| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Australian National University, ACT, Australia (K.J.A., C.R.-M., J.M., R.K., C.v.S., T.D.W., B.R.); University of Melbourne, Victoria, Australia (A.F.J.); and the University of New South Wales, NSW, Australia (W.W., P.S.).
Address correspondence and reprint requests to Kaarin J. Anstey, PhD, Centre for Mental Health Research, Australian National University, Canberra ACT 0200, Australia. E-mail: kaarin.anstey{at}anu.edu.au
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: Brain magnetic resonance imaging scans from 385 adults recruited through a community survey were analyzed. Automated segmentation and manual tracing methods were used to obtain brain subvolumes and automated methods were used to obtain quantification and localization of white matter hyperintensities. Visual measures of cortical atrophy were obtained as were data on health and lifestyle factors. Alcohol consumption was assessed with the Alcohol Use Disorders Identification Test.
Results: In men, weekly alcohol consumption had a positive linear association with ventricular volume and gray matter and a negative linear association with white matter. In women, weekly alcohol consumption had a nonlinear relationship with cerebrospinal fluid and white matter. Alcohol consumption was not associated with white matter hyperintensities, corpus callosum size, hippocampal or amygdala volumes in analyses adjusting for confounding variables.
Conclusion: An association between alcohol consumption and brain atrophy is evident at the population level. In women, detrimental effects of alcohol on the brain appear to occur at lower levels of consumption. It remains possible that low levels of alcohol consumption have neuroprotective benefits but is clear that high levels of consumption are detrimental.
Key Words: alcohol drinking • white matter • gray matter • hippocampus • sex differences • corpus callosum
Abbreviations: AC-PC = anterior commissure–posterior commissure; ACT = Australian Capital Territory; APOE = apolipoprotein allele; ARIC = The Atherosclerosis Risk in Communities; AUDIT = Alcohol Use Disorders Identification Test; CHS = Cardiovascular Health Study; CSF = cerebrospinal fluid; DNA = deoxyribonucleic acid; FFE = fast field echo; FLAIR = fluid attenuation inversion recovery; FOV= field of view; ICV = intracranial volume; MMSE = Mini-Mental State Examination; MRI = magnetic resonance imaging; NEX = number of acquisitions; NSW = New South Wales; TE = echo time; SPM = statistical parametric mapping; TI = inversion time; TR = repetition time; WMH = white matter hyperintensities.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
The three largest studies examining alcohol and the brain focused on white matter changes referred to as white matter hyperintensities (WMHs) that manifest as high signal intensities in T2-weighted magnetic resonance imaging (MRI). White matter hyperintensities are commonly detected in the brains of healthy older adults; have been associated with poorer physical health, cognitive, balance, and gait deficits; and are thought to be ischemic in origin (7). WMHs are associated with brain atrophy and although some risk factors have been established, the effect of alcohol consumption on their origin is unclear. The Atherosclerosis Risk in Communities (ARIC) Study of 1909 adults aged 45 to 64 years found no association between alcohol consumption and either stroke risk or white matter grade (8). This study did not include heavy drinkers and was based on a relatively young sample. In contrast, the Cardiovascular Health Study (CHS) involving brain MRI scans of 3660 adults aged 65 years or more found an inverse U-shaped association between white matter WMH and alcohol consumption with moderate alcohol consumption being associated with reduced prevalence of WMHs. This study also found moderate drinking to be associated with reduced risk of cerebral infarction (9). The Rotterdam study also found an inverse U-shaped association between WMH and alcohol consumption (10). The nonlinear associations observed are in keeping with epidemiologic investigations of the relationship between alcohol consumption and cognitive functioning (2,11), dementia (12), and cardiovascular disease (13).
Both case–control and population-based studies have shown that alcohol consumption is associated with whole brain atrophy. Both the ARIC and the CHS study found a dose-dependent relationship with atrophy (8,9). Studies of alcoholic men and women in young, middle, and late adulthood have found that, in comparison with healthy control subjects, they have reduced cortical gray matter and white matter volumes and increased ventricular size (5,6,14). More specifically, a large study of 769 adults aged 17 to 79 years localized the effect of alcohol consumption to reduced gray matter in the superior frontal and parietal cortices (15). Increases in total white matter, hippocampal, cerebral, and cerebellar volumes have also been observed in alcoholics after abstinence (16,17), and it has been argued that this is not wholly as a result of rehydration but may also reflect regeneration (18).
