Rapid Partial Regeneration of Brain Volume During the First 14 Days of Abstinence from Alcohol

Abstract

Background

Chronic alcohol abuse leads to severe damage of the nervous system, including a change in cerebral metabolism and brain morphology. Global volume reductions of gray matter (GM) and white matter and an increase in cerebrospinal fluid (CSF) occur after severe alcohol consumption, but abstinent alcoholics also demonstrate a brain volume recovery. The aim of this study was to investigate whether volumetric amelioration takes place already within the first 2 weeks of abstinence.

Methods

All 49 alcohol-dependent patients included in this study were scanned within the first 24 hours of detoxification and after 2 weeks of supervised abstinence. Amelioration of volumetric brain loss in alcohol-dependent patients has been investigated, and brain volumes have been compared with 55 healthy control subjects using whole-brain segmentation and a voxel-based morphometric approach.

Results

On the first day of abstinence, the global CSF volume was larger and the GM volume was smaller in alcohol-dependent patients compared with healthy controls. The largest clusters with significant volumetric differences were in the cingulate gyrus, precentral and middle frontal gyrus, cerebellum, and insula. Already after 2 weeks of abstinence, a significant albeit partial recovery of GM volume occurred in several brain regions.

Conclusions

Our results show that recovery of GM volume in alcohol-dependent patients starts within a few days after detoxification but varies between brain regions. This suggests that the general ability to recover and the rate as well as onset of the recovery diverges for different brain regions.

Chronic alcohol abuse has negative psychological, sociological, and biomedical effects. Besides well-known physical consequences like liver cirrhosis or hypertension, alcoholism also leads to a severe damage of the nervous system, including detrimental effects on cognitive functions (Mann et al., 1999), and a change in cerebral metabolism (Cardenas et al., 2011) and brain morphology (Buhler and Mann, 2011).

Several studies demonstrate a global volume reduction in gray matter (GM) and white matter (WM) and an increase in cerebrospinal fluid (CSF) volume in alcoholics (Chanraud et al., 2009; Demirakca et al., 2011; Fein et al., 2002; Mechtcheriakov et al., 2007; Nicolas et al., 2000). Volume changes occur in the entire cerebral cortex with a focus on the frontal, parietal, and cingulate cortex, and in insula, thalamus, hippocampus, and cerebellum. Brain atrophy has been found not only in alcohol-dependent patients (Mechtcheriakov et al., 2007; Nicolas et al., 2000) but also in recently detoxified alcoholics (Demirakca et al., 2011) and those who were abstinent from alcohol from months to years (Bartsch et al., 2007; Chanraud et al., 2009; Fein et al., 2009). The reported areas of significant volume loss do not completely overlap in those groups. Nonabstinent or recently detoxified alcoholics, compared with healthy controls, show volumetric differences in the cerebellum (Bartsch et al., 2007; Mechtcheriakov et al., 2007; Nicolas et al., 2000). In alcoholic patients who have been abstinent for several weeks or months, this region showed no significant volumetric difference compared with social drinkers (Chanraud et al., 2009; Demirakca et al., 2011; Wobrock et al., 2009).

Previous studies on the effects of alcohol on brain volume concentrated on gender-related differences, but the findings were inconsistent. In some studies, the interaction of diagnosis and gender is driven by greater abnormalities in male patients (Fein et al., 2009; Pfefferbaum et al., 2001), while in other studies, greater differences in women cause the interaction (Hommer et al., 2001; Mann et al., 2005). In addition, in a recent study, no diagnosis by gender interaction has been found (Demirakca et al., 2011). This discussion is additionally complicated by other differences between men and women, for example in brain volume, body weight, and fat proportion, which all together could lead to different blood alcohol levels even when the same amount of alcohol is consumed.

A period of continued abstinence leads to brain volume recovery and restoration of cognitive functions (Agartz et al., 2003; Bartsch et al., 2007; Cardenas et al., 2007, 2011; Demirakca et al., 2011; Gazdzinski et al., 2008, 2010; Mann et al., 1999, 2005; Pfefferbaum et al., 1995; Wobrock et al., 2009). Sobriety can result in recovery of GM and WM volume and a reduction in CSF volume. The brain regions where GM volume increases include the cingulate gyrus, insula, cerebellum, hippocampus, parietal lobe, temporal lobe, anterior frontal gyrus, and periventricular and frontal edges. These longitudinal studies investigated the effect of continued abstinence lasting from 1 to 9 months. However, the time point of the first scanning session with respect to the date of last alcohol consumption has not been the same for all patients, and the difference range comprised several weeks. This may mask volumetric alterations and cover rapid recovery during the first days of detoxification and continued withdrawal.

