Chronic alcohol consumption and its effect on nodes of frontocerebellar and limbic circuitry: Comparison of effects in France and the United States

Abstract

Alcohol use disorders present a significant public health problem in France and the United States (U.S.), but whether the untoward effect of alcohol on the brain results in similar damage in both countries remains unknown. Accordingly, we conducted a retrospective collaborative investigation between two French sites (Caen and Orsay) and a U.S. laboratory (SRI/Stanford University) with T1‐weighted, structural MRI data collected on a common imaging platform (1.5T, General Electric) on 288 normal controls (NC), 165 uncomplicated alcoholics (ALC), and 26 patients with alcoholic Korsakoff's syndrome (KS) diagnosed at all sites with a common interview instrument. Data from the two countries were pooled, then preprocessed and analyzed together at the U.S. site using atlas‐based parcellation. National differences indicated that thalamic volumes were smaller in ALC in France than the U.S. despite similar alcohol consumption levels in both countries. By contrast, volumes of the hippocampus, amygdala, and cerebellar vermis were smaller in KS in the U.S. than France. Estimated amount of alcohol consumed over a lifetime, duration of alcoholism, and length of sobriety were significant predictors of selective regional brain volumes in France and in the U.S. The common analysis of MRI data enabled identification of discrepancies in brain volume deficits in France and the U.S. that may reflect fundamental differences in the consequences of alcoholism on brain structure between the two countries, possibly related to genetic or environmental differences. Hum Brain Mapp 35:4635–4653, 2014. © 2014 Wiley Periodicals, Inc.

INTRODUCTION

In the U.S., the lifetime prevalence rate of alcohol abuse is 17.8% and of dependence 12.5%, with higher levels in men than women [Compton et al., 2007; Grant et al., 2004; Hasin et al., 2007]. Alcohol Use Disorders (AUDs) typically emerge during late adolescence and early adulthood (18–29 years) [Grant et al., 2009; Merikangas et al., 2010; Swendsen et al., 2012], when the brain is still undergoing maturation [Stiles and Jernigan, 2010]. In France, five million people (i.e., 8% of population) experience medical, psychological, or social difficulties attributable in part to alcohol consumption, and 2 million people (i.e., 3% of population) are alcohol dependent [Expertise Collective de l'Inserm, 2003]. More than half of all 17‐year‐old youth reported drinking more than five drinks per day at least once in the past month; 2.7% reported this consumption rate with a frequency of 10 times in the last month [Spilka et al., 2012]. Clearly, AUDs present a significant public health problem in France and the U.S., but what remains unknown is whether the untoward effect of alcohol on the brain results in similar damage in both countries. Indeed, variability in alcohol use pattern, genetic background, or nutritional status between France and the U.S. are among the many factors potentially contributing to between‐nation differences in brain volumes.

Common cognitive impairments in chronic alcoholics include diminished executive functioning affecting problem solving, abstraction, working memory, attention, response inhibition, and impulsivity [Loeber et al., 2009; Moriyama et al., 2002; Oscar‐Berman et al., 2004; Ratti et al., 2002, 1999; Uekermann et al., 2003], suggesting specific insult to frontal regions [Cummings, 1993; Fuster, 1999]. In uncomplicated alcoholics, however, volume deficits in nodes of the frontocerebellar circuitry can be better predictors of executive dysfunction than alterations in prefrontal regions alone [Chanraud et al., 2007; Sullivan, 2003; Zahr et al., 2010]. Frontocerebellar circuitry consists of parallel, multisynaptic, closed‐loop circuits [Habas et al., 2009; Kelly and Strick, 2003; Schmahmann, 2010], which include selective cerebellar lobules, thalamus, pons, and frontal cortical regions. An executive loop was identified as a specific supporting brain circuit of working memory functioning [Kelly and Strick, 2003]. Episodic memory is also impaired in chronic alcoholism and implicates compromise of limbic system components, including the hippocampus, mammillary bodies, thalamus, cingulate cortex, and amygdala, which together make up Papez's circuit. In chronic alcoholism without neurological complication, brain shrinkage has been demonstrated in the anterior hippocampus [Agartz et al., 1999; De Bellis et al., 2000; Sullivan et al., 1995], mammillary bodies [Sullivan et al., 1999], thalamus [Chanraud et al., 2007; Sullivan et al., 2003], cingulate cortex [Pitel et al., 2012], and amygdala [Makris et al., 2008].

Alcoholics are at special risk for thiamine deficiency because of poor eating habits, compromised thiamine absorption from the gastrointestinal tract, reduced thiamine storage, and phosphorylation [Lieber, 2003; Martin et al., 2003; Thomson, 2000; Thomson et al., 1987; Todd and Butterworth, 1999]; see review [Thomson et al., 2012]. The synergistic effects of excessive alcohol consumption and thiamine deficiency may lead to Korsakoff's Syndrome (KS), an amnesic syndrome occurring in alcoholics with untreated or under‐treated Wernicke's Encephalopathy (WE), which requires thiamine replacement therapy [Kopelman, 1995; Sechi and Serra, 2007; Thomson et al., 2012; Victor et al., 1959]. As in uncomplicated alcoholics, the brains of KS patients exhibit cerebral damage of nodes and connections of the frontocerebellar and limbic circuits [Pitel et al., 2009, 2012]. Alteration of the diencephalon was first proposed as a critical locus for explaining the severe and persistent amnesia characteristic of KS [for reviews Jung et al., 2012; Kril and Harper, 2012]. Neuroimaging studies conducted in France confirmed the shrinkage of the thalami and mammillary bodies [Pitel et al., 2009] and revealed hypometabolism of the thalami [Aupée et al., 2001]. Using a voxel‐based morphometric (VBM) approach, the same French research team showed graded effects of volume deficits in the mammillary bodies, medial thalamus, and left insula in alcoholics without versus those with amnesia [Pitel et al., 2012]. Thus, the French studies emphasized the role of the diencephalon in the development of alcohol‐related neurological complications resulting in KS.

In the U.S., a comparison of regional brain volumes in alcoholics with and without KS revealed graded effects of volumes deficits in the hippocampus, in addition to the mammillary bodies and thalamus [Sullivan and Pfefferbaum, 2009]. The shrinkage of the hippocampus in KS was equivalent to that observed in Alzheimer's disease [Sullivan and Marsh, 2003] and twice that reported in nonamnesic alcoholics [Sullivan et al., 1995]. Also revealed were selective relations between the severity of amnesia and extent of volume deficits in the hippocampus, but not in the mammillary bodies or temporal cortex, despite tissue volume deficits in the latter two structures [Sullivan and Marsh, 2003]. These findings indicate the relevance of hippocampal volume shrinkage as a neural substrate of KS amnesia. The apparent discrepancies observed between the diencephalic hypothesis of alcoholic‐KS amnesia of the French and the hippocampus basis of U.S. neuroimaging investigations in the development of KS may be explained by two kinds of factors: (1) the methods used for the preprocessing and analyses on the brain images (VBM approach in France versus region‐of‐interest approach in the U.S.) and (2) fundamental differences in the consequences of alcoholism on the brain structure between the two countries.