Investigations of specific brain structures such as the hippocampus and corpus callosum in relation to alcohol consumption have used mainly clinical samples or case–control designs. For example, in one study, the total area of the corpus callosum was significantly smaller in 14 hospitalized alcoholic women aged 30 to 50 years compared with 13 hospitalized alcoholic men and controls (19). The midsagittal area of the corpus callosum, the genu, and the body of the corpus callosum have also been reported as smaller in alcoholics compared with controls (20). There is evidence that smaller hippocampi may be evident in adolescent-onset (21) and adult-onset alcohol use disorders (22). The one population study to examine hippocampal and amygdala volumes found that alcohol consumption was not associated with atrophy in these regions, but that apolipoprotein (APOE) e4 allele carriers had larger hippocampal and amygdala volumes (10).
Several studies showing adverse effects of alcohol have compared hospitalized alcoholics with healthy control subjects and such studies potentially confound acute effects of heavy drinking with chronic effects of alcohol consumption on the brain. Small case–control studies may be subject to selection effects such that control groups differ from clinical groups on variables that may confound the association between alcohol and the brain. A number of the studies reviewed did not report whether the control groups were alcohol abstainers or light or moderate drinkers. Given the evidence of nonlinear relationships between alcohol and cognitive performance (23,24), as well as alcohol and a number of other variables (25), information on the control group is required to enable a full understanding of results in relation to alcohol consumption more generally. A strength of population-based studies is that they enable evaluation of the relationship between alcohol consumption and brain indices across the full distribution of alcohol consumption.
The present study sought to examine the associations among current alcohol consumption, brain atrophy, and WMHs in a community-based sample. Brain MRI measures that have been previously associated with alcohol in clinical studies (corpus callosum area, hippocampal volumes) and population-based studies (WMH, cortical atrophy, total gray matter, white matter, and cerebrospinal fluid [CSF]) were investigated. On the basis of previous findings from case–control studies and population-based studies, we hypothesized that WMH would increase with levels of alcohol consumption and that hippocampal volumes and gray and white matter would decrease with alcohol consumption, reflecting a general association between alcohol consumption and brain atrophy. We expected to find a linear association between brain atrophy and alcohol consumption and in general a stronger effect of alcohol on the brain in women than in men. Factors potentially explaining or confounding the association between brain measures and alcohol consumption were also adjusted for in regression analyses.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Magnetic Resonance Imaging Procedure
Imaging was conducted with a 1.5-Tesla Gyroscan (ACS-NT; Philips Medical Systems, Best, The Netherlands) for T1-weighted three-dimensional structural and T2-weighted fluid attenuation inversion recovery (FLAIR) sequence MRI. A two-dimensional scout midsagittal cut for anterior commissure–posterior commissure (AC-PC) plane alignment was first acquired. Then three-dimensional structural MRI was acquired in coronal orientation using a T1-weighted fast field echo (FFE) sequence (repetition time/echo time/number of acquisitions [TR/TE/NEX] = 28.05/2.64/2; flip angle = 30°; matrix size = 256 x 256; field of view [FOV] = 260 x 260 mm; slice thickness = 2.0 mm, interslice distance = 1.0 mm), yielding overcontiguous coronal slices with an in-plane spatial resolution of 1.016 x 1.016 mm/pixel. The FLAIR sequence was acquired in coronal orientation (TR/TE/inversion time [T1]/NEX = 11000/140/2600/2; matrix size = 256 x 256; FOV = 230 x 230 mm; slice thickness = 4.0 mm with no gap between slices) with in-plane spatial resolution of 0.898 x 0.898 mm/pixel. The total time of each subject’s scanning session was approximately 20 minutes. MRI scans were transferred to an independent Windows NT workstation and analyzed using the software packages ANALYZE (Mayo Foundation, Rochester, MI) and SPM99 (Cognitive Neurology Group, National Hospital for Nervous Diseases, London, U.K.).