To clarify the possibility of early recovery, we investigated volumetric brain changes within the first 2 weeks of abstinence. As the interval for possible changes seems to be comparatively short, we made certain to conduct scanning sessions in all alcohol-dependent patients within the first 24 hours of detoxification and then exactly 2 weeks later.

Materials and Methods

Participants

The studied sample consisted of 49 alcohol-dependent patients (9 women) and 55 healthy controls (13 women). Participants' characteristics are summarized in Table 1. All patients met DSM-IV and ICD-10 criteria for alcohol dependence and were recruited from the Department of Addiction Medicine at the Central Institute of Mental Health in Mannheim, Germany, where they underwent an inpatient alcohol withdrawal treatment. The patients consumed their last alcoholic drink on the day of admission or the day before. Two magnetic resonance (MR) scans (scan 1: during detoxification on the first day of abstinence; scan 2: 14 days later) were recorded, and abstinence of all patients was confirmed by unheralded alcohol breath tests within the 2 week interval. On the first day of abstinence, alcohol was still detectable in 35 patients at a level of 0.39 ± 0.40 g/l. In the remaining 14 patients, alcohol levels had reached zero shortly before magnetic resonance imaging (MRI) started. On the day of the first scan, all patients were free of benzodiazepines, but 12 patients had received clonidine (adrenergic alpha-2 receptor agonist) before the scanning session to treat withdrawal-associated hypertension or tachycardia. Within the 14 days of abstinence prior to the second MR scan, 36 alcohol-dependent patients were treated with benzodiazepines and 21 patients with clonidine. Correlation analyses revealed no association (p > 0.1) of the amount of clonidine or benzodiazepines with any detected brain volume difference or change at any time point. Alike, including administered clonidine or benzodiazepines quantities as an additional covariate in analysis of variance (ANOVA), did not significantly alter our results. Hence, we found no indication that the medications affected the results.

Table 1. Characteristics of Alcohol-Dependent Patients and Healthy Controls (mean ± SD)

	Alcohol-dependent patients			Healthy controls
	All	Female	Male	All	Female	Male
dd, drinking day; BMI, body mass index. ^a 1 drink contains 14 g alcohol.
N	49	9	40	55	13	42
Age (years)	47 ± 10.1	48.2 ± 6.9	46.8 ± 10.7	45.3 ± 11.9	43.9 ± 13.3	45.7 ± 11.5
Alcohol consumption in 3 months prior detox in g/dd (drinksa/dd)	211 ± 148 15 drinks	181 ± 101 13 drinks	218 ± 157 16 drinks	23 ± 18 2 drinks	15 ± 7 1 drink	25 ± 19 2 drinks
Proportion of cigarette smoker	80%	89%	77%	31%	31%	31%
BMI	24.3 ± 4.1	24.8 ± 3.9	24.2 ± 4.2	24.7 ± 3.5	22.4 ± 3.0	25.4 ± 3.3

The comparison group of 55 healthy controls was recruited by newspaper advertisement and was age- and sex-matched to the group of alcohol-dependent patients (see Table 1). A fraction of 20 healthy controls has also been studied twice with a time interval of 14 days between both measurements.

Exclusion criteria for all participants were any substance dependence except nicotine (and alcoholism in the patient group), any psychotropic medication in the last 3 months, positive urine drug screen, history of brain injury, other psychiatric or neurologic brain disease, any lifetime diagnosis of a psychotic disorder, hepatic encephalopathy, liver cirrhosis, severe medical illness (e.g., severe diabetes, HIV, polyneuropathy, etc.), and MRI-related exclusion criteria (e.g., metal implants). The study was approved by the ethics committee of the medical faculty Mannheim of Heidelberg University, and an informed written consent was obtained from all participants.

MRI Data Acquisition and Image Processing

The MRI images were acquired on a 3T Siemens Tim Trio system (Erlangen, Germany) using a T1-weighted MPRAGE sequence with a whole-brain coverage (192 slices, 1 mm slice thickness, 256 mm field of view, 1 mm³ resolution).

