Untitled

ALCOHOLISM: CLINICAL AND EXPERIMENTAL RESEARCH Hippocampus Volume Loss Due to Chronic Heavy Drinking Thomas P. Beresford, David B. Arciniegas, Julie Alfers, Lori Clapp, Brandon Martin, Yiping Du Dengfeng Liu, Dinggang Shen, and Christos Davatzikos Objective: No clear consensus exists regarding the effect of sustained, heavy drinking on hippo- campal volume. Our prior work hypothesized significantly lowered total hippocampus volumes inheavy chronically drinking alcohol-dependent (AD) subjects compared with light-drinking nonde-pendent control subjects matched for age and gender.
Method: Using a series of applicable exclusion criteria culled from previous published studies, we measured hippocampal volumes from MRI scan data acquired on a 3T scanner and subjected thosedata to automated volume analysis blind to the drinking history.
Results: Comparison with AD test (n 5 8) and non-AD control (n 5 8) subjects found significant lessening in total ( p 5 0.020) and left ( p 5 0.010) hippocampal volumes with a near-significant differ-ence on the right ( p 5 0.051). Linear regression demonstrated that neither total brain volume norintracranial volume affected the hippocampus measures.
Conclusions: These data support the view that heavy drinking exerts a unique and selectively injurious effect on the hippocampus. Further study in larger samples must verify this in a search forpossible mechanisms of injury.
Key Words: Hippocampus, Alcohol Drinking, Volume Loss, MRI Scan.
THE EFFECT OF sustained, heavy drinking on convenience samples of heavy drinking and control sub- hippocampal volume is a subject of continued contro- jects (Beresford et al., 1999) led us to hypothesize that the versy. Although some reports in the literature point to a mean total hippocampus volume (THV) in the AD sub- volumetric reduction in alcohol-dependent (AD) patients jects would be significantly smaller than the mean THV in (Bleich et al., 2003a; Sullivan et al., 1995), others are less con- the non-AD control subjects. Our specific aim was to rep- clusive (Agartz et al., 1999). Some have argued that lack of licate the findings of the prior study but in prospectively precise sample definition has generated this confusion, citing gathered sample groups utilizing more stringent exclusion inclusion of subjects with histories of withdrawal seizures (Sullivan et al., 1996), for example. Others have disputed suchclaims (Bleich et al., 2003b).
With development project funding, we began a prospect- ive study of the hippocampal volume in chronically This project received prior approvals from our university institu- tional review board (IRB) as well as from the Research and drinking AD subjects. To assess volume comparisons at Development Committee of our Department of Veterans Affairs baseline, we analyzed new data from heavy-drinking AD (DVA) facility where the study was conducted. All subjects were test and light-drinking control cases. Early data from voluntary and signed preapproved consent-for-study documents,consistent with IRB policy.
Medicine, Denver, Colorado (TPB); the Department of Psychiatry, Both test (AD) and control subjects were adult male veterans who University of Colorado School of Medicine, Denver, Colorado (TPB, were eligible for care in the DVA system and were recruited in DBA, JA, LC, BM, YD); the Department of Neurology, University of response to posted flyers advertising the study. Our research design Colorado School of Medicine, Denver, Colorado (DBA); the Depart- matched AD test and non-AD control subjects for age, gender, and ment of Radiology, University of Colorado School of Medicine, Denver, ethnicity. Alcohol-dependent heavy drinkers qualified for study if Colorado (YD); and the Department of Radiology, University of Penn- sylvania, Philadelphia, Pennsylvania (DL, DS, CD).
Received for publication May 12, 2006; accepted July 28, 2006.
(1) Chronic heavy drinking: drank 5 or more standard drinks daily This work was supported by Grant R21-AA14010 from the National for at least 3 days weekly, and 3 weeks monthly for at least Institute on Alcohol Abuse and Alcoholism.
9 months of the previous year, established through Time Line Reprint requests: Thomas P. Beresford, MD, Denver VA Medical Follow Back interview (Sobell et al., 1979); Center (116), 1055 Clermont Street, Denver, CO 80220; Fax: 303-315- (2) Recent heavy drinking: consumed 5 or more standard drinks daily on at least 3 days weekly for the past 30 days, established Copyright r 2006 by the Research Society on Alcoholism.
No claim to original US government works.
