuse (American Psychiatric Association, 1994), fluency in English, consumption of greater than 150 alcoholic drinks per month (one alcoholic drink equivalent =13.6 g pure alcohol) for at least 8 years prior to enrolment for men, and consumption of greater than 80 drinks per month for at least 6 years prior to enrolment for women. Primary exclusion criteria are fully detailed in Reference (20). In summary, no participant had a history of neurologic, general medical or psychiatric conditions known or suspected to influence neurocognition with the exception of the following: hepatitis C, type-2 diabetes, hypertension, major depression, substance-induced mood disorder, generalized anxiety disorder, and panic disorder. These comorbidities were permitted in recovering alcoholics (RA) given their high prevalence in those with alcohol use disorders: (25–27). RA who met DSM-IV criteria for current or past substance abuse were included, and past substance dependence (greater than 5 years prior to enrolment) was allowed, whereas current opioid replacement therapy (e.g. methadone) was exclusionary. Approximately 95% of RA participated in continued outpatient substance abuse treatment programs at the San Francisco VA Medical Center or Kaiser Permanente after their baseline magnetic resonance studies through the time of the 1-month assessment. RA attended these programs 3–4 days per week, were given random weekly drug screens, and breath alcohol levels were acquired randomly or in the case of suspected or obvious intoxication. No participant reported alcohol or substance use between enrolment and comprehensive neuro-psychological assessment, and chart review of available records confirmed that no participant tested positive for illicit/non-prescribed substances or alcohol over this interval. Additionally, prior to all assessments, participants’ urine was tested for common illicit substances, and they were evaluated for recent ethanol consumption via breathalyzer. No participant tested positive for the above-listed substances or ethanol at the time of assessment.

MRSI regions of interest and data processing

The multi-slice inversion recovery MRSI data (scan repetition time 1800 ms, inversion time 165 ms, and echo time 25 ms) were available in a comprehensive database, with each voxel (spectrum) of a particular SI slice coded by coordinates on a 64 × 64 grid that corresponded to the field of view of the SI data acquisition. Nominal voxel size was (0.8 × 0.8 × 1.5) cm³ or approximately 1 ml. The three slices of the MRSI data set were spatially aligned with a T2-weighted MRI data set (on which the specific smaller regions of interest were selected, see below). The database also contained information about anatomical location and tissue type that contributed to each of the SI voxels. It also contained average atrophy-corrected metabolite levels calculated from GM and WM of the major lobes, but not corrected for potential metabolite relaxation time differences between groups (20). In the remainder of this manuscript, we refer to these metabolite levels as ‘concentrations.’ In this retrospective analysis, we compared age equivalent groups of sRA, nsRA, and nsLD by extracting from the database spectra corresponding to coordinates of small regions of interest (ROIs) encompassing the areas of low FA described above. These ROIs were further divided into superior and anterior CR, splenium (SCC) and genu of the corpus callosum (GCC), and anterior (ALIC) and posterior limbs of the internal capsule (PLIC) (see Fig. 1). In addition, another ROI directly adjacent to the superior CR was selected (termed ‘nonCR’) for comparison, as it represents a region in which FA values did not differ as much between alcoholics and controls of our DTI analyses described above (2% vs. 5% in the CR ROI) (24).

Figure 1.

(A) Axial MRI slice through the basal ganglia with ROIs superimposed. (B) Axial MRI slice through the supraventricular brain with ROIs superimposed. nonCR, white matter region adjacent to the superior CR that is not part of the corona radiata.