The imaging data have been processed using the voxel-based morphometry toolbox (VBM8, http://dbm.neuro.uni-jena.de/vbm/) implemented in the SPM8 package (http://www.fil.ion.ucl.ac.uk/spm/software/spm8/). All images were corrected for bias effects, normalized to the Montreal Neurological Institute (MNI) template using both linear and nonlinear transformation, and classified into GM, WM, CSF, and nonbrain tissue within the same generative model (Ashburner and Friston, 2005; Luders et al., 2009). For normalization, the option of low-dimensional spatial normalization (as implemented in VBM8) was chosen. For a more thorough removal of the remaining nonbrain tissues, we used the option “Thorough Cleanup.” The GM and WM images were modulated by multiplication with the Jacobian determinant to account for volume changes resulting from affine and nonlinear transformations during normalization (Good et al., 2001). Finally, modulated GM and WM images have been smoothed with an 8 mm³ FWHM Gaussian kernel. We checked the output images at each step of processing; the sample homogeneity was controlled using the VBM8 toolbox. The preprocessed data have been further analyzed using SPM8 (see Statistical Analysis).

Global Fractional Compartment Volumes

Global GM, WM, and CSF volumes have been calculated by integrating the respective values over all voxels of the segmented images in native space (i.e., in spatial correspondence to the original data). The sum of these 3 fractions was used as an estimate for the total intracranial volume (TIV; TIV = GM + WM + CSF) (Klauschen et al., 2009; Lüders et al., 2002; Smith et al., 2007). To account for the known gender differences in brain size (larger overall GM, WM, and CSF volumes in men than in women) (Filipek et al., 1994; Lüders et al., 2002), we normalized the fractional volumes for each participant to the individual TIV. These ratios, for example, GM to TIV, have been used for further comparisons using IBM SPSS Statistics Release 20.0.0 PASW 18 (Chicago, IL).

Regional Fractional Compartment Volumes

In addition to the global volume changes, we were interested in the volumetric changes in certain regions of interest (ROIs), for which volume reduction has been shown in alcohol-dependent patients and a significant volume recovery because of abstinence was reported. These brain regions included insula, cingulate gyrus, cerebellum (including vermis), anterior frontal gyrus, hippocampus, and parietal and temporal lobe (Bartsch et al., 2007; Cardenas et al., 2007; Chanraud et al., 2009; Demirakca et al., 2011; Gazdzinski et al., 2008; Mechtcheriakov et al., 2007; Pfefferbaum et al., 1995).

Individual GM images of the ROIs were calculated by using the image calculation function in SPM8. For each ROI, a mask of the specified region was created using the WFU PIC Atlas (Maldjian et al., 2003), and ROI voxels of the modulated MRI images were cut out. The GM volume in each ROI was estimated by integrating the GM voxel values over all voxels within the ROI (Klauschen et al., 2009; Lüders et al., 2002; Smith et al., 2007). To account for brain size effects, the values were normalized individually to the TIV and data were further analyzed using SPSS (PASW 18).

Statistical Analysis

The global and regional brain volume estimates, normalized to the individual TIV, were then related to the mean of the corresponding gender subgroup of healthy controls. This approach allows visualization of interactions accounting for gender effects in global and regional volume estimates.

The tissue partition maps from the VBM image processing were used in analysis of covariance (ANCOVA) and regression analysis (based on a general linear model SPM8) to calculate voxel-wise differences between healthy controls and alcohol-dependent patients. Age, gender (except for the evaluation of gender and diagnosis interaction), and TIV served as covariates. Furthermore, we investigated a possible influence of smoking behavior on the brain volume, by applying 2-sample t-tests for smokers and nonsmokers (with SPM8) for alcohol-dependent patients and healthy controls, respectively (see Table 1).

The correlation between drinking severity (assessed by the Form 90 interview; Tonigan et al., 1997) and GM volume within the ROIs was assessed by calculating partial correlation coefficients using SPSS (PASW 18). Age was used as a nuisance variable.

The significance level for the VBM analysis has been set to p < 0.05; whole-brain family wise error (FWE) corrected. For all other statistical testing, the significance level was defined to p = 0.05.