(3) AD diagnosis: fulfilled DSM-IV criteria for AD as established through the Structured Clinical Interview for DSM-IV Axis I Alcohol Clin Exp Res, Vol 30, No 11, 2006: pp 1866–1870 Disorders (Kessler et al., 2004; Peters et al., 1998; Sbrana et al., Pennsylvania who analyzed the MRI scan data blind to the subjects’ study group membership. The steps in image analysis included (1) Similarly, non-AD light drinking comparison subjects presented removal of extracranial tissues (skull-stripping); (2) tissue segmenta- tion into gray matter (GM), white matter (WM), and cerebrospinalfluid (CSF); and (3) elastically warping a labeled atlas to all individ- (1) o2 standard drinks daily for no more than 3 days weekly, 4 ual subjects, to label and measure automatically the regions of weeks monthly for 9 months or less during the previous year and interest in the brains. These steps are described briefly.
(2) Drinking less than 2 standard drinks daily no more than 3 days (1) Skull stripping: A seed-based region growing procedure was (3) None fulfilled DSM-IV criteria, either present or lifetime, for applied first, which separates the brain parenchyma from extra- cranial material (Goldszal et al., 1998). Manual editing was thenperformed on a slice-by-slice basis, which also removes the cere- Candidates were excluded from study for SCID-verified psychi- bellum. Comparison with original, unstripped scans at manual atric illness: schizophrenia, major depressive disorder, bipolar editing assures scan data quality for the next step. The interop- disorder, posttraumatic stress disorder, or poly-substance depend- erator reliability test revealed nonsignificant differences in the ence (including concurrent antisocial personality disorder). Systemic manual editing between the 2 trained operators. For 14 image physical illness excluded those with any liver disease history, biliru- sets evaluated, the mean within-subject difference between raters bin above 1.2 mg/mL, ALT or AST above 200 U/L, Alcohol was À0.02 Æ 1.37% for white matter and 0.46 Æ 0.88% for gray Amnestic Syndrome history, HIV seropositivity, history of head matter. Correlations were greater than 0.99 for both measures.
injury resulting in loss of consciousness, seizure disorder history Finally, paired t-test comparisons yielded no significant differ- including those caused by ethanol withdrawal, blood evidence of folate or vitamin B-12 deficiency, dementia of any type, history of (2) Tissue segmentation: The SPGR data are segmented into GM, endocrine dysfunction (including Addison’s disease, Cushing’s WM, and CSF (Goldszal et al., 1998). An automated segmenta- disease, or exogenous steroid use within the past 5 years), and any tion algorithm (Segal et al., 1995) based on k-means clustering history of genetically based reactions to alcohol use (Asian ancestry and Markov random fields, which has been validated extensively with a history of the ethanol flush response). Alcohol-dependent (Davatzikos and Resnick, 1998; Goldszal et al., 1998), is used at subjects were excluded if withdrawal symptoms at the time of study this step. This method also applies correction for magnetic field (3) Automated measurement of brain structures: A labeled atlas is transformed spatially into spatial coregistration with each In this study, heavy-drinking subjects were required to provide a tissue-segmented individual brain scan via an elastic warping negative breath ethanol test administered by the study staff on the algorithm (Davatzikos et al., 2001a, 2001b; Shen and Davatzi- day of consent, the first day of the study protocol and again on the kos, 2002, 2003), referred to as Hierarchical Attribute Matching day of the first monitored disulfiram ingestion, as well as previous to Mechanism for Elastic Registration (HAMMER), thus obtain- each subsequent witnessed disulfiram administration. This was to ing the automatic labeling of brain structures in each subject assure (1) informed consent for study entry, that is, the absence of brain. By calculating the total volume in each structure with inebriation and (2) subject safety in disulfiram administration. The identical label, the volumetric measurement for each brain struc- MRI scan was performed within 3 days of the last reported alcohol ture in each subject can be obtained. To achieve this, we first use, within 4 days of the consent; breath alcohol testing was not adopted a finely parcellated brain image as a labeled atlas devel- required on the day of MRI scan. No subjects were scanned while oped by Kabani and colleagues at the Montreal Neurological obviously inebriated and none were observed in a disulfiram-ethanol Institute (Kabani et al., 1998), including the hippocampal reaction on the scan day. The 3-day limit for inclusion was designed regions of interest (ROIs). Left and right hippocampi have sep- to assure MRI study very early in the course of any structural healing arate labels in this atlas. We then used a HAMMER registration processes that might have begun with abstinence. All of the subjects algorithm (Shen and Davatzikos, 2002, 2003) to warp this atlas were ambulatory volunteers and none were inpatients at the time of to each individual subject’s image, and thus obtain the auto- study entry. Control subjects did not receive disulfiram, but reported matic labeling of the ROIs in each subject and further obtain negative alcohol use histories at the time of study entry (TLFB) and their volumetric measurements by computing the volume in each offered no evidence of intoxication.