The MRSI data set and its coaligned T2-weighted MRI were loaded into the FITT module of SITOOLS (28), a software that has been used extensively to process spectroscopic imaging data (see e.g. Reference (20)). On the MRI, ROIs were circumscribed as shown in Fig. 1 and the corresponding MRSI voxel coordinates for each ROI were noted. For the superior CR, anterior CR, nonCR, and the corpus callosum, a rectangular region containing the desired white matter volume was identified on the MRI. Given the morphology of the internal capsule, its voxels were selected by noting the individual coordinates of MRSI voxels that contained the targeted tissue. For the superior CR, anterior CR, nonCR, PLIC, and ALIC ROIs, voxel coordinates from left and right hemispheres were noted separately. For each ROI, the corresponding spectra were then extracted from the database. The database also contained information about tissue composition for each MRSI voxel, based on the co-aligned MRI data set that had been segmented previously into tissue types and major anatomical subdivisions. Thus, for each ROI, voxels that were included in the final ROI-specific analysis were also selected based on their WM tissue fraction (volume of WM in the voxel divided by the total tissue volume in voxel) and their exclusion of voxels with large amounts of GM and with more than 33% of cerebral spinal fluid. Only voxels with sufficiently high WM tissue fraction were used in this analysis. ‘Sufficiently high’ was defined as 70% or more frontal WM (frontal WM) for anterior CR, superior CR, and nonCR; 50% or more WM for the GCC and SCC; and 30% or more WM for the ALIC and PLIC. For the purpose of this retrospective analysis only, and to aid in the analysis of spatial metabolic heterogeneity within the large frontal lobe WM, metabolite peak integrals were also averaged over all voxels with 70% or more WM contribution from the entire frontal lobe. The average number of spectra that contributed to the ROIs, after screening out those with insufficient WM content and unsatisfactory data quality, were not statistically different for RA and nsLD groups and were as follows: bilateral frontal WM (114.2 ± 30.9 voxels per subject), bilateral superior CR (41.6 ± 10.8), bilateral anterior CR (4.6 ± 1.6), bilateral nonCR (5.8 ± 3.7), bilateral PLIC (4.5 ± 2.0) and SCC (6.6 ± 5.4), GCC and ALIC voxels were not further analyzed because the number of available spectra with acceptable data quality and WM content precluded meaningful analyses.

Statistical analyses

Average and standard deviation of metabolite concentrations from the selected MRSI voxels of each ROI and the frontal WM volume were computed in institutional units. As preliminary analyses indicated no concentration differences between hemispheres, spectra from left and right hemispheric ROIs were combined for further analyses. Analyses of variance (ANOVAs) were evaluated for group differences in each of the individual ROIs (α =0.05) for all metabolites. Significant ANOVAs for Cho, mI and Cr were followed by pairwise t-tests comparing sRA, nsRA, and nsLD. All ANOVAs for NAA were followed by pairwise t-tests to evaluate the hypothesized regional NAA concentration differences between groups. A significantly higher lifetime average drinks per month ( p =0.01) was observed in sRA (279 ± 97) relative to nsRA (194 ± 114); therefore all comparisons between these groups were covaried for lifetime average drinks per month. α levels of follow-up pairwise t-tests among groups were corrected for multiple comparisons via a modified Bonferroni procedure (29) that adjusts α (0.05) according to the number of pairwise comparisons (3) and the average intercorrelation among metabolites across regions of interest (r =0.40). This procedure yielded a corrected α of p =0.026 for all pairwise comparisons. To compare the magnitude of group differences for metabolites in different ROIs, effect sizes (ES) were calculated according to the differences between two group means divided by the average of the two group standard deviations (i.e. Cohen’s D).

RESULTS

Mean NAA and Cho concentrations for each group are listed in Table 1, separately for the entire frontal WM and the five different ROIs, together with group statistics (ANOVA and follow-up tests) and ES.

Table 1.

Regional metabolite concentrations in institutional units (mean ± standard deviation)