Brain structural manifestations of alcoholism are highly variable across individuals with alcohol use disorders. It remains unknown, however, why some alcoholics exhibit severe abnormalities of brain structure and function, whereas others with similar overall alcohol consumption levels appear to drink with impunity [Fein and Landman, 2005; Fein et al., 2006]. The brain insults attributed to chronic alcohol dependence might occur directly, with alcohol thereby acting as a neurotoxic substance, and could vary according to the alcohol use pattern (quantity, frequency, duration), or they might result indirectly from alcohol‐related liver disease, genetic background, environment, poor nutritional history, or impaired vitamin absorption [Thomson, 2000].

The first goal of the present study was to compare the effects of alcoholism on the principal components of the frontocerebellar and limbic circuits in large samples of subjects from France and the U.S., controlling for demographic variables notably age, intracranial volume, and sex, and for methodological variables, notably image preprocessing and analyses. Moreover, we aimed to characterize influences of specific determinants of alcoholic history (lifetime alcohol consumption, duration of alcoholism, and length of alcohol abstinence) that could contribute to the heterogeneity and gradient of alcohol‐related brain damage on nodes of the frontocerebellar and limbic circuits. To this end, we conducted a retrospective collaborative investigation across two French sites (Caen and Orsay) and a U.S. laboratory (SRI/Stanford University) on a combined subject population including uncomplicated alcoholics and alcoholics with KS, diagnosed at all sites with Diagnostic and Statistical Manual (DSM IIIR or DSM IV) criteria. All MRI data were T1‐weighted (SPGR) sequences acquired on 1.5T General Electric whole‐body magnet systems. Data from all sites were pooled, preprocessed, and analyzed together at the U.S. site using an atlas‐based parcellation approach [Rohlfing et al., 2010]. Region‐of‐interest (ROI) measurement of brain volumes has been recommended over the voxel‐based morphometry approach because of the formers' robustness to between‐center variability, including temporal‐linked changes in scanner hardware and software [Focke et al., 2011; Jovicich et al., 2009; Pfefferbaum et al., 2012a]. We expected that similar patterns of brain structural damage but different extents of regional volume shrinkage would be observed from uncomplicated alcoholics to Korsakoff patients and between patients from France and the U.S. In addition, we questioned whether the pattern of the relationships between drinking history variables and alcoholism‐related brain changes would differ between patients from France and the U.S., potentially accounting for national differences in brain damage from alcoholism.

METHODS

Participants

Our goal was to assemble the largest available set of volumetric studies of controls, alcoholics, and KS across the three sites. Thus, results from subsets of the assembled data set have previously appeared in other publications.

Caen

The Caen sample comprised 11 patients with an alcoholic Korsakoff's syndrome (KS), 23 nonamnesic patients with alcohol‐dependence (ALC), and 25 normal control participants (NC). The KS, ALC, and NC were included in previous studies [Le Berre et al., in press, 2012; Pitel et al., 2009, 2012]. Patients with KS were diagnosed with DSM‐IV criteria for persisting amnesic disorder [American Psychiatric Association, 1994]. A detailed neuropsychological examination confirmed that all patients with KS presented disproportionately severe episodic memory disorders compared with other cognitive functions (see [Pitel et al., 2009] for more details). All patients with KS had a history of heavy drinking as reported by family members and medical records, but their memory disorder precluded obtaining valid drinking histories directly from the participants. They were abstinent from alcohol and no longer presented any physiological symptoms of alcohol withdrawal. Patients with ALC were recruited by clinicians on the basis of the DSM‐IV criteria for alcohol dependence [American Psychiatric Association, 1994] while they were receiving treatment for alcohol dependence as inpatients at Caen University Hospital. NC subjects were social drinkers as defined by the National Institute on Alcohol Abuse and Alcoholism and were recruited via word of mouth. None of the participants was taking psychotropic medication or presented with psychiatric or medical histories (head injury, coma, epilepsy, depression, hepatic encephalopathy, or others) influencing the diagnosis. The participants gave their informed consent before entry to the study, which was conducted in line with the Declaration of Helsinki and was approved by the local ethics committee for human investigations.

Orsay

The Orsay sample comprised 25 nonamnesic patients with ALC and 28 NC. The ALC and NC were included in previous studies [Chanraud et al., 2009a, 2007,b; Chanraud‐Guillermo et al., 2009]. Patients with ALC were recruited by clinicians on the basis of the DSM‐IV criteria for alcohol dependence [American Psychiatric Association, 1994] while they were receiving clinical treatment as inpatients in addiction departments of the Paul Brousse and Emile Roux hospitals in the Paris area (Assistance Publique, Hôpitaux de Paris). Inclusion criteria required patients to have had fewer than three periods of withdrawal; or have been detoxified for at least 3 weeks and to be abstaining as assessed by biological norms (normal levels of gamma‐glutamyltransferase (GGT) and normal levels of carbohydrate‐deficient transferring (CDT)); and not to have been taken lorazepam or sedative medication for at least 7 days. Control participants were social drinkers as defined by the National Institute on Alcohol Abuse and Alcoholism and were recruited via word of mouth. For controls and alcoholics, exclusion criteria were: younger than age 25 or older than 65 years, left‐handedness, non‐fluency in French, presence of drug abuse (other than nicotine), anxiety or depressive disorders, and neurological, somatic, or other psychiatric symptoms, including a history of head injury with loss of consciousness, stroke, or other major brain abnormalities observed on MRI scans. The participants gave their written informed consent before their inclusion in the study, which was conducted in line with the Declaration of Helsinki and was approved by the Bicètre ethics committee for human investigations.

United States

The study groups included 15 patients with KS (10 included from a Veterans Administration (VA) Medical Center and 5 from SRI International), 117 nonamnesic patients with ALC (59 from VA and 58 from SRI), and 235 NC (188 from VA and 47 from SRI). Results from subsets of these 15 KS patients were reported in earlier studies [Fama et al., 2004a, 2006; Shear et al., 1996; Sullivan and Marsh, 2003; Sullivan and Pfefferbaum, 2009]. Subsets of the ALC and NC subjects were reported previously [Fama et al., 2004b, 2006; Pfefferbaum et al., 1992, 1995, 2004, 2007, 2012b; Rosenbloom et al., 2004, 2007; Sullivan et al., 2000]. The KS subjects were recruited from inpatient units of a VA Medical Center or from local substance abuse treatment programs. Each KS subject had an extensive history of alcoholism, which was verified through medical charts or family reports, and met criteria for DSM‐IV Alcohol‐Induced Persisting Amnestic Disorder [American Psychiatric Association, 1994]. Because of their amnesia, the KS patients were unable to provide a credible account of their drinking quantities.

