Cross-Sectional Volumetric Analysis of Brain Atrophy in Alcohol Dependence: Effects of Drinking History and Comorbid Substance Use Disorder

Male subjects with alcohol dependence (N=134; Caucasian: N=113, African American: N=17, Hispanic: N=4) were admitted at the National Institutes of Health Clinical Center in Bethesda, Md., for an inpatient alcoholism treatment program. Community-recruited male subjects (N=32; Caucasian: N=22, African American: N=5, Hispanic: N=2, other: N=3) with no history of significant medical illness or psychiatric disorders were also included for comparison. To help minimize cohort effects in a cross-sectional analysis, the age range of both alcoholic and healthy comparison subjects was restricted to 30–50 years. Alcohol-dependent patients whose IQ was below 80, who had gross neurological abnormalities, who had a history of psychotic symptoms, or who were not eligible for an MRI scan (e.g., metal implants) were excluded. All subjects were assessed with the Structured Clinical Interview for DSM-IV (27), which confirmed that each patient met criteria for alcohol dependence and that no comparison subject met criteria for a psychiatric disorder.

For assessment of comorbid drug effects, alcohol-dependent subjects were assigned into one of three groups on the basis of substance use disorder comorbidity. Among alcohol-dependent subjects treated at NIAAA, the vast majority of patients with collateral substance use disorders are diagnosed with either cocaine or marijuana use disorders. Those with abuse or dependence of other substances (singly) were too rare for meaningful analysis and were excluded. The subjects with alcohol dependence alone (N=51) were those whose use of other psychoactive drugs (if any) did not meet substance use disorder criteria. The alcohol and marijuana use disorders group (N=33) comprised alcohol-dependent subjects who met DSM-IV criteria for marijuana abuse or dependence at some point in their life but never met criteria for substance use disorder for any other substance. Finally, the alcohol and cocaine use disorders group (N=50) comprised alcohol-dependent subjects who at some time in their life met criteria for either cocaine abuse or dependence. Only seven of the 50 alcohol-dependent subjects with a comorbid cocaine use disorder met criteria for only cocaine abuse or dependence and alcohol dependence. The majority (N=39) also met lifetime criteria for marijuana abuse or dependence, with 23 meeting lifetime criteria for two additional substance use disorders (besides alcohol and cocaine) such as sedatives or hallucinogens. The large majority of patients were also diagnosed with a comorbid mood disorder (41 of 51 subjects with alcohol dependence alone [80.4%], 25 of 33 alcohol-dependent subjects with comorbid marijuana use disorder [75.8%], and 41 of 50 alcohol-dependent subjects with comorbid cocaine use disorder [82.0%]). Patients had typically been drinking heavily for many years before seeking treatment, and it could not be confidently established whether the mood symptoms anteceded or were a consequence of the drinking. Therefore, differential diagnoses of primary or secondary mood disorder are not delineated.

A social worker administered a semistructured lifetime drinking history interview to each subject. Alcohol use history was divided into epochs of various use patterns according to each respondent’s history, and from these epochs we calculated three drinking history parameters: 1) age at onset of heavy drinking, defined a priori as the age at which the subject reported first consuming the equivalent of 90 drinks in a 1-month period; 2) years of heavy drinking, defined as the cumulative total contiguous or noncontiguous months during which the subject drank 90 drinks per month (note: since subjects often maintain this high a level of alcohol use for at least 12 consecutive months, months are summed into years); and 3) estimated lifetime alcohol consumption (in kg), which is a summation of all alcohol ingestion, including during periods where ingestion did not reach 90 drinks per month. Data were not normally distributed, so data were analyzed by rank. Subject age correlated positively with age at onset of heavy driinking (r_s=0.35, p<0.0001) and mildly with years of heavy drinking (r_s=0.17, p<0.05) but not with estimated lifetime alcohol consumption (r_s=0.09, p=0.31).

Subjects underwent imaging with a 1.5-T MRI scanner (GE Medical Systems, Milwaukee) using a fast spoiled gradient/recall acquisition in the steady-state sequence. The scans were performed no sooner than 3 weeks after the subject’s last alcohol use. We used a gapless series of high-contrast, 2-mm thick T₁-weighted coronal images (repetition time=25 msec; inversion time=5 msec; echo time=16 msec). The images were acquired by using a 256×256 matrix with a 240×240-mm field of view. Each volumetric brain consisted of 124 coronal slices with voxel size of 0.9375×0.9375×2.0 mm. We manually separated intracranial tissue from the skull on coronal sections with a hand-driven cursor. The intracranial volume included the cerebrum and CSF spaces covering the cortex but excluded the cerebellum and CSF of the posterior fossa. Interrater reliability for manual identification of the intracranial volume of 10 randomly selected MRI volumes was high (intraclass correlation=0.97). Next, intracranial volume was automatically segmented into gray matter, white matter, sulcal CSF, and ventricular CSF according to a previously described and validated computerized method that used voxel intensity information to perform a nonsupervised K-means clustering procedure. (For further details, see reference 28.) Cerebral brain volume was calculated as the sum of gray matter and white matter.

