OUP user menu


Ingrid Agartz, Svante Brag, Johan Franck, Anders Hammarberg, Gaku Okugawa, Katarina Svinhufvud, Hans Bergman
DOI: http://dx.doi.org/10.1093/alcalc/agg020 71-78 First published online: 1 January 2003


Aims: The brain volume of chronic drinkers is known to partially recover with abstinence from alcohol. To investigate the relative contribution of grey and white brain matter to this process, magnetic resonance imaging and brain tissue segmentation was used to study brain tissue in acute alcohol withdrawal and abstinence in seven alcohol-dependent men. Methods: The patients were studied on three occasions; within 48 h after the last drink and approximately one month and two and a half months later. Total brain tissue class volumes [grey matter (GM) and white matter (WM) and cerebrospinal fluid (CSF)] were measured. Eleven healthy volunteers were scanned twice to serve as a control group. The alcohol-dependent patients were investigated with regard to drinking variables, neuropsychological performance and blood biochemistry. Results: In the alcohol-dependent patients, intracranial volume and total GM volume did not change between scan occasions, except in a single patient who demonstrated a GM increase of 4.8% (4.2% relative volume) between scans 2 and 3. For all patients, the increase in total WM volume ranged between 1.9 and 22.4% (absolute volumes) and 2.1 and 21.2% (relative volumes). Between scans 2 and 3, the increase in total WM volume ranged between 0.3 and 13.2% (absolute volumes), and between 1.5 and 14.0% (relative volumes). One patient resumed drinking and was investigated a second time during acute withdrawal. In this patient, the measured decrease of 8.1 and 8.5% of absolute and relative WM volumes corresponded to the size of the volume increase between scans 1 and 2. CSF, GM and WM volumes in the healthy subjects were constant over time. Conclusions: The results demonstrate that changes in brain volume during short-term abstinence in chronic alcohol-dependent patients are confined to the WM. The time limit of WM volume restitution is variable and continues longer than 3 weeks after withdrawal.


The partial recovery of brain volume, metabolism and cognitive function with prolonged abstinence from alcohol in chronic alcohol-dependent individuals is a well-documented phenomenon. Evidence for this has been accumulated from computer tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET) and neuropsychological studies.

Different investigators have quantified brain volume recovery with short-term abstinence from alcohol. Zipursky et al. (1989) reported reversibility of ventricular enlargement with short-term abstinence from alcohol. Shear et al. (1994) demonstrated that, in recently detoxified alcoholic men who were examined with MRI 1 month after their last drink, those who still abstained 3 months later exhibited white matter (WM) volume increases and cerebrospinal fluid (CSF) reductions over the follow-up interval, whereas those who resumed drinking did not. Pfefferbaum et al. (1995) obtained MRI scans on chronic alcoholics after an average of 12 and 32 days of sobriety with healthy controls scanned at comparable intervals and at 2–12 months after the second MRI. The results suggested that improvement in cortical grey matter (GM) volumes, sulcal and lateral ventricular volumes occur early in the course of abstinence, and that improvement in third ventricular volume appears later with continued abstinence. Resumption of drinking after a short period of abstinence arrests third ventricular volume improvement and produces WM loss. Schroth et al. (1988) found CSF volume reduction up to 31% after 5 weeks of abstinence but WM T2 values did not change on MR spin-echo sequences, which was interpreted to speak against rehydration. Mann et al. (1993) measured CT density measurements in 29 alcoholics during a 5-week period of controlled abstinence. The results revealed a significant reduction of CSF volume and a slight increase in CT density measures over the study period, which also speaks against the rehydration hypothesis. Trabert et al. (1995) found significant reversibility of alcohol brain shrinkage within 3 weeks of abstinence in younger (<38 years), but not in older, subgroups of alcoholic men. Studies of long-term brain changes with abstinence show that brain shrinkage is potentially reversible to some degree with abstinence for several months or years, but even after several years the brain may remain abnormal (review by Moselhy et al., 2001).