Results

Volumetric Differences Between Controls and Alcohol-Dependent Patients on the First Day of Abstinence

Whole-Brain Volume

Group comparisons of the fractional cerebral compartment volumes revealed significant differences between patients and healthy controls. CSF/TIV (p < 0.001) and GM/TIV (p < 0.001), but not WM/TIV (p > 0.4) differences gained significant results. A gender effect was detected only for the GM/TIV ratio: it was significantly higher in women compared with men (p = 0.002). Analyzed for both genders separately, the CSF/TIV ratio was 22% higher (p = 0.001) in alcohol-dependent women and 20% higher in male patients (p < 0.001), compared with the respective gender-matched healthy control group. Furthermore, we observed a smaller GM/TIV ratio in alcohol-dependent men and women (both 7%; p < 0.001 in men, p = 0.002 in women), while the WM/TIV ratios did not differ significantly from healthy control values (p > 0.2). No gender by diagnosis interaction was found (Table 2 and Fig. 1). Smoking status had a significant influence on brain volume, with the nonsmoking subgroup exhibiting smaller GM/TIV (p = 0.006) and WM/TIV (p = 0.049) ratios and larger CSF/TIV ratio (p < 0.001) compared with the smoking subgroup in alcohol-dependent patients. In healthy controls, only the WM/TIV ratio was significantly higher (p = 0.037) in the nonsmoking subgroup.

Figure 1. Percent differences of global gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) and regional GM volumes of alcohol-dependent patients relative to healthy controls (=100%) for each gender, respectively. TIV, total intracranial volume.

Table 2. Relative and Regional Brain Volumes: Results of a 2 × 2 Factorial Analysis of Variance with the Factors Sex and Diagnosis (mean ± SD)

Global volumes	Male		Female		p-Value
	HC (42) (in%)	Pat (40) (in%)	HC (13) (in%)	Pat (9) (in%)	Gender	Diagnosis	Gender × Diagnosis
HC, healthy controls; Pat, alcohol-dependent patients; CSF, cerebrospinal fluid; TIV, total intracranial volume; GM, gray matter; WM, white matter.
Age (years)	45.7 ± 11.5	46.8 ± 10.7	43.9 ± 13.3	48.2 ± 6.9	0.960	0.321	0.553
CSF/TIV	17.1 ± 2.0	20.5 ± 3.7	16.0 ± 1.9	19.5 ± 2.1	0.106	<0.001	0.989
GM/TIV	44.6 ± 2.0	41.6 ± 3.0	46.6 ± 2.0	43.3 ± 2.3	0.002	<0.001	0.751
WM/TIV	38.3 ± 1.5	37.8 ± 2.2	37.5 ± 1.4	37.3 ± 1.4	0.099	0.423	0.681
Regional GM volumes
Insula/TIV	1.04 ± 0.08	0.96 ± 0.11	1.07 ± 0.09	0.99 ± 0.06	0.157	<0.001	0.929
Cingulate Gyrus/TIV	1.87 ± 0.13	1.71 ± 0.14	1.95 ± 0.14	1.79 ± 0.09	0.010	<0.001	1.000
Cerebellum/TIV	6.56 ± 0.53	6.07 ± 0.62	7.06 ± 0.43	6.53 ± 0.51	0.000	<0.001	0.848
Frontal Gyrus/TIV	6.01 ± 0.38	5.59 ± 0.48	6.34 ± 0.47	5.72 ± 0.29	0.027	<0.001	0.322
Hippocampus/TIV	0.58 ± 0.03	0.55 ± 0.05	0.59 ± 0.04	0.59 ± 0.03	0.027	0.217	0.128
Parietal Lobe/TIV	5.22 ± 0.39	4.73 ± 0.39	5.46 ± 0.31	4.95 ± 0.36	0.014	<0.001	0.914
Temporal Lobe/TIV	6.85 ± 0.39	6.46 ± 0.48	6.98 ± 0.43	6.65 ± 0.40	0.122	0.001	0.799