Demographic, Health, and Lifestyle Variables
Participants were asked a series of questions about educational activities and these were used to calculate years of education. Depression and anxiety were assessed with the Goldberg depression and anxiety inventory (26). Cognitive impairment was assessed using the Mini-Mental State Examination (MMSE (27)). Histories of stroke, diabetes, hypertension and head injury were coded as binary variables. Current smoking status was recorded as current smoker, exsmoker, or never smoker. We have previously found that hormone replacement therapy is not associated with brain structure in our sample so we did not include this variable in our analyses (28).
Alcohol Consumption
Recent alcohol consumption was measured using three items of the World Health Organization Alcohol Use Disorders Identification Test (AUDIT (29)). Weekly consumption was estimated from frequency of alcohol intake and the number of standard drinks (10 g alcohol) consumed on typical drinking days following quantity-frequency assessment procedures (30,31). Quantity–frequency methods of estimating consumption have been found to correlate closely with daily diary records of alcohol intake and to have good test–retest reliability (32). A scale variable representing estimated number of drinks consumed per week was used in regression analyses. Highest level of past drinking was also estimated from two additional quantity–frequency items referring to drinking at the "highest level over a period of 3 months or longer." Responses to these items were used to classify participants who had previously consumed alcohol at hazardous or harmful levels (i.e., over 28 and 14 standard drinks per week for females and males, respectively) as defined by National Health and Medical Research Council guidelines (33). Highest level of past drinking was coded as 1 = hazardous/harmful or 0 = other.
Intracranial Volume and Tissue Segmentation
First, a customized T1 template was created by spatially normalizing, smoothing with an 8-mm full-width half-maximum kernel, and averaging the included image data. Second, the structural T1 images were spatially normalized into the new template and then segmented into gray matter, white matter, and CSF partitions. Each partition was then averaged and smoothed to create the study specific gray matter, white matter, and CSF templates. Third, raw structural images were segmented into gray matter, white matter, and CSF partitions from which the total volumes of gray matter, white matter, and CSF and intracranial volume, which was the sum of these three components, were computed for each participant. As measures of individual volumes of brain tissue compartments correcting for differences in head size, the ratios of total gray matter volume, white matter volume, and CSF volume to total intracranial brain volume were calculated for each participant. The computation for the volume of CSF was found to be difficult as a result of the fact that there was no clear boundary between CSF and skull in the T1-weighted scans. As a result of this, CSF could often be overestimated. This problem was partially resolved in our case by using the brain mask provided with the SPM2 package to remove the "out-of-boundary" voxels that would otherwise have been regarded as CSF. The brain mask defined in the "standard" space was then inversely mapped back to each individual CSF segmentation, which was defined in their "native" space, and the voxels outside the mask were then removed. Inverse mapping of brain masks was performed in SPM2. Subarachnoid CSF was calculated as the total CSF volume minus the ventricular volume. Ventricular volume was calculated by manual tracing using ANALYZE software.
White Matter Hyperintensities
WMHs were identified on FLAIR sequences and coregistered with a T1 image of the same subject. They were normalized spatially in Talairach space so that WMH could be identified and then localized. Falsely classified WMH were identified through visual inspection of WMH maps and removed. Using a standard atlas (34), we traced anatomic regions (deep white matter—frontal, parietal, temporal, and occipital; periventricular white matter—anterior cap, posterior cap, and periventricular body) on the standard single brain included in SPM99 software. Using these brain region of interest masks, the WMH volumes, number, location, and size were calculated automatically by implementing FSL (Image Analysis Group, FMRIB, Oxford University, U.K.: http://www.fmrib.ox.ac.uk/fsl/) in the script language PERL. Because both linear and nonlinear transforms were applied onto each individual MRI, the WMH thus measured did not equate exactly to absolute volume. The relative value, i.e., the "density" (ratio of WMH against white matter expressed as a percentage) of the WMH of each individual is used. The details of the method, which has excellent reliability and good validity, are described elsewhere (35).