We chose to include both intracranial volume and cerebral brain volume in our analyses because they are conceptually distinct. The skull (and thus intracranial volume) grows in tandem with the brain during childhood until about age 15, at which point the volume of the brain (but not the skull) begins to shrink in late adolescence and throughout adulthood (29). Therefore, intracranial volume in an adult is a trait characteristic that reflects maximal brain growth during development, but brain volume in adults represents the outcome of both growth and atrophy processes. We also calculated intracranial-volume corrected volumes for gray matter and white matter (matter type divided by the difference of intracranial volume minus matter type) as an indirect measure of specific age-related atrophy of gray matter and white matter (30). The additional terms in the denominator were included to control for compositional data (interrelated correlates), since intracranial volume is a function of the sum of maximum CSF, white matter, and gray matter during development. Finally, we also calculated the gray matter/white matter ratio.

To disentangle independent effects (partial correlations) of age, drinking history, and comorbid marijuana or cocaine use disorder on volumetric measures, a series of multiple regression analyses were performed with age, comorbidity group, and each of the drinking history variables entered as independent variables, and each of the brain morphometric measures as the dependent variable. These analyses indicated that neither age, comorbidity status, nor any drinking history variable independently accounted for significant variance in intracranial volume.

With regard to total and proportional brain tissue volumes, age showed a negative partial correlation with brain volumetric measures in all analyses. Comorbid lifetime cocaine or marijuana use disorder, however, did not account for a significant portion of variance in any regression. In contrast, the number of years of heavy drinking showed a significant negative partial correlation with several brain volumetric measures after age and comorbidity were controlled (Table 1). To demonstrate minimal collinearity between age and years of heavy drinking as independent variables in model fitting, and to illustrate the independent relationships between both age and years of heavy drinking with brain atrophy, we used the model-fitting module of the statistical package JMP-SAS (SAS Institute; Cary, N.C.) to create leverage plots whose slopes represent the partial correlations between the cerebral brain volume/intracranial volume ratio and both age and years of heavy drinking (Figure 1).

The three alcohol-dependent groups did not significantly differ in race or age (Table 2). Tukey’s honestly significant difference tests indicated that in relation to those subjects with alcohol dependence alone, the alcohol-dependent subjects with a comorbid marijuana use disorder had a significantly earlier age at onset of heavy drinking (>90/month), and the alcohol-dependent subjects with a comorbid cocaine use disorder consumed significantly more alcohol. There were no significant group differences, however, in total and proportional brain volumes between the three groups, and mean intracranial volume was nearly identical across groups. While this report focuses on sources of volumetric differences among alcohol-dependent subjects, and comparisons between alcohol-dependent subjects and healthy subjects are presented in detail elsewhere, to put the brain volume measures of the alcohol-dependent subjects into broader perspective, we also present data from the 30–50-year-old male comparison subjects (Table 2). Parenthetically, analyses of variance followed by Dunnett’s post hoc comparisons with a healthy group indicated that all three alcoholic groups had significantly lower raw and intracranial volume-corrected brain volume, gray matter volume, and white matter volume than did the comparison subjects.

To assess whether alcohol-dependent subjects with and without comorbid marijuana or cocaine use disorders differed in how age correlated with brain volumes, we correlated age-by-brain volumes in each comorbidity group separately. Since age was not normally distributed, and to minimize outlier effects, Spearman (rank) correlations were performed (Table 3). We note that alcohol-dependent subjects treated at NIAAA often began heavy alcohol use in adolescence and were already showing significantly altered brain morphology by age 30 relative to healthy comparison subjects (data not presented), resulting in generally blunted age correlations with brain volumes later in adulthood compared with similar correlations in healthy subjects. The relatively steeper age correlations with volumetric measures in comparison subjects are presented for reference (Table 3). Age correlated negatively with intracranial volume-corrected gray matter in all three alcohol-dependent groups, and age also correlated negatively with intracranial volume-corrected cerebral brain volume in subjects with alcohol dependence either alone or with a comorbid cocaine use disorder (with a nonsignificant negative correlation seen for alcohol-dependent subjects with comorbid marijuana use disorder). The correlation between age and intracranial volume-corrected white matter was significant only in alcohol-dependent subjects with a comorbid cocaine use disorder, and the gray matter/white matter ratio decreased with age only in the subjects with alcohol dependence alone. There was no correlation between age and intracranial volume in any group.

We also compared correlations between proportional brain volumes with age by using one-tailed Fisher z tests with an a priori expectation of more severe white matter atrophy in alcohol-dependent subjects with comorbid cocaine use disorder. The negative correlation between age and the intracranial volume-corrected white matter volume tended to be steeper in the alcohol-dependent subjects with comorbid cocaine use disorder than in the subjects with alcohol dependence alone (p=0.061). The subjects with alcohol dependence alone demonstrated a steeper negative correlation between age and the gray matter/white matter ratio than did the alcohol-dependent subjects with comorbid cocaine use disorder (p=0.006) and, to a lesser extent, the alcohol-dependent subjects with comorbid marijuana use disorder (p=0.05). For illustrative purposes, linear fits of age-by-gray matter/white matter ratio for each group are overlaid in Figure 2.