Regional Volumes

The estimated TIV scaled GM volumes of the ROIs (insula, cingulate gyrus, cerebellum, anterior frontal lobe, hippocampus, parietal lobe, and frontal lobe) were significantly (p < 0.01) smaller in alcohol-dependent patients compared with healthy controls. The regional volume loss (ROI-GM/TIV), compared with healthy controls of the same sex, was similar in the investigated ROIs except for the hippocampus and temporal lobe, and the mean difference between the groups ranged from 6 to 9% in men (p < 0.005) and 5 to 10% in women (p < 0.03). In the hippocampus and the temporal lobe, the TIV scaled ROI-GM difference was significant only for men (p < 0.005) but not for women (p > 0.08). A 2 × 2 factorial ANOVA with gender and diagnosis as factors revealed a significant effect of diagnosis for all ROIs except the hippocampus and a significant effect of gender for all ROIs except insula (see Table 2). Comparable to the relation between gender and the whole-brain GM/TIV ratio, ROI-GM/TIV ratios were higher for women than for men. Nevertheless, the interaction of gender and diagnosis was not significant for any region. In the ROIs (except for the hippocampus and anterior frontal gyrus), the ROI-GM/TIV ratio was significantly (p < 0.03) smaller in the nonsmoking subgroup of alcohol-dependent patients (N = 10) compared with the smoking subgroup. In the hippocampus and the anterior frontal gyrus, the same trend has been observed, but the difference did not comply with our significance level. For healthy controls, no significant differences (p > 0.1) between the smoking and nonsmoking subgroup were found.

Correlating GM/TIV ratio for each ROI with parameters of drinking severity controlling for effects of age revealed no significant correlation.

VBM Analysis

To compare voxel-wise GM volumes of alcohol-dependent patients with those of healthy controls, a VBM analysis was performed, where age, gender, and TIV have been used as covariates of no interest. Several regions with significantly less GM volume in alcohol-dependent patients compared with healthy controls were found, while a significant contrary effect did not occur. Regions with the largest GM volume loss in alcohol-dependent patients were cingulate gyrus, precentral gyrus, middle frontal gyrus, cerebellum, and insula (listed in declining order for cluster size) (Fig. 2A and Table 3). In a voxel-based approach, the extent of volume loss in alcohol-dependent patients did not correlate with the amount of alcohol consumed over the last 3 months.

Figure 2. (A) VBM analysis of gray matter volume loss in alcohol-dependent patients compared with healthy controls for the first scan at the first day of abstinence (yellow) and 14 days later (orange) and (B) the results of a paired t-test (increased gray matter: Scan 2 > Scan 1) for alcohol-dependent patients (green). Family wise error corrected, p < 0.05. In A and B, the first image of each pair displays the same slice, and the white bar indicates the position of the second image in each pair.

Table 3. Peak Voxel of Gray Matter Volume Loss in Alcohol-Dependent Patients Compared with Healthy Controls (Age, Gender and TIV as Covariates), p < 0.05 FWE Corrected, Cluster Size ≥30 Voxels

MNI label	p (FWE)	T	Z	x	y	z	Cluster size
MNI, Montreal Neurological Institute; FWE, family wise error; TIV, total intracranial volume.
Cingulate gyrus	<0.001	7.71	6.81	−9	−25	45	6,475
Precentral gyrus	<0.001	7.70	6.80	−45	−13	49	4,924
Middle frontal gyrus	<0.001	7.19	6.43	51	15	30	4,184
Cerebellum (Uvula)	<0.001	6.71	6.07	32	−82	−35	2,134
Insula	<0.001	6.53	5.94	−39	−12	13	912
Insula	<0.001	6.51	5.92	39	−18	12	529
Middle frontal gyrus	<0.001	6.40	5.84	24	35	39	189
Cerebellum (Uvula)	<0.001	6.38	5.83	−32	−84	−36	263
Superior parietal lobule	<0.001	6.27	5.74	−22	−70	43	190
Superior frontal gyrus	<0.001	6.24	5.72	−15	42	45	273
Middle temporal gyrus	<0.001	5.99	5.52	64	−4	−11	179
Superior frontal gyrus	<0.001	5.93	5.47	−27	54	19	133
Cerebellum (Pyramis)	0.001	5.81	5.38	8	−73	−36	65
Inferior frontal gyrus	0.001	5.79	5.36	−54	18	3	66
Insula	<0.001	5.75	5.33	39	6	6	137
Middle frontal gyrus	0.001	5.61	5.22	34	−10	67	70
Middle temporal gyrus	0.002	5.35	5.00	60	−31	1	55
Cerebellum (inferior semi-lunar lobule)	0.001	5.28	4.94	32	−76	−54	70
Precuneus	0.005	5.27	4.93	33	−72	30	31

To investigate a possible diagnosis by gender interaction, a 2 × 2 factorial ANCOVA with gender and diagnosis as 2 factors (age and TIV as covariates) revealed no significant interaction. Also, no gender-related difference in GM volume was found for any voxel cluster.