The precise neurobiological underpinnings of the volume recovery during abstinence are not known. Proposed explanations have been rehydration (vasopressin secretion may be suppressed during alcohol intoxication), regeneration (augmented dendritic growth) or changes in perfusion (vascularization) (Trabert et al., 1995). The rehydration hypothesis has not been confirmed by the majority of investigators using post-mortem examination, CT density measures or MR relaxation time constant estimates (Harper et al., 1988; Mann et al., 1993; Harper, 1998; Moselhy et al., 2001). It has been shown that, during high-dose alcohol ingestion and acute withdrawal, there is a significant but transient reduction in perfusion of the brain as shown by single photon emission tomography (SPECT), PET and measures of regional cerebral blood flow using Xenon 133 (133Xe-rCBF) investigations (Berglund and Risberg, 1981; Berglund et al., 1987; Tutus et al., 1998). The reduction was thought to lead to atrophic brain changes, as it is known for other chronic hypoperfusion states.

The recovery of total brain volume with alcohol abstinence in alcoholics appears greatest in the first weeks of sobriety (Pfefferbaum et al., 1995; Kril and Halliday, 1999). The effects of abstinence on metabolism have been studied by Volkow et al. (1994), who assessed the rate of recovery of regional brain metabolism using PET in alcoholics at 8–15, 16–30 and 31–60 days after the last use of alcohol. Recovery of brain metabolism (which is decreased during alcohol intake) occurred predominantly within 16–30 days of abstinence. From neuropsychological studies, improve-ments in short-term memory, abstract reasoning, spatial ability and visuo-motor co-ordination have been shown to occur during the first 2 weeks of abstinence (Moselhy et al., 2001).

To understand the mechanisms behind the recovery of brain volume, it is important to describe the role of the different brain tissues in this process. In the current study, we describe the quantitative change in brain tissue volumes during acute withdrawal and abstinence from alcohol in chronic alcohol-dependent patients and compared the results with the those obtained from healthy volunteers.


The protocol was approved by the Ethics Committee at the Karolinska Hospital, Stockholm, Sweden. The study was conducted in accordance with the Declaration of Helsinki.

Subjects characteristics

Alcoholic subjects.

Seven Caucasian male patients with chronic alcohol dependence, fulfilling DSM-IV (American Psychiatric Association, 1994) criteria for alcohol dependence and, upon entry into the study, also those of the Substance (Alcohol) Withdrawal Syndrome (DSM-IV). Six of the patients were scanned on three occasions and one was scanned twice. One patient (patient E in Tables 1 and 2, and Fig. 1) started drinking after scan 2. In this case, scan 3 was obtained less than 24 h after the last drink.

View this table:
Table 1.

Current age, age at onset of drinking, moderate and heavy drinking (years), respectively, drinking pattern, drinking style, abstinence symptoms, concomitant drug misuse, and intellectual impairment in the alcohol-dependent patients

View this table:
Table 2.

MR measurements, weight and scan interval in alcoholic patients A–G

Fig. 1.

Relative intracranial WM tissue volumes (ml) in seven alcohol-dependent patients and a reference group of 11 healthy control subjects measured over time (days). The alcoholic patients (named A–G and marked by thick lines) were, on day 1, in acute withdrawal from alcohol. The subsequent measures were obtained during sobriety. MR scan 3 of the alcoholic patient E was not included. The healthy subjects are numbered 1–11 and shown in thin black lines.

The alcohol-dependent patients were recruited from the Detoxification Unit, St Göran Hospital, and participated in an alcohol treatment programme at the Magnus Huss Clinic, Karolinska Hospital, Stockholm. Alcoholic patients who misused addictive drugs other than alcohol or nicotine were not recruited. Other exclusion criteria were presence of illicit drugs in the urine, signs or symptoms of current or past malnutrition, severely elevated liver function tests (i.e. elevated transaminases three times the normal level), signs and symptoms of liver cirrhosis [elevated PK (prothrombin complex assay) and albumin], significant somatic or psychiatric disease other than alcohol dependence, or loss of consciousness for more than 5 min due to head trauma.

Clinical testing was performed at the Magnus Huss Clinic, Karolinska Hospital, Stockholm. Drinking variables and related characteristics were assessed by the following standardized methods: Addiction Severity Index (ASI), alcohol module (McLellan et al., 1992), Alcohol Timeline Followback (TLFB) (Sobell and Sobell, 1992), 90 days back, Life Time Drinking History (LDH) (Skinner and Scheu, 1982), and Alcohol Use Disorders Identification Test (AUDIT) (Saunders et al., 1993). A Swedish version of the Alcohol Use Inventory was also used (Berglund et al., 1988). Age of onset, drinking pattern and style, recent alcohol intake, abstinence symptoms and other drug misuse for each patient are presented in Table 1.