The ALC subjects were recruited from a 28‐day inpatient alcohol program in a VA Medical Center or from local substance abuse treatment programs and met Research Diagnostic Criteria (RDC) for alcoholism [Spitzer et al., 1975] or DSM‐IV criteria for alcohol dependence [American Psychiatric Association, 1994], depending on recruitment year. The Research Diagnostic Criteria for alcohol dependence (RDC), a forerunner of DSM criteria, demonstrated a high level of agreement with the DSM III, even with stricter criteria for determining alcohol dependence than required by DSM [Leonard et al., 1984] The NC subjects were recruited from the local community. Screening for all participants (KS, ALC, and NC) included a structured psychiatric diagnostic interview and medical examination. Subjects were excluded for significant history of psychiatric or neurological disorder not related to their primary diagnosis, past or present alcohol or drug abuse or dependence in the NC group, substance dependence other than alcohol in the ALC and KS groups, or a serious medical condition. All non‐KS subjects underwent a semi‐structured interview [American Psychiatric Association, 1982; Pfefferbaum et al., 1988; Skinner and Sheu, 1982] to quantify lifetime alcohol consumption. All participants gave written informed consent to participate in the study; in a few cases of KS, written consent was also granted by a legal custodian. The study was approved by the Institutional Review Boards (IRB) of Stanford University School of Medicine and the Veterans Administration, and later study components were approved by IRBs of Stanford University School of Medicine and SRI International.

Demographic data for KS, ALC, and NC groups at each site [Caen, Orsay, the U.S. Veterans Administration Medical Center (U.S. VA), and U.S. SRI International (U.S. SRI)] are listed in Table 1. The three diagnostic groups differed in age; the sex‐ratio and age were significantly different by site (Table 1). The alcoholic groups differed significantly in duration of alcoholism, with U.S. SRI ALC patients having a longer duration of alcohol dependence than ALC patients from the U.S. VA and French sites. The ALC patients from Orsay and the U.S. VA had similar durations of alcoholism but longer durations than ALC patients from Caen. The ALC groups did not differ in lifetime alcohol consumption or length of alcohol abstinence prior to study participation.

Table 1.

Demographic data (mean ± SD) of the samples from Caen, Orsay, the U.S. VA, and the U.S. SRI for the three diagnoses: Healthy control subjects (NC), patients with alcoholism (ALC), and patients with Korsakoff's syndrome (KS)

	CAEN			ORSAY		U.S. VA			U.S. SRI			Effect of the site		Effect of the diagnosis
	NC	ALC	KS	NC	ALC	NC	ALC	KS	NC	ALC	KS	χ 2, F value	P	χ 2, F value	P
n	25	23	11	28	25	188	59	10	47	58	5	/	/	/	/
Sex (women/men)	14/11	3/20	4/7	0/28	0/25	72/116	36/23	2/8	15/32	14/44	2/3	37.99a	<.001* Orsay ≠ (Caen= U.S. VA = U.S. SRI)	0.52a	0.77
Age (yrs)	60.20 ± 4.86	43.39 ± 8.60	52.91 ± 10.10	44.36 ± 7.96	47.48 ± 7.25	50.51 ± 17.01	44.78 ± 11.70	68.00 ± 4.88	45.38 ± 15.33	46.19 ± 11.41	57.00 ± 15.89	3.51b	<.02* (Orsay = U.S. SRI) < (U.S. VA = Caen)	13.04b	<0.001* KS > NC > ALC
Lifetime alcohol consumption (kg)	/	1312.36 ± 1278.79	/	/	1046.09 ± 1109.96	/	820.66 ± 690.07	/	/	1183.28 ± 987.79	/	2.06b	.11	/	/
Duration of alcoholism (years)	/	15.13 ± 10.89	/	/	8.44 ± 7.26	/	16.95 ± 9.71	/	/	23.28 ± 11.34	/	13.32b	<.001* Orsay <(Caen = U.S. VA) < U.S. SRI	/	/
Alcohol abstinence prior to the study (days)	/	12.74 ± 7.76	/	/	321.92 ± 645.81	/	285.93 ± 467.18	/	/	315.24 ± 580.82	/	2.20b	.09	/	/

^aPearson χ ².

^bUnivariate ANOVA: F value.

*Significant at P < 0.05.

The Orsay sample had no KS patients; the U.S. SRI sample comprised only five KS patients. In addition, the Orsay sample included only men; the sex‐ratio also differed by sites with men and women. Therefore, in order to take into account the KS diagnosis and to consider sex as a covariate in the statistical analyses, controls, alcoholics and KS patients from Caen and Orsay as well as participants from the U.S. VA and the U.S. SRI were pooled for the comparison between France and the U.S. in the subsequent analyses.

Demographic data for the KS, ALC, and NC groups in each country (France vs. U.S.) are listed in Table 2. The three diagnostic groups differed in age, and the sex‐ratio was significantly different by site (Table 2). The French and U.S. alcoholic groups differed in duration of alcoholism, with U.S. ALC patients having a longer duration of alcohol dependence than ALC patients from the French sites (Table 2). The country‐specific ALC groups did not differ in lifetime alcohol consumption or length of alcohol abstinence prior to study participation.

Table 2.

Demographic data (mean ± SD) of the samples from France and the U.S. for the three diagnoses: Healthy control subjects (NC), patients with alcoholism (ALC), and patients with Korsakoff's syndrome (KS)

	FRANCE			U.S.			Effect of the site		Effect of the diagnosis
	NC	ALC	KS	NC	ALC	KS	χ 2, F, t value	P	χ 2, F value	P
n	53	48	11	235	117	15	/	/	/	/
Sex (women/men)	14/39	3/45	4/7	87/148	50/67	4/11	14.83a	<0.001*	0.52a	.77
Age (yrs)	51.83 ± 10.37	45.52 ± 8.10	52.91 ± 10.10	49.48 ± 16.78	45.48 ± 11.52	64.33 ± 10.78	0.27c	0.79	13.04b	<.001*KS>NC>ALC
Lifetime alcohol consumption (kg)	/	1173.68 ± 1188.55	/	/	1000.42 ± 866.43	/	1.04c	0.30	/	/
Duration of alcoholism (yrs)	/	11.65 ± 9.69	/	/	20.09 ± 10.98	/	−4.64c	<0.001*	/	/
Alcohol abstinence prior to the study (days)	/	173.77 ± 487.20	/	/	300.45 ± 524.52	/	−1.44c	0.15	/	/

^aPearson χ ².