Regarding the influence of smoking behavior on GM volume, we found no voxels with significantly less GM volume in smoking alcohol-dependent patients compared with nonsmoking alcohol-dependent patients and vice versa. In the healthy control group, only 1 cluster (cluster size = 20 voxels) in the cerebellum (MNI coordinates: −30, −45, −57) showed significantly lower GM volume for smoking healthy controls, compared with nonsmoking healthy controls; no voxels with significantly higher GM volume for the smoking subgroup were detected.

Differences in WM volume were also analyzed despite the fact that VBM is optimized for GM differences only. A significant WM volume loss (p < 0.05, FWE corrected) was found for 2 clusters in the frontal lobe (subgyral) in the right and left hemisphere with a cluster size of 34 and 25 voxels, respectively (no figure shown).

Volumetric Changes After 2 Weeks of Abstinence

Whole-Brain Volume and Regional Volumes

When comparing the data acquired on the first day of abstinence and then 14 days later, paired t-tests revealed a significant decrease in the CSF/TIV ratio and an increase in GM/TIV in the group of alcohol-dependent patients. No significant change of WM/TIV ratio occurred (see Table 4). Significant GM/TIV volume changes in the paired analysis were observed for all ROIs except the hippocampus. No significant differences were present in healthy controls, which have been measured twice (Table 4).

Moreover, a significant difference (p = 0.003) occurred in the GM/TIV increase in the cerebellum of smoking alcoholic patients compared with nonsmoking alcoholic patients, with nonsmoking alcoholics exhibiting larger GM/TIV increase in the cerebellum. Neither other ROIs nor the global volume measures displayed any significant differences between smokers and nonsmoking patients.

VBM Analysis

A paired t-test in alcohol-dependent patients revealed several brain regions with significant GM volume recovery after 2 weeks of abstinence (Fig. 2B). The largest clusters with significant recovery were found in the cingulate gyrus, temporal gyrus, parietal lobule, cerebellum, and precuneus (Table 5).

Table 4. Global and Regional Gray Matter Volume Changes (mean ± SD) Measured at the First Day of Abstinence (1. Scan) and After 14 Days of Abstinence (2. Scan)

Global volumes	Alcohol-dependent patients (49)			Healthy controls (20)
	1. Scan (in%)	2. Scan (in%)	p paired t-test	1. Scan (in%)	2. Scan (in%)	p paired t-test
CSF, cerebrospinal fluid; TIV, total intracranial volume; GM, gray matter; WM, white matter.
CSF/TIV	20.3 ± 3.5	20.1 ± 3.4	0.006	17.0 ± 2.0	17.0 ± 2.0	0.508
GM/TIV	41.9 ± 2.9	42.3 ± 2.8	0.001	45.0 ± 1.9	45.1 ± 2.0	0.317
WM/TIV	37.7 ± 2.0	37.6 ± 2.0	0.130	38.0 ± 1.6	37.9 ± 1.6	0.527
Regional volumes
Insula/TIV	0.96 ± 0.10	0.97 ± 0.11	<0.001	1.06 ± 0.09	1.05 ± 0.09	0.970
Cingulate gyrus/TIV	1.73 ± 0.14	1.75 ± 0.14	<0.001	1.89 ± 0.14	1.88 ± 0.15	0.421
Cerebellum/TIV	6.15 ± 0.63	6.27 ± 0.56	<0.001	6.73 ± 0.50	6.77 ± 0.54	0.276
Frontal gyrus/TIV	5.61 ± 0.45	5.68 ± 0.45	0.001	6.05 ± 0.39	6.04 ± 0.40	0.886
Hippocampus/TIV	0.56 ± 0.05	0.56 ± 0.05	0.936	0.58 ± 0.04	0.58 ± 0.03	0.722
Parietal lobe/TIV	4.78 ± 0.39	4.85 ± 0.40	<0.001	5.27 ± 0.40	5.29 ± 0.40	0.258
Temporal lobe/TIV	6.50 ± 0.47	6.56 ± 0.46	0.003	6.79 ± 0.41	6.82 ± 0.40	0.179

Table 5. Peak Voxel of GM Volume Recovery (Increased GM: Scan 2 > Scan 1) in Alcohol-Dependent Patients, p < 0.05 FWE Corrected, Cluster Size ≥10 Voxels