All patients (except one) were neuropsychologically assessed on one occasion a few weeks after the clinical investigation. The following tests were given: the SRB tests (Synonyms, Reasoning and Block Design) of general intelligence to assess vocabulary, logical reasoning and visuo-spatial pattern reproduction; the Trail Making Test of visuo-spatial scanning, visuo-motor coordination and flexibility; the Claeson-Dahl Verbal Learning and Retention Test; the Benton Test of visuo-graphic short-term memory (Bergman, 1987). An overall assessment of intellectual impairment was also carried out using a 3-point rating scale (1 = no impairment, 2 = slight signs, 3 = definite impairment). The result is presented in Table 1.

The alcohol-dependent patients were scanned between January 2000 and April 2001 between 12.00 and 14.00. The mean (± SD) interval time between scans 1 and 2 was 28 ± 8 days (range 20–45 days; median 27 days) and between scans 2 and 3 was 63 ± 50 days (range 27–166 days; median 43 days). Scan interval time as well as body weight changes between scans 1 and 2, and between scans 2 and 3 for each patient are shown in Table 2. For mean age, height and body mass index (BMI) see Table 3.

View this table:
Table 3.

Characteristics of patients and controls

All patients received oral benzodiazepine treatment to control withdrawal symptoms at the time of the first MR scan. The mean ± SD dosage of benzodiazepines (oxazepam) was 325 ± 174 mg administered on a symptom triggered basis. The dose was tapered within 1 week. One patient (C) reported chronic daily use of 30 mg oxazepam without legal prescription. All patients had an intravenous tube during scanning to enable treatment of seizures or other acute effects of alcohol withdrawal if required. None of the patients received extra medication during scanning. There were no somatic or psychiatric complications during the MR investigation.

The patients were treated as out-patients during the study period. Abstinence from alcohol was ascertained from patient self-report in combination with clinical interviews performed by nurses, a psychiatrist (J.F.) and a behavioural scientist (A.H.), who were familiar with the patients, and from blood biochemistry [liver function markers, carbohydrate deficient transferrin (CDT), mean corpuscular volume (MCV) and blood alcohol concentration determined by breathalyser]. Blood biochemistry was obtained on each MR scan occasion. Values of gamma-glutamyl-transferase (GGT), transaminases, CDT, haemoglobin (Hb), haematocrit (Hct) and MCV at the times of scans 2 and 3, except in the case of the patient (E) who resumed drinking after scan 2, were continually reduced from scan 1 and approached normal values. One patient (patient A) was found to have a severely elevated GGT of 50 μkat/l (reference interval of GGT: 0.00–1.20 μkat/l) upon entry. Transaminases were within the permitted range. This patient was not excluded from the study. Patients who were found to drink alcohol during the study period were excluded.

Healthy subjects.

Eleven healthy subjects, all Caucasian, were included as a reference group. The 11 control subjects were scanned twice. The time interval between investigations was 6–77 days (range), mean interval ± SD was 35.3 ± 25.3 days, median interval was 36 days. All scans were obtained between 26 November 1999 and 9 June 2000.

The healthy subjects were recruited at the Section of Psychiatry, Department of Clinical Neuroscience, Karolinska Hospital. Five of the 11 control subjects were women. The healthy subjects had either taken part in previous studies as healthy subjects or were staff members at the Psychiatry Section. Exclusion criteria for the healthy volunteers were the same as for the patients, except the presence of alcohol dependence and alcohol withdrawal syndrome upon entry. They were all found to be physically and mentally healthy according to a psychiatric interview (SCID) (Spitzer et al., 1986), physical examination and blood biochemistry. Male subjects with 8 points or higher and women with 6 or higher in AUDIT were considered to be at risk for alcohol dependence and were therefore excluded. The healthy subjects did not receive any instructions with respect to alcohol consumption during the study period.