^bUnivariate ANOVA: F value.

^cIndependent samples T‐test: t value.

*Significant at P < 0.05.

MRI Data Acquisition

Caen

All MRI data were acquired on the same scanner (1.5 T Signa Advantage Echo Speed; General Electric Healthcare, Milwaukee, WI) using the same parameters for all participants: axial SPoiled Gradient Recalled (SPGR) images, 128 slices, slice thickness = 1.5 mm, repetition time (TR) = 10.3 ms, and echo time (TE) = 2.1 ms.

Orsay

All subjects underwent volumetric MRI brain scanning on a 1.5 T Signa system (General Electric Healthcare, Milwaukee, WI) with axial T1‐weighted inversion recovery fast‐SPGR sequence with the following parameters: 124 slices locations, slice thickness = 1.3 mm, TR = 10 ms, and TE = 2 ms.

United States

For participants from the VA Medical Center (10 KS patients, 68 ALC, and 188 NC), MR data were acquired on a GE Signa 1.5 T MRI scanner as sagittal T1‐weighted SPGR images with the following parameters: 124 slices, slice thickness = 1.5 mm, TR = 24 ms, and TE = 5 ms.

All the other participants (5 KS patients, 58 ALC, and 47 NC) underwent a structural MRI examination at SRI International. MR data were collected on a GE 1.5 T scanner as coronal T1‐weighted SPGR images with the following parameters: 94 slices, slice thickness = 2 mm, TR = 26 ms, and TE = 5 ms.

MRI Data Preprocessing

The same MRI preprocessing and quantification were applied to data from all sites.

Image preprocessing

All structural images were first corrected for intensity bias by applying a second‐order polynomial multiplicative bias field computed via entropy minimization [Likar et al., 2001]. The SPGR images were each skull stripped using FSL's Brain Extraction Tool, BET [Smith, 2002].

Registration and atlas‐based parcellation

For each subject, the skull‐stripped SPGR image was registered to the SPGR channel of the SRI24 atlas [Rohlfing et al., 2010] (http://nitrc.org/projects/sri24) via nonrigid image registration [Rohlfing and Maurer, 2003]. We chose the SRI24 atlas over other available brain templates (e.g., MNI152) because of its ability to discern detailed anatomical structures, which can thus be unambiguously outlined directly in the atlas images without the need to access the images that were used to create the atlas itself.

Tissue segmentation

All bias‐corrected and skull‐stripped SPGR images were segmented into three tissue compartments (gray matter, white matter, CSF) using FSL's FAST tool [Zhang et al., 2001]. As tissue priors to both initialize and guide the classification, we used the tissue probability maps provided with the SRI24 atlas, reformatted into subject SPGR space via the same transformations described above.

Regions of interest (ROIs)

Because our focus was to compare the effects of chronic alcohol consumption on regional brain volumes within the frontocerebellar and limbic circuits and between France and the U.S., we focused our analyses on nine bilateral regions of cortex and allocortex (anterior/middle sections of the cingulate gyrus, posterior section of the cingulate gyrus, lateral frontal cortex, medial frontal cortex, superior cerebellum, and inferior cerebellum), subcortical structures (thalamus, hippocampus, amygdala), and two midline ROIs (cerebellar vermis, pons) [Pfefferbaum et al., 2010] (Fig. 1). Gray matter volume was computed for each cortical region, and tissue volume for each subcortical region, derived from the SRI24 atlas [Rohlfing et al., 2010].

Figure 1.

Frontocerebellar and limbic parcellated regions of interest (ROIs). Sample axial slices (top left, superior to bottom right, inferior) are from the SRI24 atlas displaying color‐coded ROIs used in the analyses. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Statistical Analyses

Z‐scores for each regional brain measure were calculated to adjust for normal variation in intracranial volume (ICV) and for age by a two‐step regression approach to adjust for age differences across diagnostic groups [Mathalon et al., 1993; Pfefferbaum et al., 1992]. Thus, control subjects from Caen, Orsay, and the U.S. were pooled for the computation of the z‐scores. For each subject, the standardized ICV‐ and age‐corrected z‐scores were expressed as deviation from the mean, where, by definition, the expected mean of the 288 controls was 0 and standard deviation was 1.

Because the sex ratio differed by sites, sex was entered as a covariate in the statistical analyses. The effect of alcoholism in France and the U.S. was compared with three‐diagnosis (KS vs. ALC vs. NS) × two‐site (France vs. U.S.) × two‐sex (men vs. women) MANOVAs on the six cortical/allocortical gray matter volumes, three subcortical structures volumes, and on two midline volumes. When an overall F‐test yielded significant effects for diagnosis, site, or diagnosis‐by‐sites interactions, separate three‐diagnosis (KS vs. ALC vs. NS) × two‐site (France vs. U.S.) ANOVAs were conducted. When necessary, follow‐up tests (Fisher's LSD) were carried out for between‐group comparisons.

Examination of the relationships between regional brain volumes in each ROI and alcohol history variables (i.e., lifetime alcohol consumption, duration of alcoholism, and length of sobriety) used Pearson's correlation analyses in the French and U.S. alcoholic groups separately. In addition, the length of sobriety data were log‐transformed because they were non‐normally distributed, and the correlations were recalculated on the log‐transformed data. Alcohol history correlations were not conducted in the KS given the lack of reliability of collected alcohol history variables due to amnesia of the KS patients. A probability level threshold of 0.05 was adopted for significance of all analyses.

RESULTS

Initial omnibus MANOVAs coupled with follow‐up paired comparisons were based on ICV‐ and age‐corrected z‐scores and tested for site and diagnosis differences. A description of two‐diagnosis (ALC vs. NS) × four‐site (Caen vs. Orsay vs. U.S. VA vs. U.S. SRI) MANOVAs is provided in the Supporting Information. These analyses indicated that there was no difference in regional brain volumes between subjects from Orsay and Caen for any of the ROIs.

Comparison of Regional Volumes in the Control, Alcoholic, and KS Groups (NC vs. ALC vs. KS) in France versus U.S.

Because the sex ratio was different by country, sex was entered as an additional factor in the MANOVA analyses.