MNI label	p (FWE)	T	Z	x	y	z	Cluster size
MNI, Montreal Neurological Institute; FWE, family wise error; GM, gray matter.
Cingulate gyrus	<0.001	7.70	6.18	6	−46	37	2,082
Superior temporal gyrus	<0.001	6.47	5.46	54	−10	6	246
Posterior cingulate	<0.001	6.21	5.29	−16	−67	4	138
Superior parietal lobule	<0.001	6.20	5.29	−22	−64	45	167
Cerebellum (Culmen)	<0.001	6.19	5.28	38	−54	−30	112
Cerebellum (Uvula)	<0.001	5.86	5.07	−26	−81	−36	247
Precuneus	0.004	5.81	5.03	−10	−51	63	46
Precuneus	0.008	5.78	5.01	8	−49	51	27
Precuneus	0.011	5.66	4.93	33	−51	49	21
Precuneus	0.007	5.63	4.91	9	−37	51	31
Middle occipital gyrus	0.018	5.54	4.85	−40	−75	6	12
Cerebellum (Uvula)	0.020	5.53	4.84	24	−84	−35	10
Cerebellum (Pyramis)	0.005	5.52	4.84	45	−67	−44	42
Precuneus	0.020	5.47	4.80	−15	−63	28	10
Insula	0.014	5.43	4.77	39	−12	12	16

Two weeks of abstinence produced no significant changes in WM, but we would like to point out that the VBM method is not optimized to detect WM changes.

The amount of alcohol consumed over the previous 3 months showed no significant influence on the changes within 2 weeks abstinence. In addition, no effect of smoking on volume recovery was detected in this voxel-wise approach.

Discussion

In this study, we focused on early recovery of brain volume occurring during the first 14 days after detoxification. Volumetric differences were studied on 3 partially nested levels: whole-brain volumes, regional GM volumes (ROIs), and voxel-wise analyses. The information content is different for all 3 levels, and hence, the results are not necessarily the same. All patients were scanned within 24 hours after the last alcohol consumption, and at this time point, the comparison of the whole-brain volumes in patients and healthy controls revealed significantly smaller GM/TIV ratio and higher CSF/TIV ratio in alcoholic patients, while WM/TIV ratio showed no significant difference. After 2 weeks of abstinence, a significant increase in GM volume and a significant decrease in CSF were found. In contrast to previous studies (Agartz et al., 2003; Demirakca et al., 2011; Hommer et al., 2001), no global WM volume difference was detected. GM/TIV ratios in the investigated ROIs were significantly smaller in the alcohol-dependent patients, and VBM analyses showed that volume loss in these regions spanned over large clusters (except for the hippocampus, where VBM found no voxels with significant differences). Nonetheless, GM volume reduction affects other regions as well, namely the precentral gyrus, and precuneus. Volume reduction in alcohol-dependent patients in the frontal lobe, parietal lobe, temporal lobe, and the cerebellum was previously reported by Cardenas and colleagues (2007). These authors additionally found a reduction in the occipital lobe volume. In another study, Demirakca and colleagues (2011) observed no volume shrinkage in the cerebellum in alcohol-dependent patients 4 to 37 days after the last alcohol consumption but GM shrinkage most prominent in the insula and the cingulate gyrus, regions which also showed a significant volume decrease in our study.

Cigarette smoking is common among alcohol-dependent patients, and hence, it was previously discussed whether changes in brain volume in alcohol-dependent patients can be solely attributed to alcohol consumption. In contrast to previous studies (Das et al., in press; Gazdzinski et al., 2005; Liao et al., in press), we found no evidence of smoking effects on GM volume in our sample; among the healthy controls, the whole-brain GM/TIV ratio showed no significant differences for the smoking and nonsmoking subgroup, and the VBM analyses revealed significant differences only in 1 small-sized cluster. On the contrary, the nonsmoking subgroup (N = 10) from the alcoholic group exhibited larger brain volume loss on global and regional level (not in the VBM analyses). On the first day of abstinence, nonsmoking alcohol-dependent patients had smaller GM/TIV and WM/TIV and larger CSF/TIV ratios than the smoking subgroup, while neither age nor alcohol consumption was significantly different between the subgroups. Brain volume shrinkage prevailed also 14 days later except for the cerebellum, where the difference between the smoking and nonsmoking alcohol-dependent patients has faded. This finding agrees with the significantly higher larger GM/TIV increase in the cerebellum within the first 2 weeks of abstinence for nonsmoking alcoholics compared with smoking alcoholics. In agreement with Yeh and colleagues (2007), we suggest that smaller regional brain volumes in the first measurement are related to faster volume gains. Nevertheless, we would be cautious with the interpretation of results derived from only 10 nonsmoking alcohol-dependent patients, as there is still a chance of obtaining a random result because of the small sample size.