		All fWM	Superior CR	Anterior CR	nonCR	SCC	PLIC
NAA	nsLD	30.90 ± 3.14	33.61 ± 3.17	29.36 ± 3.62	32.04 ± 3.57	30.68 ± 4.33	31.67 ± 4.35
	nsRA	30.26 ± 3.91	32.31 ± 4.58	28.63 ± 3.77	31.92 ± 3.79	29.18 ± 4.14	31.39 ± 3.72
	sRA	28.95 ± 3.06	30.66 ± 3.15	27.63 ± 3.46	30.56 ± 4.24	27.44 ± 3.35	29.23 ± 3.10
	ANOVA	p =0.11	p =0.017	p =0.26	p =0.37	p =0.037	p =0.059
	sRA versus nsLD	p =0.014	p =0.001	p =0.049	p =0.10	p =0.006	p =0.015
		ES =0.66	ES =0.94	ES =0.50	ES =0.38	ES =0.84	ES =0.65
	sRA versus nsRA	p =0.10	p =0.074	p =0.18	p =0.14	p =0.078	p =0.01
		ES =0.38	ES =0.43	ES =0.28	ES =0.34	ES =0.46	ES =0.63
	nsRA versus nsLD	p =0.27	p =0.13	p =0.25	p =0.46	p =0.12	p =0.41
		ES =0.18	ES =0.34	ES =0.20	ES =0.033	ES =0.36	ES =0.068
Cho	nsLD	6.05 ± 1.04	6.33 ± 1.09	5.96 ± 1.05	5.57 ± 0.85	5.56 ± 0.95	6.17 ± 0.88
	nsRA	6.17 ± 0.72	6.41 ± 0.88	5.96 ± 1.05	5.88 ± 0.75	5.13 ± 0.83	5.78 ± 0.94
	sRA	6.06 ± 0.89	6.29 ± 0.91	5.92 ± 0.94	5.62 ± 0.96	5.21 ± 0.82	5.83 ± 0.51
	ANOVA	p =0.87	p =0.90	p =0.98	p =0.46	p =0.21	p =0.23
Cr	nsLD	18.85 ± 2.02	19.16 ± 2.10	18.69 ± 2.22	18.36 ± 2.34	18.21 ± 2.62	21.75 ± 2.12
	nsRA	19.74 ± 1.99	19.95 ± 2.16	18.52 ± 2.26	19.09 ± 3.02	18.92 ± 2.86	21.75 ± 3.02
	sRA	19.36 ± 2.22	19.59 ± 2.50	18.67 ± 1.78	18.83 ± 2.78	17.71 ± 2.84	21.51 ± 2.16
	ANOVA	p =0.34	p =0.48	p =0.96	p =0.59	p =0.25	p =0.94
mI	nsLD	17.25 ± 2.68	17.86 ± 2.68	17.40 ± 2.73	16.59 ± 2.57	18.33 ± 2.69	16.64 ± 3.17
	nsRA	18.55 ± 2.64	19.02 ± 2.88	18.09 ± 3.13	18.16 ± 2.37	18.64 ± 2.56	17.95 ± 2.42
	sRA	17.88 ± 3.10	18.42 ± 3.05	17.24 ± 3.28	17.78 ± 2.98	17.52 ± 2.76	16.28 ± 2.31
	ANOVA	p =0.29	p =0.38	p =0.62	p =0.12	p =0.38	p =0.22

nsLD, (n =23–26 per region of interest); nsRA, (n =19–22) and ; sRA, (n =19–26).

Group differences in NAA concentration for small WM ROIs with abnormal FA

Metabolite concentrations were compared across the groups for the entire frontal WM (conventional analysis) and for each smaller ROI separately (DTI-guided MRSI analyses). A trend for group differences in the entire frontal lobe WM was observed [F (2, 73) =2.27, p =0.11]. In follow up t-tests, sRA demonstrated significantly lower NAA than nsLD in the entire frontal WM ( p =0.014); nsRA did not differ significantly from sRA or nsLD. Examination of ES suggests that the greatest NAA concentration difference occurred between sRA and nsLD (see Table 1). ANOVAs for the smaller ROIs showed significant differences among groups for NAA in the superior CR [F (2, 73) =4.30, p =0.017] (Fig. 2), the SCC [F (2, 66) =3.47, p =0.037]), and a trend for group differences in the PLIC [F (2, 65) =2.96, p =0.059]. Pairwise t-tests indicated sRA had significantly lower NAA than nsLD in the superior CR ( p <0.001) and SCC ( p =0.006), and there were trends for lower NAA in sRA than nsRA in the superior CR ( p <0.11) and SCC ( p <0.06). In the PLIC, sRA demonstrated significantly lower NAA than both nsRA ( p =0.01) and nsLD ( p =0.015). nsRA and nsLD were not significantly different in any ROI (all p >0.11). Importantly, no significant group differences were observed for NAA in the nonCR ROI, [F (2, 64) =1.01, p =0.37] (Fig. 2). Furthermore, Cho, Cr, or m-Ino concentrations were not significantly different between groups in any of the ROIs.

Figure 2.

NAA concentrations (i.u.) in frontal white matter by regions of interest. Significance levels indicated above each set of data is from an ANOVA (analysis of variance) test comparing the NAA concentration across the three subject groups.