labeled ROI. When computing the volume in each labeled ROI,all voxels, whether they are gray matter voxels or white matter voxels, were counted. Figure 1 shows the automatically labeledhippocampus in representative control and AD subjects. The All entered subjects completed a baseline 3T-MRI brain scan.
accuracy of the HAMMER registration algorithm has been Scan data were collected through whole brain-volume acquisition extensively validated by both real data and simulated data (Shen using a 3D inversion-recovery spoiled grass (IR-SPGR) pulse sequence in the coronal plane with an image matrix 256Â192Â124 on a 3T-MRI scanner (General Electric Company,Milwaukee, WI). The image resolution was 0.94Â0.94Â1.7 mm3, with a lower resolution along the anterior/posterior direction. The Scan-derived volumetric measurement data acquired in blinded inversion time of 450 ms was selected to optimize the gray/white fashion were analyzed for test or control group membership. Differ- matter contrast. The data acquisition time for the 3D volume was 13 ences in AD test versus non-AD control group means in total (combined), right, and left hippocampal mean volumes, respec-tively, were compared and assessed for statistical significance using Student’s t-test of the means for samples with differing variances.
Hippocampal, total brain, and intracranial volumes (ICVs) were Significance was set at 0.05 in a one-tailed test as indicated by the a derived using an automated segmentation process on the 3T-MRI priori prediction of the direction of mean values. We conducted a images of the brain (Shen and Davatzikos, 2002). These were secondary analysis using multiple linear regressions to include assessed by an imaging analysis research group at the University of potential confounding variables. These included total brain volume Fig. 1. Labeled hippocampus in representative subjects: control subject (left) and heavy-drinking subject (right).
(TBV) and ICV measures to assess the possible effects of individual criteria for a substance-related diagnosis: past alcohol abuse (the variations in brain and calvarium volumes. In this analysis, TBV is abuse occurred over 30 years before study entry).
the sum of all white matter, gray matter, and ventricular CSF, whileICV is the total volume inside the skull. If no confounding variableswere identified at the 0.1 level of probability, they were removed from analysis. If no variables remained, the model was reduced to a simple t-test of the means. As a convergent analysis, we conductedpartial correlation coefficient (PCC) tests between THV and TBV As shown in Table 1, the total (p 5 0.02) and left (p 5 0.01) mean volumes were significantly smaller in theAD heavy drinkers than in the light-drinking, non-AD controls. The right hippocampus mean difference narrowlymissed significance (p From a total of 54 screened cases (35 test and 19 control), we accrued a sample of 18 matched subjects who proceeded to MRI ed independently, mean measures of TBV and ICV each scan: 9 AD test and 9 non-AD control. The most common reasons differed significantly between the 2 groups (Student’s t-test, for failing study qualification included stopping drinking more than 2 tails). The mean TBV was 917.4 Æ 98.1 mL in the alcohol 3 days before study, medical illness exclusion, and presence of Axis I/ group versus 1,040.5 Æ 74.3 mL in the control group II disorders mentioned above. Two subjects, 1 from each group, were removed from final data analysis because of potentially significant 5 0.014). The mean ICV was 1,024.4 Æ 104.1 mL among anatomic abnormalities on the MRI scans that were not regarded as the drinkers versus 1,162.1 Æ 96.9 mL among the control related to alcohol use and that indicated clinical referral (evidence of previous anoxic injury and congenital malformation, respectively).
Both subjects were Caucasians; dropping them did not affect eitherage or ethnic distribution. Data were analyzed from the remaining Because the data sufficiently fit a normal distribution and did not require any transformation, we proceeded to multiple linear regression analysis. Drinking status was Owing to the matching procedure, the mean subject age was forced into the model during backward stepwise regression equivalent between groups: AD-test group 47.25 Æ 10.71 years, to assess our primary scientific question at all times. To non-AD control group 47.75 Æ 10.78 years. There were 7 Cau- evaluate whether hippocampus volumes were related to casians and 1 African American in each group. Within the brain or calvarium size, TBV and ICV measures were heavy-drinking group, the subjects’ reported average number of tested in the regression model as covariates. In the total drinks in the 30 days before study entry was 392 Æ 259 stand-ard drinks; however, the average alone is somewhat misleading. Of multivariate model for each side of the hippocampi, the 8 heavy drinkers, 5 drank daily and 3 drank during binges that neither covariate contributed significantly (right hippo- lasted 3 to 4 days weekly. For perspective, all drank an average of campus, p 5 0.42, p 5 0.65, respectively; left hippocampus, 16 Æ 7 standard drinks on those days when they drank.