Hippocampus and Amygdala
The volumes of brain anatomic regions were determined by two researchers (J.M., C.M.) manually outlining the periphery of the regions of interest on the coronal T1-weighted image. Outlining of the structures always proceeded from anterior to posterior, and the amygdala was traced first. Amygdala tracing began a maximum number of four slices anterior to the slice where the anterior tip of the temporal horn was visible and was traced according to the protocol outlined by Watson and colleagues (36). The hippocampus included the dentate gyrus, the hippocampus proper, and the subicular complex. The rostral end was the anatomic beginning of the head when it first appeared below the amygdala. The caudal end was taken as the section on which the crux of the fornix departed from the lateral wall of the lateral ventricles and a path of CSF was clearly seen. Data were normalized within person by dividing individual subvolumes by intracranial volume. The average interrater coefficient of variation calculated on 110 brains for amygdala and hippocampus was 0.035.
Corpus Callosum
The corpus callosum was traced on the midline sagittal slice which was chosen using anatomic landmarks in an hierarchical order as suggested by Talairach and Tournoux (37). These landmarks were, first, no or only minimal white matter in the cortical mantle surrounding the corpus callosum; second, the interthalamic adhesion; and third, the transparent septum and the visibility of the aqueduct of Sylvius. Intrarater reliability was tested in 15 randomly selected brains in which the corpus callosum area was traced twice by each researcher at different dates; the intraclass coefficients were 0.997 and 0.977, respectively. Interrater reliability was tested in 54 randomly selected brains; the intraclass correlation was 0.956.
Visual Atrophy Ratings
Scans were also visually rated by two independent clinicians experienced in examining MRI scans on a scale adapted from Victoroff (38) for the present study. Frontal, midtemporal, and midparietal atrophy was rated as either nil, mild, moderate, or severe atrophy using visual reference standards. The interrater reliability (weighted kappa) for these ratings, established on 57 cases, was 0.78 for frontal atrophy, 0.72 for midtemporal atrophy, and 0.72 for midparietal atrophy.
Statistical Analysis
Education, health, and brain measures were compared between males and females using t-tests for continuous variables and Pearson 
2 tests for categorical variables.
The unadjusted relationships between alcohol consumption (number of drinks per week) and brain measures for males and females were examined using a series of regression analyses. Nonlinearity in the relationships between alcohol consumption and brain variables was modeled by including polynomials in the regression analyses. Regression analyses of the brain measures were then further adjusted for years of education, comorbid conditions (hypertension, diabetes, stroke, head injury), depression, anxiety, MMSE, smoking status, and past drinking. Two male and one female participant reported very high number of drinks per week compared with the rest of the participants in their groups. To reduce the effect of these potential outliers, the two extreme values for males were replaced by one and two (for the more extreme of the two) units above the next highest number of drinks per week. Similarly, the extreme value for the female was replaced by one unit above the next higher number of drinks per week reported for that group. The relationships between alcohol consumption and atrophy ratings were tested using a series of ordinal logistic regressions conducted separately for males and females. All analyses were conducted using SPSS version 12.0.1 and significance level was set to 0.05.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
|
Table 2 displays unadjusted and adjusted relationships between weekly alcohol consumption with brain variables for males and females. For males, significant linear associations were found between drinks per week and ventricular volume, gray matter, white matter, and hippocampal volume left, respectively. After adjusting for years of education, depression, anxiety, MMSE, hypertension, diabetes, stroke, head injury, smoking status, and past drinking, the relationship between alcohol consumption and ventricular volume, gray matter, and white matter remained significant. Significant confounding variables were hypertension (standardized ß = 0.151, p = .035) for ventricular volume, MMSE (standardized ß = 0.161, p = .026) for gray matter, and diabetes (standardized ß = –0.206, p = .003) for white matter.
|
For females, significant unadjusted quadratic associations were found between drinks per week and CSF volume and subarachnoid CSF and white matter. These associations remained significant after adjusting for covariates. Stroke (standardized ß = 0.172, p = .027) was also associated with CSF volume. Figure 1 shows the relationships between drinks per week with ventricular volume, CSF volume (subarachnoid CSF is not shown because it is so highly correlated with CSF), gray matter, and white matter for males and females.