Cortical and allocortical gray matter volumes

The three‐diagnosis × two‐country × two‐sex MANOVA conducted on the cortical and allocortical gray matter volumes revealed significant effects of diagnosis [F _(12,922) = 4.79; P < 0.001], site [F _(6,461) = 4.61; P < 0.001], and sex [F _(6,461) = 2.48; P < 0.05] but no country × diagnosis interaction [F _(12,922) = 0.74; P = 0.71], country × sex interaction [F _(6,461) = 0.22; P = 0.97], diagnosis × sex interaction [F _(12,922) = 0.84; P = 0.61], or country × diagnosis × sex interaction [F _(12,922) = 1.06; P = 0.39]. As this MANOVA analysis did not reveal a significant effect of country × diagnosis × sex interaction, sex was removed as a factor in the subsequent ANOVA analyses of cortical and allocortical regions, the results of which are presented in the respective figures.

The three‐diagnosis (KS vs. ALC vs. NC) × two‐country (France vs. U.S.) ANOVA carried out on the lateral frontal volumes showed significant effects of diagnosis and country but no interaction. Follow‐up t‐tests indicated the expected step‐wise effect, with the smallest volumes in KS, larger volumes in ALC, and the largest volumes in NC (Fig. 2A). In addition, the lateral frontal volumes were smaller in the subjects from France than in those from the U.S. Statistical results appear in Table 3.

Figure 2.

Scatterplots of cortical and allocortical gray matter volumes in controls (NC), alcoholics (ALC), and Korsakoff's Syndrome patients (KS) in France and the U.S. Means of each column of data are noted by group and by country. Blue dots = France (light blue = Orsay; dark blue = Caen); red dots = U.S. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Table 3.

ICV‐ and age‐corrected z‐scores (mean ± S.D.) of the samples from France and the U.S. for the three diagnoses: Healthy control subjects (NC), patients with alcoholism (ALC), and patients with Korsakoff's syndrome (KS)

	FRANCE			U.S.			Effect of the diagnosis		Effect of the site		Effect of the interaction
	NC	ALC	KS	NC	ALC	KS	F value	P	F value	P	F value	P
Cortical/allocortical gray matter volumes
Lateral frontal cortex	−0.38 ± 0.92	−1.01 ± 0.95	−1.52 ± 1.39	0.10 ± 0.98	−0.48 ± 1.25	−1.53 ± 1.04	25.58	<0.001* KS < ALC < NC	4.25	<0.05* France < U.S.	0.70	ns
Medial frontal cortex	−0.60 ± 0.85	−1.24 ± 0.94	−1.52 ± 1.21	0.14 ± 0.98	−0.65 ± 1.06	−1.44 ± 1.29	29.78	<.001* KS < ALC < NC	9.56	0.002* France < U.S.	1.29	ns
Superior cerebellum	0.43 ± 1.10	0.29 ± 0.86	0.16 ± 0.90	−0.10 ± 0.95	−0.03 ± 1.12	−0.88 ± 0.90	3.08	<.05* KS < (ALC=NC)	16.98	<0.001* U.S. < France	1.47	ns
Inferior cerebellum	0.24 ± 1.11	0.12 ± 1.04	0.33 ± 1.20	−0.05 ± 0.97	−0.07 ± 1.05	−0.80 ± 0.56	1.18	ns	12.05	<0.001* U.S. < France	2.32	ns
Anterior and middle cingulate gyri	−0.68 ± 0.91	−0.96 ± 0.88	−1.01 ± 1.02	0.16 ± 0.95	−0.26 ± 1.14	−0.32 ± 1.10	5.49	<0.005* (KS = ALC) < NC	23.78	<0.001* France < U.S.	0.21	ns
Posterior cingulate gyrus	−0.20 ± 1.11	0.12 ± 0.89	−0.49 ± 1.23	0.04 ± 0.97	−0.04 ± 1.02	−0.08 ± 1.22	1.32	ns	1.17	ns	1.89	ns
Subcortical structures volumes
Hippocampus	−0.24 ± 1.03	−0.26 ± 1.08	0.24 ± 1.24	0.05 ± 0.98	−0.29 ± 0.97	−1.47 ± 2.04	3.16	<0.05* KS < ALC < NC	9.11	0.003* France < U.S.	10.25	<.001* KS U.S. < KS France
Amygdala	0.01 ± 1.14	0.05 ± 1.10	−0.28 ± 1.08	−0.01 ± 0.97	−0.22 ± 1.05	−1.60 ± 2.04	8.63	<0.001* KS < (ALC=NC)	10.74	0.001* U.S. < France	4.19	<0.02* KS U.S. < KS France
Thalamus	−0.27 ± 1.11	−1.39 ± 1.30	−1.56 ± 1.50	0.06 ± 0.96	−0.39 ± 0.99	−1.52 ± 1.28	34.05	<0.001* KS < ALC < NC	8.17	<0.005* France < U.S.	4.71	<0.01* France NC < U.S. NC France ALC < U.S. ALC
Midline volumes
Vermis	0.06 ± 0.98	−0.12 ± 1.03	0.01 ± 0.68	−0.01 ± 1.00	−0.34 ± 1.02	−1.39 ± 0.65	6.72	0.001* KS < ALC < NC	13.98	<0.001* U.S. < France	4.98	<0.01* KS U.S. < KS France
Pons	0.32 ± 0.82	−0.02 ± 0.89	−0.40 ± 0.78	−0.08 ± 1.04	0.26 ± 1.24	−0.40 ± 1.00	2.86	ns	0.06	ns	4.06	<.02* U.S. NC < France NC

*Significant at P < 0.05.

The same analysis conducted on the volume of the medial frontal cortex revealed significant effects of diagnosis and country but no interaction. Again pair‐wise t‐tests indicated a step‐wise effect: KS < ALC < NC (Fig. 2B). Regardless of diagnosis, subjects from France had smaller age‐, head size‐, and sex‐controlled medial frontal volumes than those from the U.S.

The ANOVA analysis of superior cerebellum volumes revealed significant effects of diagnosis and country but no interaction. The KS group had smaller superior cerebellum volumes than the ALC or NC, which did not differ from each other (Fig. 2C). In addition, the superior cerebellum volumes were smaller in the subjects from the U.S. than in those from France.

The analysis carried out on the inferior cerebellum volumes showed a significant effect of country but no effect of diagnosis or interaction. Generally, U.S. subjects had smaller inferior cerebellar volumes than those from France (Fig. 2D).

The ANOVA analysis revealed significant effects of diagnosis and country but no interaction for volumes of the anterior and middle cingulate gyri. Both the alcoholic and KS groups had smaller anterior cingulate volumes than the control group (Fig. 2E). Moreover, the anterior cingulate gyrus was smaller in the subjects from France than the U.S.

There was neither a significant effect of diagnosis nor effect of country and no interaction for the volume of the posterior cingulate gyrus (Fig. 2F).

All statistical results for three‐diagnosis (KS vs. ALC vs. NC) × two‐country (France vs. U.S.) ANOVAs carried out on each cortical and allocortical gray matter volumes appear in Table 3.