By contrast, the control subjects reported far less alcohol exposure Table 1. Mean Hippocampal Volumes (mean mL Æ SD) in the prior 30 days. The 8 of them reported a total intake of24 standard drinks for the previous 30 days, an average of 3.0 Æ3.3 standard drinks each for the entire month, or about 0.75 Æ 0.8 standard drinks per person weekly. The range was 0 to 8 drinks during the month. Of the 8 test subjects, 1 subject met SCID criteria for current cannabis abuse but not dependence; 1 met criteria for current stimulant abuse but not dependence; and 1 met criteria for current cocaine dependence. None of the 8 control subjects met cri-teria for current dependence or abuse of any substance; only 1 met p 5 0.25, p 5 0.46, respectively). When the total hippo- this sample includes no female subjects and some reports campal volume was considered, neither TBV nor ICV suggest that gender may be a confounding factor in any significantly contributed to the model (p 5 0.31, p 5 0.53, comparison with male and female AD drinkers (Pfeffer- respectively) when drinking status was included. With baum et al., 2001; Gianoulakis et al., 2003). Other possible volume covariates showing no effect, we concluded that variables of interest were not recorded including handed- Student’s t-test was the appropriate statistic for assessment ness, socioeconomic status, or body size. Our previous of between-group hippocampus volumes measures.
research (Lucey et al., 1999), however, casts doubt as to Using the whole sample (n 5 16), we calculated the par- whether body size is a contributing variable in a design tial correlation coefficients (PCC) between THV and TBV, controlling for age and gender. Finally, this study did not as well as THV and ICV, controlling for drinking status as address any possible secondary molecular effects from either a heavy or a light drinker. We found no association high ethanol exposure (Bleich et al., 2003a).
between THV and either variable: for total brain volume, The test subjects in this study were seen in middle age PCC 5 0.281, p 5 0.31; for intracranial volume, PCC 5 after long, heavy-drinking careers. Although a recent report found that adolescents with alcohol use disorderswho are free of psychiatric comorbidities experience areduction in the left hippocampus (Nagel et al., 2005), our data offer no comment on heavy, sustained drinking at an The data presented here support the hypothesis that earlier age, for example, binge drinking in young adults chronic, heavy drinking of ethyl alcohol is associated with when the course of heavy drinking is comparatively early.
reduced THV and that observed volume reductions are Future directions suggested by this line of research likely independent of total brain and intracranial volumes.
include enlarging the sample beyond middle-aged, male, Although the data are derived from a relatively small DVA subjects in an effort to arrive at more generalizable sample, the subjects represent a group selected to be conclusions. Future replication study should include a free of variables previously reported as potentially con- wider sampling of heavy drinking men and a large sample founding volumetric MRI data—the exclusion criteria of heavy-drinking women. If the data continue to suggest listed above. As a result, we offer these results as clearly THV lessening, studies at earlier points in the drinking implicating an injurious role of chronic heavy ethanol use career—such as in heavy-drinking adolescents—as well as in specific minority groups would be indicated. The data For an added perspective, we construed the group mean observed here relate only a cross-sectional view of THV differences as a drug effect of ethanol. The calculated and raise the importance of recording the natural history effect difference in THV between the AD and control of hippocampus volume change, if any, over the course of groups yielded Cohen’s d 5 1.1. For left and right hippo- abstinence from ethanol. While the present report suggests campus volumes, d 5 1.3 and 0.9, respectively. Cohen’s injury to the hippocampus, injuries are often capable of statistic defines effect sizes as large equaling 0.6 to 0.8 or healing in a healthy environment. It is our hope, as well, to greater, medium 0.3 to 0.5, and small 0.0 to 0.2 (Cohen, explore this in serial, controlled MRI studies of active AD 1988). The large effect size in this case appears best attrib- uted to the difference in drinking status between these 2groups. As a comparison for discussion purposes, wecalculated this statistic from the reported effect data ofnaltrexone on days abstinent as reported in a recent multicenter trial (Anton et al., 2006). Those data yielded Agartz I, Momenan R, et al (1999) Hippocampal volume in patients with d 5 0.24, only a small effect. In the same study, medical alcohol dependence. Arch Gen Psych 56:356–363.
management without the study medication resulted in Anton RF, O’Malley SS, et al (2006) Combined pharmacotherapies and behavioral interventions for alcohol dependence: the COMBINE d 5 0.49, a medium effect. By contrast, the effect of study; a randomized controlled trial. JAMA 295:2003–2017.