After 14 days of abstinence, brain volume differences in alcohol-dependent patients could be identified in basically the same regions observed on the first day of abstinence, except for precuneus and middle temporal gyrus that exhibited significant volumetric differences compared with healthy controls. In several regions, the differences were reduced as indicated by decreased cluster sizes but still significant. It is important to note that the extent of volume recovery from the first day of abstinence to 14 days afterward depends on the brain region, and hence, the recovery rate seems to vary between regions. For example, recovery in the cerebellum seems to happen faster in comparison with the cingulate gyrus, as indicated by a greater volume gain within the 2-week interval. This is also reflected in the intraindividual GM volume increase within the 2-week interval. A significant increase in GM volume was found only in the cingulate gyrus, temporal gyrus, parietal lobule, cerebellum, and precuneus, but not in the precentral gyrus or frontal gyrus. The brain volume recovery in these 2 latter regions seems either to be slower or starting later in comparison to the recovery in the cerebellum, and might in the long run not be completely reversible (as has been suggested by Fein et al., 2006). These differences in recovery rate or recovery onset of atrophic GM regions become more evident when we compare our results with previous VBM studies investigating the effects of different time intervals with respect to withdrawal.

Volumetric shrinkage in the cerebellum, including the vermis, was only found in those VBM studies that were performed in nonabstinent alcoholics (Mechtcheriakov et al., 2007; Nicolas et al., 2000). Studies that started at a later time point after detoxification (Chanraud et al., 2009; Fein et al., 2009) reported no volumetric difference in the cerebellum. In addition, a gain in cerebellar brain volume was reported in a longitudinal study where the first scan was obtained within the first week after detoxification (Bartsch et al., 2007). In the study by Demirakca and colleagues (2011) where the time interval for the first scan spanned 4 to 37 days into abstinence, neither cerebellar shrinkage nor its gain was observed in a second MR scanning session after 3 months. Our results show that already after 2 weeks of abstinence, the cerebellar volume significantly increases compared with the first day of abstinence. Volumetric differences in the cerebellum are unlikely to be found if the scan is performed after 2 weeks of abstinence. In addition, we found a small volumetric recovery in the insula that showed a large volume gain after 3 months of abstinence in the previous study by Demirakca and colleagues (2011). Studies on brains of alcoholics that were abstinent for months to years (Chanraud et al., 2009; Fein et al., 2009) indicate that some alcohol-induced brain damage might not be reversible at all. Altogether this leads to the conclusion that the general ability for recovery, the recovery rate, and/or its onset varies between different brain regions.

Limitations

The aim of the study was to examine the brain volume differences between healthy controls and alcohol-dependent patients on their first day of abstinence and to study the recovery within the first 14 days of abstinence. To examine alcohol effects, the participants had to meet several exclusion criteria (e.g., no substance dependence except nicotine [and alcohol for the patients], no history of brain injury, no lifetime diagnosis of a psychotic disorder, etc.), and we matched both groups for age and gender to make them as homogenous as possible. However, we could not control for every possible difference between the groups that might possibly affect brain morphology (e.g., IQ, education, risk factors for Alzheimer disease, etc.). Hence, we cannot conclude that our findings of differences in brain morphology between the healthy controls and the alcohol-dependent patients are solely attributed to the misuse of alcohol. Nonetheless, to the best of our knowledge, morphologic changes in the order observed have not been described for any of the risk factors or group differences not taken into account here, and the results on the brain volume changes within the 14 days of abstinence are unaffected by those considerations.

To perform the VBM analyses, we used the SPM8 software package, which allows a whole-brain VBM analysis including the cerebellum. Compared with the cortex, the cerebellum has a different structure, and for VBM analysis, this creates a particular challenge in terms of accurate normalization. A specific cerebellum-optimized procedure was established (Diedrichsen, 2006). A recent publication shows that common whole-brain VBM methods tend to underestimate volumetric changes in the cerebellum (Kuhn et al., 2012).

Acknowledgments

This study was funded by the DFG through a center grant (SFB636, project D7) and within the framework of Nationales Genomforschungsnetz (NGFN⁺ No. 01GS08152, see www.ngfn-alkohol.de and Spanagel et al., 2010) by the Bundesministerium für Bildung und Forschung.