Tissue (subcortical structures) volumes

The three‐diagnosis × two‐country × two‐sex MANOVA carried out on the tissue (subcortical structures) volumes showed significant effects of diagnosis [F _(6,928) = 7.49; P < 0.001], country [F _(3,464) = 4.26; P = 0.005], sex [F _(3,464) = 4.97; P = 0.002], country × diagnosis interaction [F _(6,928) = 2.96; P = 0.007], and diagnosis × sex interaction [F _(6,928) = 2.52; P = 0.02], but neither a country × sex interaction [F _(3,464) = 0.32; P = 0.81] nor a country × diagnosis × sex interaction [F _(6,928) = 0.71; P = 0.64]. As this MANOVA analysis did not reveal any significant country × diagnosis × sex interactions, sex was removed from the subsequent ANOVA analyses.

The three‐diagnosis × two‐country ANOVA conducted on the volume of the hippocampus revealed significant effects of diagnosis, country, and an interaction. Follow‐up t‐tests indicated the expected step‐wise effect of volumes: KS < ALC < NC (Fig. 3A). Overall, subjects from France had smaller hippocampal volumes than those from the U.S., but hippocampal volumes were smaller in the KS group from the U.S. than from France.

Figure 3.

Scatterplots of tissue (subcortical structures) volumes in controls (NC), alcoholics (ALC), and Korsakoff's Syndrome patients (KS) in France and the U.S. Means of each column of data are noted by group and by country. Blue dots = France (light blue = Orsay; dark blue = Caen); red dots = U.S. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

The same analysis conducted on the volume of the amygdala revealed significant effects of diagnosis, country, and an interaction. Follow‐up t‐tests indicated that the KS group had smaller amygdala than the alcoholic and control groups, who did not differ from each other (Fig. 3B). Overall, the amygdala volumes were smaller in the subjects from the U.S. than in those from France. Further, amygdala volumes were smaller in KS patients from the U.S. than in those from France.

The analysis carried out on the thalamus volumes showed significant effects of diagnosis, country, and an interaction. Post‐hoc t‐tests indicated a step‐wise effect: KS < ALC < NC (Fig. 3C). Overall, subjects from France had smaller thalamic volumes than those from the U.S. In addition, thalamic volumes were smaller in the control group from France than in the control group from the U.S. and in the alcoholic group from France than in the alcoholic group from the U.S.

All statistical results for three‐diagnosis (KS vs. ALC vs. NC) × two‐country (France vs. U.S.) ANOVAs carried out on each tissue (subcortical structures) volumes appear in Table 3.

Midline regions

The three‐diagnosis × two‐country × two‐sex MANOVA conducted on the two midline regions revealed significant effects of diagnosis [F _(4,930) = 3.38; P < 0.01], country [F _(2,465) = 6.11; P = 0.002], and country × diagnosis interaction [F _(4,930) = 2.43; P < 0.05] but no significant sex effect [F _(2,465) = 2.86; P = 0.06], country × sex interaction [F _(2,465) = 0.04; P = 0.96], diagnosis × sex interaction [F _(4,930) = 0.79; P = 0.53], or country × diagnosis × sex interaction [F _(4,930) = 1.72; P = 0.14]. As this MANOVA analysis did not reveal a significant effect of country × diagnosis × sex interaction, sex was removed as a factor in the subsequent ANOVA analyses.

The three‐diagnosis × two‐country ANOVA carried out on the vermis volumes showed significant effects of diagnosis, country, and an interaction. Post‐hoc t‐tests indicated a step‐wise effect: KS<ALC<NC (Fig. 4A). Overall, subjects from the U.S. had smaller vermis volumes than those from France, and the U.S. KS patients had smaller vermis volumes than those from France.

Figure 4.

Midline regions volumes in controls (NC), alcoholics (ALC), and Korsakoff's Syndrome patients (KS) in France versus the U.S. Means of each column of data are noted by group and by country. Blue dots = France (light blue = Orsay; dark blue = Caen); red dots = U.S. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

The same analysis conducted on pons volumes revealed a significant group‐by‐country interaction, but no simple effects of diagnosis or country. Post‐hoc t‐tests indicated that pons volumes were smaller in the control group from the U.S. than in the controls from France (Fig. 4B). This result could be the consequence of segmentation and parcellation limitations in delineating the pons from the T1‐weighted images as the anterior border is difficult to segment due to flow artifact from the basilar artery and the posterior border of the pons is particularly difficult to dissociate from the rest of the brainstem tissue.

All statistical results for three‐diagnosis (KS vs. ALC vs. NC) × two‐country (France vs. U.S.) ANOVAs carried out on each midline regions appear in Table 3.

Relationships Between Regional Brain Volumes and Alcohol History and Age in the Alcoholics From the Two Countries (France vs. U.S.)

In the French alcoholics without KS, significant relationships emerged: (1) smaller volumes of lateral frontal cortex, vermis, and pons correlated with greater lifetime alcohol consumption, and (2) larger volumes of lateral frontal cortex and thalamus correlated with longer sobriety (Tables 4 and 5 and Fig. 5a). In the U.S. alcoholics without KS, greater lifetime alcohol consumption and shorter duration of alcoholism were each related to smaller thalamic volumes (Tables 4 and 5 and Fig. 5b). No relationship with age was demonstrated in the alcoholics group from either country.

Table 4.

Relationships between regional brain volumes and alcohol history for limbic circuit in the French and U.S. alcoholic groups

Country	ROIs (bilateral)	Sobriety (days)a	Lifetime alcohol consumption (kg)	Duration of alcoholism (yrs)	Age (yrs)
France	Thalami	0.30*	−0.15	−0.19	−0.02
	Hippocampi	−0.08	0.12	0.19	0.18
	Amygdala	0.22	−0.01	0.01	0.29
	anterior/middle cingulate gyrus	0.27	−0.16	−0.13	0.22
	posterior cingulate gyrus	0.26	0.12	0.16	0.08
U.S.	Thalami	−0.01	−0.24 *	−0.31*	−0.17
	Hippocampi	0.08	0.26	0.42	0.19
	Amygdala	−0.02	0.09	0.32	0.13
	Anterior/middle cingulate gyrus	0.04	−0.18	−0.08	−0.05
	Posterior cingulate gyrus	−0.01	−0.03	0.06	0.14

^aLog−transformed variable.

*Significant correlation at P < 0.05.

Table 5.