DISCUSSION

The primary goal of this study was to determine if abnormalities of ¹H MRSI derived metabolite concentrations were apparent in those brain regions that also showed microstructural abnormalities by DTI in two similar cohorts of alcohol dependent individuals. We retrospectively analyzed an existing ¹H MRSI data set to study metabolite concentrations of the selected small regions of WM that demonstrated abnormalities in DTI measures of FA. Examining the ROIs that demonstrated lower FA in alcohol dependent participants compared to non-drinking controls allowed for complementary assessment of abnormalities in MRSI-derived surrogate markers of neuronal viability, gliosis, cellular bioenergetics, and membrane synthesis and turnover. The regional analyses of metabolite levels indicated significant group differences for NAA (marker of neuronal viability) in superior CR and SCC, but not when NAA was averaged over the entire volume of the frontal lobe WM. This suggests alcohol dependence (and/or cigarette smoking) may have localized adverse effects in anatomically and functionally distinct regions in the frontal WM. Taken together, the localized NAA and DTI findings suggest regional axonal vulnerability within the frontal lobe to the effects of chronic and excessive alcohol consumption and/or chronic smoking (1,30). Such multimodality neuroimaging approaches may offer unique and complementary information and provide a more detailed understanding of the nature and extent of neuropathology in AUD and other biomedical conditions.

Using conventional analysis of NAA concentration in the entire frontal WM, the group differences in this cohort were consistent with those of a previous study of a smaller cohort utilizing average concentration data from large lobar regions (20). In that study, the NAA concentration difference in frontal WM between the alcoholic and non-alcoholic groups was statistically significant, whereas in the present analysis we only found a trend to a group difference ( p =0.11). This lack of significance may be explained by the percentage of WM contributing to the SI voxels being analyzed. In our earlier study, all voxels contained 90% or more WM, whereas in this study, all voxels contained 70% WM or more. The lower percentage was required to insure that sufficient numbers of voxels were included in the analyses of the small anterior CR and nonCR ROIs. When using the criterion of 90% or more frontal WM for analyzed voxels in this cohort, NAA differences across groups had similar ES as our previous smaller cohort, and the previously reported significantly lower frontal lobe WM NAA levels in sRA compared to nsRA were apparent. Comparisons of these two analyses also suggest that the metabolite differences in alcohol dependent patients are greater in WM than in neocortical GM, which constitutes the other tissue type in the analyzed voxels. This is also consistent with significant WM injury reported in a treatment-naive heavily drinking cohorts.

In the small ROI analyses of this report, significant NAA concentration differences between groups in the superior CR and the corresponding absence of NAA differences in the nonCR indicate regionally localized NAA abnormalities within frontal WM, and specifically lower NAA in regions of low FA. The co-localization of group differences in FA and NAA concentrations in the absence of corresponding Cho and m-Ino abnormalities suggests abnormalities in axonal microstructure or metabolism, or reduced axonal density in this cohort of 1-month-abstinent alcohol dependent individuals. Furthermore, the recurring trends for lower NAA in sRA relative to nsRA (which were statistically significant in PLIC), the significantly lower levels in sRA compared to nsLD in multiple regions, and the absence of significant NAA differences between nsRA and nsLD suggest that chronic cigarette smoking in this alcohol dependent sample was associated with greater region-specific axonal injury in frontal and subcortical WM. The corona radiata and the internal capsule are components of cortical–subcortical circuits implicated in higher-level cognitive functions (31,32), whereas the corpus callosum contains the majority of fibers interconnecting the cerebral hemispheres. Thus, compromised integrity of the WM in these brain regions may contribute to the poorer neurocognitive function observed in our sRA versus nsRA cohorts (33,34).

Axonal injury (as suggested from decreased WM NAA) may not be the only substrate for the compromised neurocognition observed in AUD. Our longitudinal MRSI studies in similar cohorts show that initially abnormally low lobar Cho concentrations at entry into AUD treatment recover more rapidly than NAA levels during abstinence from alcohol (21), suggesting more swift remyelination and/or glial changes than improvement of neuronal integrity. The data in this analysis were obtained from treatment-seeking alcohol dependent individuals after approximately 1 month of abstinence from alcohol. At this time, Cho and m-Ino levels may have normalized. Thus, our spectroscopic data may not be sensitive to very early (<4 weeks) glial changes during abstinence from alcohol. However, NAA reductions at 1 month of abstinence from alcohol were also demonstrated by other groups (35,36). The absence of significant Cho and m-Ino reductions in both sRA and nsRA, at 1 month of abstinence, suggests that axonal microstructural injury may be more persistent than abnormalities in glial or membrane function in this cohort.