Weight fluctuation between scans 1 and 2 was seen in only three healthy subjects and represented an increase of 2 kg in one woman and a 1 kg increase and decrease, respectively, in two men. GGT, transaminases, CDT, Hb, Hct and MCV were within the normal range. For mean age, height and BMI for the healthy control subjects, see Table 3.

MR evaluation methods

MR data acquisition.

The subjects were investigated in a 1.5 T GE Signa Echospeed MR system (Milwaukee, WI, USA) at the MR-Centre, Karolinska Hospital, Stockholm. Selected pulse sequences were used in one scanning session, with a total duration of approximately 45 min. A fast spin echo (T2-weighted) axial oblique sequence was used for clinical evaluation. For segmentation, T1-weighted images, using a 3D spoiled GRASS sequence, were acquired with the following parameters: 1.5 mm coronal slices, 35° flip angle, TR 24 ms, TE 6.0 ms, 1 NEX, FOV 24 cm, acquisition matrix 256 × 192. T2-weighted images were acquired with the following parameters: 2.0 mm coronal slices, no gap, TR 6000 ms, TE 84 ms, 2 NEX, FOV 24 cm, acquisition matrix 256 × 192. From visual inspection, all scans were judged to be excellent, without any obvious motion artifact. All scans were evaluated for gross clinical pathology by a neuroradiologist. The quantitative analysis was performed blindly with regard to the diagnostic categories and scanning order.

Automated segmentation.

MR data analysis was performed using the software BRAINS (Andreasen et al., 1992, 1993). Measures in BRAINS, generated by automatic segmentation, have been carefully validated against manual tracings considered to be the gold standard (Harris et al., 1999). The segmentation is based on a 3D MR acquisition. The tissue classification programme can process unimodal (typically T1 volumes) or multimodal (T1 and T2, or T1, T2 and PD) data sets. The first step in the tissue classification process is the automatic identification of training classes that are entered into a discriminant analysis function, which classifies the tissue into GM, WM, CSF and venous blood. The generated image is a continuous representation of the tissue types. The continuous classification identifies the predominant (containing more than 50% of the tissue type) and next most likely tissue component (class) within each volume element (voxel). Details on the segmentation procedure and evaluation of validity were described by Harris et al. (1999). Reproducibility and reliability of this segmentation procedure have also been ascertained previously (Harris et al., 1999; Agartz et al., 2001).

The intracranial volume (ICV) was automatically delineated in each segmented MR volume using artificial neural networks (Magnotta et al., 1999). Total (intracranial) GM, WM and CSF tissue class volumes were obtained. Volumes from each hemisphere were combined in the calculations.

Statistical analysis

Since the number of subjects was too limited to permit a statistical comparison with the control subjects, the latter subjects were used as a reference group. In the alcohol-dependent patients, we calculated the per cent difference for the tissue class volumes between the first and the second scans and between the second and the third scans. The difference between the first and the second scans (∂1–2) was given as a per cent of the values obtained at the first scan and the difference between the second and the third scans (∂2–3) was given as a per cent of the values obtained at the second scan.

To correct for individual differences in head size, we calculated relative tissue class brain volumes (absolute volume/ ICV). Both the absolute and the relative values of GM, WM, brain volume and CSF tissue volume were reported.


Total intracranial, absolute and relative GM, WM, and CSF tissue class volume measures, scan interval, and body weight for each of the alcohol-dependent patients on the different scan occasions are presented in Table 2.

Median volume changes for each tissue class volume between scans 1 and 2 and between scans 2 and 3 for the alcoholic and the healthy subjects are presented in Table 4.

View this table:
Table 4.

Median volume change (ml) for alcoholic subjects between MR scans 1 and 2, and MR scans 2 and 3 and of healthy subjects between MR scans 1 and 2 of intracranial volume (ICV) and absolute (A) and relative (R) white matter (WM), grey matter (GM) and CSF tissue class volumes

The relative intracranial WM tissue volumes (ml) in the seven alcohol-dependent patients and the 11 healthy reference subjects from the first, second and third MR scans measured over time (days) are shown in Fig. 1.

In the seven alcohol-dependent patients, on all scan occasions, ICV changed within the range of –1.2 to 1.0%. The change in GM volume was within the range of –2.1 to 1.8% (absolute volumes) and –1.3 to 1.4% (relative volumes) in all subjects on all scan occasions, with the exception of patient A. GM of patient A changed 4.8% (4.2% relative GM volume) between scans 2 and 3 with scan 2 taking place after 20 days of abstinence.