Relationships between regional brain volumes and alcohol history for frontocerebellar circuit in the French and U.S. alcoholic groups

Country	ROIs (bilateral)	Sobriety (days)a	Lifetime alcohol consumption (kg)	Duration of alcoholism (yrs)	Age (yrs)
France	Thalami	0.30*	−0.15	−0.19	−0.02
	Vermis	−0.06	−0.30*	−0.22	−0.08
	Pons	0.13	−0.29*	−0.22	0.04
	Lateral frontal cortex	0.41*	−0.30*	−0.28	0.40
	Medial frontal cortex	0.28	0.10	−0.24	0.24
	Superior cerebellum	−0.09	−0.26	−0.25	−0.24
	Inferior cerebellum	−0.06	−0.15	0.01	0.02
U.S.	Thalami	−0.01	−0.24*	−0.31*	−0.17
	Vermis	0.04	0.13	−0.12	−0.05
	Pons	−0.01	0.13	0.14	0.06
	Lateral frontal cortex	−0.01	0.01	0.04	−0.11
	Medial frontal cortex	0.07	−0.11	−0.08	−0.04
	Superior cerebellum	0.13	0.20	0.09	0.01
	Inferior cerebellum	0.04	0.24	0.24	0.06

^aLog‐transformed variable.

*Significant correlation at P < 0.05.

Figure 5.

a. Scatterplots illustrating the relationships between regional brain volumes and alcohol history in the combined French alcoholic patients. b. Scatterplots illustrating the relationships between regional brain volumes and alcohol history in the U.S. alcoholic patients. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

DISCUSSION

Using a single, common quantification approach for processing and analysis of brain images from the two international sites enabled (1) determination of regional brain volume differences between the U.S. and France (Fig. 6), (2) confirmation of a universal effect of diagnostic, graded volume shrinkage from uncomplicated alcoholism to KS in specific brain regions (Fig. 7), and (3) recognition that volumes within the frontocerebellar and limbic circuitry were differentially affected by chronic excessive alcohol consumption depending on the country. Principal national differences indicated that thalamic volumes were smaller in non‐KS alcoholics in France than the U.S. despite similar alcohol consumption levels in both countries, whereas volumes of the hippocampus, amygdala, and cerebellar vermis were smaller in KS in the U.S. than France even after adjustment for variation due to age and intracranial volume.

Figure 6.

ROIs showing differences, regardless of diagnosis, between the samples from France and the U.S. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 7.

Frontocerebellar ROIs (green tones) and limbic ROIs (yellow tones) exhibiting graded volume deficits (ALC > KS). The color scales indicate the level of volume deficit by z‐score. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Regional Brain Volume Differences Between France and the U.S.

Macrostructural brain volume disparities were found between subjects irrespective of diagnosis from the two countries: larger lateral and medial frontal cortices, anterior/middle cingulate gyrus, thalamus, and hippocampus in the U.S. than in France, and larger superior cerebellum, inferior cerebellum, cerebellar vermis, and amygdala in France than in the U.S. These brain volume discrepancies may be attributable to variability in genetic polymorphisms underlying ethnic diversity or environmental differences in nutrition related to cultural eating habits. Compared with the French sample, the U.S. sample is characterized by a greater ethnic diversity with following ethnic distribution: 69% White, 13% African American, 7% Asian, 7% Hispanic, less than 1% Native American, less than 1% Pacific Islander, and the remaining 4% with missing information concerning ethnicity; conversely, White Caucasian is the only representative ethnicity in the French group.

Diagnostic Graded Effect

Comparison of the three diagnostic groups revealed differences in the extent of the effect of alcoholism on regions of the frontocerebellar and limbic circuitry. In particular, the lateral frontal cortex, medial frontal cortex, hippocampus, thalamus, and cerebellar vermis were especially vulnerable to the harmful effects of alcoholism, with graded effects of volume deficits from mild‐to‐moderate shrinkage in alcoholics without KS to more severe abnormalities in alcoholics with KS. These findings suggest that a compounded effect of high‐dose, chronic alcohol consumption, and nutritional deficiencies (notably thiamine) may explain this graded shrinkage in specific brain regions involved in the frontocerebellar and limbic circuitry and could be a source of heterogeneity in the patterns of brain abnormalities documented in alcoholism. Therefore, the previously observed neuropsychological continuum in episodic memory and executive functions could be the reflection of neuroanatomical continuity in alcoholics with and without Korsakoff's syndrome [Butters and Brandt, 1985; Pitel et al., 2008; Sullivan and Pfefferbaum, 2009].

Frontocerebellar Versus Limbic Volume Differences by Country

Components of frontocerebellar and limbic circuitry were not homogeneously compromised by alcoholism in the two countries. In particular, the thalamus was more affected in French than U.S. alcoholics. Thiamine deficiency can lead to acute Wernicke's Encephalopathy (WE) [for review, Sechi and Serra, 2007] with brain abnormalities particularly observed in the thalamus and mammillary nuclei of the hypothalamus [Victor, 1990; Victor et al., 1971]. Alcoholic patients represent a high‐risk population for thiamine deficiency marked by a considerable variability for developing this vitamin depletion [Thomson, 2000; Thomson and Marshall, 2006]. Further, in vivo [Pitel et al., 2011] and postmortem studies [Harper, 2006] suggest that WE is underdiagnosed in alcoholics. Taken together, these data lead to the speculation that a higher prevalence of undetected, subclinical WE in French alcoholic patients may explain the greater thalamic volume deficit in French than U.S. patients.

In contrast with the findings in the thalamus, the hippocampus, amygdala, and cerebellar vermis were more sensitive to the compounded effect of alcoholism and presumed thiamine deficiency in KS in the U.S. than France. Follow‐up analysis revealed a graded effect of volume shrinkage in the hippocampus and the vermis from uncomplicated alcoholics to KS in the U.S. and a specific volume deficit in the amygdala of U.S. KS patients. By contrast, this volume gradation was not present in these regions in the French group. Country‐related differences in patterns, rather than severity alone, of regional brain shrinkage suggest specificity in regional brain damage by country. In addition to the combined effect of poor diet quality and putative alcohol toxicity itself, thiamine (B1) deficiency observed in patients with WE is often associated with other B‐vitamin deficiencies including pyridoxine (B6), folate (B9), and cobalamin (B12). These micronutrients are linked to homocysteine (Hcy) metabolism and their deficiencies could contribute to the high Hcy blood levels (i.e., hyperhomocysteinemia) associated with chronic alcoholism [Cravo et al., 2000; Harper and Matsumoto, 2005]. Given that Hcy has been considered a risk factor for brain atrophy in general [Sachdev, 2005 for a review], alcoholism‐related brain damage could be potentially explained by high Hcy [cf., Bleich et al., 2003, 2004]. Another relevant factor to consider is the type of alcoholic beverages consumed. For example, lower concentrations of homocysteine have been demonstrated in beer drinkers compared with drinkers of wine or spirits [Cravo et al., 1996]. Similarly, magnesium is a significant co‐factor in many thiamine‐dependent enzymes, and its lack of replacement during clinical treatment of WE could hamper the efficacy of parenteral thiamine [Sechi and Serra, 2007 for a review]. Therefore, in the early symptomatic stages of WE, treatment must be promptly administered that includes adequate parenteral thiamine doses in association with other B vitamins and magnesium supplementation when WE is suspected [Thomson et al., 2012].