|
Visual Ratings of Cortical Atrophy
Further evaluation of the association between alcohol consumption and brain atrophy was conducted by analysis of visual atrophy ratings. For females, no significant association was found between alcohol consumption and atrophy ratings. For males, alcohol consumption was significantly associated with frontal lobe atrophy (ß = 0.056, Wald statistic = 13.85, p = .000) and midtemporal lobe atrophy (ß = 0.044, Wald statistic = 9.25, p = .002) with participants consuming higher number of drinks more likely to receive poorer ratings.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Evidence for nonlinear relationships between alcohol and indicators of atrophy such as increased CSF and reduced white matter were observed in women. This is consistent with the view that low levels of alcohol consumption are associated with less brain atrophy and may even be neuroprotective with animal studies showing that ethanol can enhance cell proliferation in the hippocampus (40). Alternatively, low levels of alcohol may have benefits through reduction in cardiovascular risk factors that also influence brain aging. Our results clearly show that high levels of alcohol consumption are associated with increasing rates of brain atrophy; however, the nonlinearity in our findings must be treated with caution given the limited sample size, the nature of our consumption measure, and the small numbers of women in the hazardous/harmful drinking category.
The results also suggest that alcohol consumption is detrimental for women at lower levels of consumption than for men. This is consistent with previous studies of alcoholic men and women (6,19). Studies have shown that women achieve higher blood levels of alcohol after consuming the same amount as men, possibly as a result of lower levels of overall body water in females and lower levels of gut enzymes that metabolize alcohol in women (41). The nonlinearity present in women in this study is more consistent with the threshold view supported by some animal studies (6,42) than with a dose–response relationship between alcohol and brain atrophy, which is suggested by the findings in males.
It has been suggested the most likely explanation for the loss of white matter associated with alcohol consumption is demyelination (43) or with changes in axonal integrity (4). One longitudinal study of alcoholic men also suggested a dose–response relationship between alcohol consumption and loss of gray matter (44), although our results did not support a loss of gray matter associated with alcohol intake in men or women.
Consistent with previous studies we also found greater cortical atrophy in the midfrontal and midtemporal lobes in heavier male drinkers. Our failure to find any link between alcohol consumption and corpus callosum size differs from previous studies (19,20,45). This may be the result of selection effects operating in case–control studies or that these effects are only present in more clinically severe cases of alcoholism. Our finding of smaller hippocampal volumes in males being associated with higher levels of alcohol consumption did not remain after adjusting for other demographic and health factors. It is possible that in a larger or older sample, or in a sample with more men drinking at hazardous levels, this effect would have been stronger because it is consistent with brain damage associated with long-term heavy drinking (46). The fact that this effect was only evident in males is likely the result of the larger proportion of males classified as hazardous or harmful drinkers.
Our results are consistent with one previous study in not finding an association between WMH and alcohol consumption (8). The study that did find an association had a larger and older sample than the present one. It is therefore possible that an association between WMH and alcohol consumption will emerge in the follow up of the present study when the sample is older.
Although the narrow age range enables examination of the associations free from confounding with age differences, it means that our results are restricted to an age group in which there is less variability in WMH and atrophic changes, thus reducing the possible strength of associations that may be found in a sample with a wider age range or a sample that is older. The present study was also limited by lack of information on the type of alcohol consumed, and it is possible that some of the statistically significant findings were the result of chance as a result of the large number of statistical tests. The strengths of the present study include random sampling of the population; the inclusion of a very large sample of manually traced hippocampi, amygdalae, and corpus callosa; automation of WMH ratings; and detailed information on alcohol consumption. Longitudinal follow up of this sample will allow for examination of how alcohol consumption affects further brain atrophy in this sample and whether alcohol-related atrophy affects brain reserve in relation to dementia risk (47).
We thank Helen Christensen, Florian Wertenauer, National Capital Diagnostic Imaging Group, Patricia Jacomb, Karen Maxwell, June Cullen, and the Neuroimaging Group, NPI, Prince of Wales Hospital.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Received for publication December 13, 2005; revision received May 30, 2006.
DOI:10.1097/01.psy.0000237779.56500.af
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. A. Paul, R. Au, L. Fredman, J. M. Massaro, S. Seshadri, C. DeCarli, and P. A. Wolf Association of Alcohol Consumption With Brain Volume in the Framingham Study Arch Neurol, October 1, 2008; 65(10): 1363 - 1367. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