ethanol that we observed in reducing THV appears to be Beresford T, Arciniegas D, et al (1999) Hippocampal to pituitary volume ratio: a specific measure of reciprocal neuroendocrine alterations in This study has several limitations that prevent general- alcohol dependence. J Studi Alcohol 60:586–588.
ization to all heavy, sustained users of alcohol. As Bleich S, Bandelow B, et al (2003a) Hyperhomocysteinemia as a new risk factor for brain shrinkage in patients with alcoholism. Neurosci Lett mentioned, the data presented from a small and highly select sample gathered to establish that hippocampus vol- Bleich S, Sperling W, et al (2003b) Lack of association between hippo- ume loss can be reliably observed. The small sample size campal volume reduction and first-onset alcohol withdrawal seizure.
may have to do with the smaller mean ICV that we observed in the drinking subjects; while previous research Cohen J (1988) Statistical Power Analysis for the Behavioral Sciences.
Lawrence Earlbaum Associates, Hillsdale, NJ.
strongly suggests lessened mean TBV in heavy drinkers, no Davatzikos C, Genc A, et al (2001a) Voxel-based morphometry using the reports in our awareness note lessened ICV as a general RAVENS maps: methods and validation using simulated longitudinal characteristic in a single gender sample. Concomitantly, Davatzikos C, Li HH, et al (2001b) Accuracy and sensitivity of detection Pfefferbaum A, Rosenbloom M, et al (2001) Sex differences in the of activation foci in the brain via statistical parametric mapping: a effects of alcohol on brain structure. Am J Psychiatry 158: study using a PET simulator. Neuroimage 13:176–184.
Davatzikos C, Resnick SM (1998) Sex differences in anatomic measures Sbrana A, Dell’Osso L, et al (2003) Acceptability, validity and reliability of interhemispheric connectivity: correlations with cognition in men of the Structured Clinical Interview for the Spectrum of Substance but not in women. Cereb Cortex 8:635–640.
Use (SCI-SUBS): a pilot study. Int J Methods Psychiatri Res 12: Gianoulakis C, Dai X, et al (2003) Effect of chronic alcohol consumption on the activity of the hypothalamic-pituitary-adrenal axis and pituit- Segal DL, Kabacoff RI, et al (1995) Update on the reliability of diagnosis ary beta-endorphin as a function of alcohol intake, age, and gender.
in older psychiatric outpatients using the structured clinical interview of DSM IIIR. J Clin Geropsychol 1:313–321.
Goldszal AF, Davatzikos C, et al (1998) An image processing protocol Shen D, Davatzikos C (2002) HAMMER: Hierarchical attribute match- for the analysis of MR images from an elderly population. J Comput ing mechanism for elastic registration. IEEE Trans Med Imaging Kabani N, MacDonald D, et al (1998) A 3D atlas of the human brain.
Shen DG, Davatzikos C (2003) Very high resolution morphometry using mass-preserving deformations and HAMMER elastic registration.
Kessler RC, Abelson J, et al (2004) Clinical calibration of DSM-IV diag- noses in the World Mental Health (WMH) version of the World Sobell LC, Maisto SA, et al (1979) Reliability of alcohol abusers’ self- Health Organization (WHO) Composite International Diagnostic reports of drinking behavior. Behav Res Therapy 17:157–160.
Interview (WMHCIDI). Int J Methods Psychiatric Res 13:122–139.
Sullivan EV, Marsh L, et al (1995) Anterior hippocampal volume deficits Lucey MR, Hill EM, et al (1999) The influences of age and gender on blood in nonamnesic, aging chronic alcoholics. Alcohol Clin Exp Res ethanol concentrations in healthy humans. J Stud Alcohol 60:103–110.
Nagel BJ, Schweinsburg AD, et al (2005) Reduced hippocampal volume Sullivan EV, Marsh L, et al (1996) Relationship between alcohol with- among adolescents with alcohol use disorders without psychiatric drawal seizures and temporal lobe white matter volume deficits.
comorbidity. Psychiatry Res 139:181–190.
Peters RH, Greenbaum PE, et al (1998) Prevalence of DSM-IV substance Ventura J, Liberman RP, et al (1998) Training and quality assurance abuse and dependence disorders among prison inmates. Am J Drug with the Structured Clinical Interview for DSM-IV (SCID-I/P).

Source: http://www.rad.upenn.edu/sbia/papers/Hippocampus%20Volume%20Loss%20Due%20to%20Chronic%20Heavy%20Drinking.pdf