Variability in medical decisions concerning the adequate treatment protocol for possible WE [Thomson et al., 2012] could contribute to the heterogeneity of brain damage in KS within and between countries. Further, fundamental differences such as the access to health care could contribute to national variability, especially during the occurrence of WE requiring a timely intervention. In 2011, the French population gave up or postponed health care due to financial difficulties at approximately the same level as in the United States (France = 29% vs. U.S. = 25%) [http://www.europ-assistance.com/sites/default/files/ea_cham2011_synthesis_en.pdf]. However, French generally are more likely to defer dental or vision care, whereas U.S. populations delay routine primary medical care and costly treatment. In the U.S., health coverage disparity is even more pronounced in those of low socioeconomic status.

In addition to vitamin deficiencies per se, a genetic vulnerability to these deficiencies and to alcohol effects could contribute to these national differences [Guerrini et al., 2009]. A selective genetic component in the pathogenesis of WKS may partly explain the specificity in brain abnormalities between the two countries. For example, the thiamine transporters related to expression of the SLC19A2 and SLC19A3 genes could play a crucial role in pathophysiology of alcohol‐related thiamine deficiency and, therefore, in heterogeneity of alcohol‐related brain volume deficits in KS [Guerrini et al., 2009; Kono et al., 2009]. The interaction of poor nutritional status and selective genetic polymorphisms may contribute to differential brain damage in the frontocerebellar and limbic circuitry in U.S. compared with French KS patients.

Contributions From Alcohol Factors

Different relationships between regional brain volumes and alcohol history variables (lifetime alcohol consumption, duration of alcoholism, and length of sobriety) were present in the alcoholic groups from each country. In French alcoholics, greater volume shrinkage of lateral frontal cortex, vermis, and pons occurred with higher lifetime consumption of alcohol, and the lateral frontal cortex and thalamus were smaller in alcoholics with shorter sobriety. Moreover, the thalamus demonstrated a greater sensitivity to lifetime alcohol exposure in the U.S. than French alcoholic group despite the fact that French alcoholics had smaller thalamus volumes than U.S. alcoholics. Heterogeneity in the expression of alcoholism on brain structure may well be associated with patterns of drinking. Alcohol consumption can be characterized in terms of quantity (number of drinks per day) and frequency (number of days drinking per month); together, they are predictive of harmful alcohol‐related health consequences [Li et al., 2007]. Maximum amount of alcohol drunk on any one day is a relevant factor for predicting both sociological and medical alcohol‐related harm, particularly when heavy drinking occurs in an intermittent rather than continuous pattern [Greenfield et al., 2006] and with multiple withdrawals [O'Daly et al., 2012].

Limitations

A limitation inherent in any retrospective study is the restricted access to detailed social, demographic, and clinical descriptive data of the available subject samples. Indeed, investigation of the effect of patterns of drinking on brain abnormalities in alcoholics with or without KS is essential, given the association between alcohol drinking pattern and diet quality [Breslow et al., 2006]. Moreover, tobacco use is a common comorbidity in alcoholism and must be considered as a potential confounding variable in light of brain volume differences found between smokers and non‐smoker alcoholics [Brody et al., 2004; Durazzo et al., 2014; Gallinat et al., 2006; Gazdzinski et al., 2005; Meyerhoff et al., 2006]. Consideration of potential family history of alcoholism could also provide a crucial covariate given that adolescents with such a family history have higher‐risk for developing AUD and having premorbid structural brain volume abnormalities [Hill et al., 2001; Nurnberger et al., 2004]. In the context of these limitations, this retrospective study is incorporated within the framework of a current international multicenter research project between France and the U.S. where one of the aims is the identification of environmental and genetic factors (pattern of alcohol use, history of withdrawal symptoms including seizures, diet and nutritional status, genes, environment, socioeconomic status, and neuropsychological status) mediating the heterogeneity of brain structural manifestations of alcoholism.

Additional limitations relate to potential differences in multisite image data acquisition, processing, and quantification methods. Our study averted some fundamental acquisition differences because the image data collected at all sites were T1‐weighted SPGR, acquired on the same GE platform at the same magnet strength (1.5T). However, differences in acquisition protocols among sites, such as slice thickness or slice orientation, could have a systematic effect on registration accuracy and segmentation and thus effect region and tissue volume estimates factors beyond adjustment through pre‐processing methods. Nonetheless, to minimize the impact of these differences, our processing pipelines are designed to be agnostic to differences in acquisition protocols and make no assumptions, for example, about slice orientation or thickness. Additionally, all tissue and region volumes were computed in the native space of each image, to avoid interpolation artifacts and therefore were based on the exact pixel size and slice thickness of that particular image. Further, data from all sites were subjected to the same post‐processing method in the aggregate, and a common atlas‐based parcellation approach was applied for regional structural identification and volume determination. The ROI parcellation method used has considerable anatomical validity but involves having a priori hypotheses. Conversely, the fast, automated VBM approach enables whole‐brain investigation and yields relatively similar results for brain volumes assessment relative to the ROI method [Giuliani et al., 2005; Good et al., 2002; Whitwell et al., 2005]. VBM is problematic, however, when used for multi‐site studies and requires accounting for hardware and software factors that could result in a volume difference bias [Focke et al., 2011; Jovicich et al., 2009] but were not at issue herein.

CONCLUSIONS

This retrospective analysis of MRI data collected in large samples of controls, alcoholics with KS, and alcoholics without KS from France and the U.S., all diagnosed with a common interview schedule, enabled identification of discrepancies in brain volume deficits that may reflect fundamental national differences in the consequences of alcoholism on brain structure while minimizing differences in methods for brain image analysis. The substantial heterogeneity in the presentation of alcoholism‐related brain dysmorphology precludes generalization regarding drinking‐related consequences to the overall population given the many environmental and biological differences idiosyncratic to an individual. Thus, it is clinically meaningful to identify patients with the most profound deficits and clinical antecedents that lead to their extensive brain damage. It is likely that those with the greatest neuropathology are those in the most need of urgent intervention (social, psychological, nutritional, and pharmacological) and prevention efforts, and for which the cost of inadequate or not treatment far outweighs that of treatment.

Supporting information

Supplementary Information