The methodological combination of DTI and ¹H MRSI was previously used in multiple sclerosis to examine the nature of injury to the corpus callosum (37). The investigators found no correlation between callosal NAA concentration and callosal FA, which they interpreted to reflect that the decreased callosal NAA might be due to reversible metabolic dysfunction rather than irreversible structural damage to axons. We were not able to correlate these measures in this study, because our DTI and MRSI data were obtained at different time points and not in the exact same participants. Nevertheless, we found NAA reductions (at 1 month of abstinence) in regions of abnormal FA (measured at 1 week of abstinence), but not in regions with normal FA. This information from two MR modalities suggests that the WM microstructural abnormalities in this cohort are associated primarily with axonal metabolic disturbances, and not necessarily with demyelination, oligodendroglia or astrocyte abnormalities.

Due to better spatial resolution than provided by MRSI, DTI analyses can identify specific WM regions and fiber bundles that are specifically and severely affected by disease. These regions would be indistinguishable when analyzed as part of a larger WM region as done in our conventional MRSI data analysis, simply because frontal WM has a rather homogeneous appearance on structural MRI that generally serves as a guide for MRSI data extraction. Thus, our conventional MRSI data analysis that takes the average metabolite concentrations of entire regions (e.g. all frontal WM) will not capture metabolic differences in particular fiber tracts that appear most affected by alcohol dependence or other pathology. On using DTI to provide more regionally specific or localized WM abnormalities, examination of metabolite concentrations in corresponding regions of lobar and subcortical WM becomes easier. Combining MR modalities in this manner may increase our understanding of the nature and magnitude of neurobiological abnormalities associated with AUD and their changes during abstinence from alcohol, and may have use in other neurodegenerative pathologies as well.

Limitations of this study include limited overlap between participants of this study and the previous FA study; lack of data from earlier in the recovery process that may show differences in those metabolites that may recover faster than NAA; and the manual extraction of SI voxels that does not rely on DTI abnormalities directly overlaid on co-aligned multimodal images. It is also possible that the results were influenced by factors not directly evaluated in this study, such as nutrition, exercise, previous exposure to environmental cigarette smoke or genetic predispositions. Additionally, we did not assess for personality disorders in our cohort which may contribute to the neurobiological and neurocognitive abnormalities observed in AUD, particularly antisocial personality disorder (38–40). Finally, majority of the participants were male, which did not allow for the examination of the potential effects of sex on our outcome measures. Therefore, the results of this report must be considered preliminary in nature. Nevertheless, multimodal neuroimaging provides unique, complementary, and converging data that suggest compromised tissue in regions implicated in the development and maintenance of substance dependence. On a more general note, DTI-guided MRSI analysis allows the identification of most severely impacted WM that manifests no gross structural abnormalities on clinical imaging, thereby increasing the power to detect statistically significant differences in localized metabolite concentrations between study groups.

Abbreviations used

ALIC

anterior limbs of the internal capsule

antCR

anterior corona radiata

AUD

alcohol use disorders

Cho

choline-containing compounds

Cr

creatine-containing compounds

CR

corona radiata

DTI

diffusion tensor imaging

ES

effect sizes

FA

fractional anisotropy

fWM

frontal white matter

GCC

genu of the corpus callosum

GM

gray matter

¹H MRSI

proton magnetic resonance spectroscopic imaging

m-Ino

myo-inositol

NAA

N-acetylaspartate

nonCR

ROI directly adjacent to the superior CR

nsLD

nonsmoking light drinkers

nsRA

nonsmoking recovering alcoholics

PLIC

posterior limbs of the internal capsule

RA

recovering alcoholics

ROIs

small regions of interest

SCC

splenium of the corpus callosum

sRA

smoking recovering alcoholics

supCR

superior corona radiata

WM

white matter