Between the first and the second scans, the increase in absolute WM volume ranged between 1.9 and 22.4%, and between 2.1 and 21.2% for relative WM volume. Between the second and the third scans, the increase in absolute WM volume ranged between 0.3 and 13.2% and between 1.5 and 14.0% for relative WM volume. One alcohol-dependent patient (E) resumed drinking after scan 2 and was investigated a second time during acute withdrawal (measures from scan 3 are not included in Fig. 1). He demonstrated a decrease of 8.1 and 8.5% of absolute and relative WM volumes, respectively, compared with the absolute and relative volumes at the second scan.

In the alcohol-dependent patient B, the WM change occurred between scans 1 and 2. The third MR scan was not performed until 2 months after the second and, by then, no WM change was detectable. The WM volume of the alcoholic subject (F) did not change between scans 1 and 2 (28 days interval) but demonstrated a 13.2% (absolute) and 14.0% (relative) change between scans 2 and 3 (28 days interval). On neuroradiological inspection, this patient was found to have high signal intensity patches of degenerative/vascular origin in the central pons (size 0.5 × 0.5 × 2 cm) on T2-weighted images. He developed marked systemic hypertension during withdrawal. Patient A with severely elevated GGT and who demonstrated a GM as well as a WM increase between scans 2 and 3, had a marked enlargement of the subarachnoid space and WM hyperintensities, but no increase of the ventricular system. Significant organic brain pathology was not found in the other patients. The other two alcohol-dependent patients (D and C) had the greatest change in WM volume between scans 1 and 2, but showed at least some change between scans 2 and 3.

The tissue class volumes in the brains of the healthy subjects did not change (GM range: absolute, –1.8 to 1.2%; relative, –1.0 to 0.7%; WM range: absolute, –1.4 to 1.7%; relative, –1.3 to 4.1%), except for the CSF volumes that changed marginally (range: absolute, –3.2 to 4.1%; relative, –2.2 to 3.8%). ICV changed less than ±0.42%, except in one subject (–1.1%).

Relative WM and GM measurements were lower and CSF measurements greater in the alcohol-dependent patients, than in the healthy subjects, on all three scan occasions.


The present study is in keeping with previous findings that, in chronic alcohol-dependent patients, there is an increase in brain volume with abstinence from alcohol after a period of significant alcohol intake. The brain volume accordingly decreased in the patient who resumed drinking after a period of abstinence. We found that the brain volume increase in alcoholics who abstain from alcohol after a period of heavy drinking is almost entirely due to expansion of the WM.

ICV and GM volumes remained virtually constant across scans. In the alcohol-dependent patients, the change in ICV and GM volumes was equal or less than 1.2 and 2.1%, respectively, and occurred in both directions [except in one case (A) of severe brain atrophy and liver affection in which the GM and WM increases were found between scans 2 and 3]. The range of ICV and GM volume change was within the range of the healthy group, except that the alcoholics showed a marginally greater variability in GM (maximum 2.1% compared with maximum 1.2% in the healthy subjects). Since the proportion of GM is greater than WM in the brain, the same absolute difference in GM and WM gives a greater percentage difference in the WM (i.e. a certain per cent GM difference reflects a greater change in the absolute volume of the GM than the same per cent change of the WM). Still, the GM change was consistently small and fluctuated in both directions.

The decrease in the CSF volume can be explained by the corresponding expansion of the WM volume. Our results are in accordance with the results of Shear et al. (1994) who found altered WM and CSF volumes, but not with the results of Pfefferbaum et al. (1995), who also found GM changes. The reason that other groups have found significant reduction also in GM volume may depend on differences in the MR pulse sequences that have been used. Previous studies have shown that the segmentation programme which we use is robust and has high reliability and reproducibility. It is able to detect differences also in GM volumes which previous investigations have shown (Okugawa et al., 2002). The reason may also be that the number of subjects in this study is too limited.

Previous investigators have not been able to explain conclusively the reason for the volume increase. The proposed explanation of altered brain water content has not been confirmed by the majority of investigations (Schroth et al., 1988; Mann et al., 1993; Harper, 1998; Moselhy et al., 2001). Earlier studies have shown that, in alcoholics, the brain is hypo-perfused during acute withdrawal from alcohol (Berglund and Risberg, 1981; Berglund et al., 1987; Tutus et al., 1998). The fact that the volume increase that took place over the first month interval — but also later on — was limited to the WM, needs to be considered. In our patients, the change in WM volume was considerable and in the range of 17–22% in the two older subjects. A plausible explanation for the WM volume increase with continued abstinence from the acute withdrawal state is that changes in water regulation and distribution/perfusion in the brain are differently regulated in WM and GM, respectively. Abstinence-related release from the direct toxic effects of alcohol may also offer an explanation of the findings. However, other methods are required to determine the precise physiological events underlying the process of restitution of brain volume.

In four (A, C, D, F) of the five alcohol-dependent patients who remained abstinent and were investigated three times, a WM volume increase occurred after at least 1 month after acute withdrawal. In one (F) of these four patients, the WM increase occurred exclusively between scans 2 and 3. Patient B had scan 3 after 62 days of sobriety and, at that time, there was no WM volume change. Patient A had a GM increase after scan 2 on day 20 along with a small WM increase, but the bulk of the WM increase occurred between scans 1 and 2 when there was no GM change. The fact that we found a WM volume increase at the time of scan 3 after 2 months of sobriety in four of the five subjects shows that WM volume increase continued past at least 1 month (minimum 20 days) of sobriety. This is interesting, since, in alcoholics, a 3-week abstinence from alcohol has been considered to be adequate to obtain reliable and valid tissue volume measurements.

The temporal course of the WM change was different between the alcohol-dependent patients. Interestingly, the two older patients had the greatest WM change. The variability among these patients in regard to volume change and other organ manifestations secondary to alcoholism makes larger patient materials called for. The effects of lifetime alcohol intake as well as recent drinking, liver function and possibly also benzodiazepine treatment on brain volume plasticity need to be addressed and clarified.

Even after 2 months of abstinence from alcohol, the relative GM and WM values were lower in the alcohol-dependent patients, than in the control group. Damage not only to the WM, but also to the GM, is found in alcoholism and has been demonstrated by many investigators using brain imaging and post-mortem methods (Harper, 1998; Hommer et al., 2001). In post-mortem studies, the proportional distribution of the different brain tissues in non-alcoholic subjects is approximately 50% GM, 40% WM and 10% CSF (Miller et al., 1980). This is in accordance with the results we have obtained in the healthy subjects using this MRI segmentation method (Agartz et al., 2001).

The strength of our study is the excellence of MR methodology. The methodology has been previously evaluated with regard to validity and reliability (Harris et al., 1999; Agartz et al., 2001). However, we have not tested whether fluctuations in brain water content in different tissue compartments affect the validity of the segmentation. The limitations are that the sample size was too limited to determine predictors of the magnitude of WM change or the temporal course of the restitution. The intervals between the scans were not identical across subjects. We were also able only to use one gender, i.e. men. In conclusion, the volume expansion of the brain during an approximately 1-month interval following acute withdrawal from heavy alcohol consumption in seven chronic alcohol-dependent men was confined to the WM. In four of the five patients who were scanned three times the increase in WM volume continued past at least 20 days after cessation of drinking, suggesting that WM restitution after heavy drinking takes longer than 3 weeks. Since the sample size was limited with respect to number, further studies of clinically well characterized alcoholic patients will be necessary to clarify the mechanism behind the change in the brain matter that occurs with withdrawal and abstinence from alcohol as well as the factors which determine the magnitude and time course of the change.


The staff at the MR-Centre, Karolinska Hospital, are acknowledged for generous and qualified support. Margareta Gard-Hedander SRN and Else-Britt Hillner SRN, Magnus Huss Clinic, Karolinska Hospital, are thanked for help with patient recruitment, relapse prevention and collection of blood samples. Financial support was generously provided by the Fredrik and Ingrid Thuring Stiftelse and the Swedish MRC.


  • * Author to whom correspondence should be addressed at: Department of Clinical Neuroscience, Psychiatry Section, Karolinska Hospital, Stockholm SE-171 76, Sweden.


View Abstract