Effects of Chronic Alcoholism on Six Transcallosal Fiber Bundles
The six-bundle repeated measures ANOVA for FA revealed lower values across all fiber bundles in alcoholics than controls: group effect [F(1,173) = 20.917, p < 0.0001] but no bundle effect or group-by-bundle interaction. The repeated measures ANOVA for λL revealed higher values in alcoholics than controls [F(1,173) = 11.51, p = 0.0009] but no bundle effect or group-by-bundle interaction. The repeated measures ANOVA for λT revealed higher values in alcoholics than controls [F(1,173) = 26.642, p < 0.0001] and a trend toward a bundle-by-group interaction [F(5,865) = 2.55, p = 0.027]. Follow-up t-tests revealed significant group differences at the p < 0.0001 level for five bundles and a trend for the sensory-motor bundle [t(193) = 2.39, p = 0.018].
The group-by-three-DTI metric (FA, λL, and λT) repeated measures ANOVA revealed significant group effects at the p = 0.0001 level for five callosal bundles and a trend for the sensory-motor bundle [F(1,173) = 4.8, p = 0.03]. Metric [F(2,346) = 7.67, p = 0.0006] effect and metric-by-group interaction [F(2,346) = 6.04, p = 0.0026] were significant only for the frontal bundle, where FA [t(173) = 4.8, p < 0.0001] was lower and λT [t(173) = 5.19, p < 0.0001] was higher in alcoholics than controls, but λL only showed a trend in this direction [t(173) = 2.03, p = 0.044]. A similar but weaker pattern of results was seen for the occipital bundle: metric main effect [F(2,346) = 4.675, p = 0.0099]; group by metric interaction [F(2,346) = 3.78, p = 0.024].
lists raw values for each metric and fiber bundle. lists t-test comparisons between alcoholics and controls for each metric and fiber bundle.
Group Comparisons (t-tests) for Fractional Anisotropy (FA), λL, and λT in Six Transcallosal Fiber Bundles. Controls (n = 88) Versus Alcoholics (n = 87)
Sex Differences in Regional FA and Diffusivity of Transcallosal Fibers
plots means and standard errors for each DTI metric at each corpus callosum bundle separately for men and women in each diagnostic group. Two-factor (group and sex) ANOVAs were performed to test for interactions that would support the putative greater vulnerability of women than men to the deleterious effects of alcoholism. Regardless of diagnosis, men had lower FA [F(1,171) = 7.77, p = 0.006] and higher λT [F(1,171) = 7.37, p = 0.0073] than women in the parietal bundle and higher λL in the sensory-motor bundle [F(1,171) = 9.85, p = 0.002]. However, there was no significant group-by-sex interaction for any bundle or metric.
The alcoholic women had been sober for about twice as long as the alcoholic men, and some men had drunk considerably more than any woman; both conditions may have mitigated observing greater vulnerability to alcohol in the women. Therefore, we derived a subgroup of 40 men and 25 women who were matched in length of sobriety (91 days for the men and 103 days for the women), age (45 years for men, 45.7 years for women), and lifetime consumption of alcohol (557 kg for men, 517 kg for women) and tested for sex differences in FA, λT, and λL in each callosal bundle. This analysis also found no evidence for differences between alcoholic men and women.
Other Contributors to Transcallosal Fiber Degradation
Regression analyses tested the hypotheses that among the combined group of alcoholic men and women, older age, higher current depressive symptom scores, greater lifetime alcohol consumption, a shorter period of sobriety, lower than normal body mass indicative of poor nutrition would contribute to greater deviations from age norms for each fiber bundle and DTI metric. Older age, greater lifetime alcohol consumption, or shorter sobriety were not associated with greater deficits in any of the age-normalized fiber bundle metrics. BMI, which was lower in alcoholic women than alcoholic men (), was negatively associated with λL (r = −0.28, p = 0.0085) of the premotor bundle.
BDI scores, available only on a subgroup of participants, were not associated with any DTI metrics in alcoholic or control groups examined separately. However, across a combined sample of 103 alcoholics and controls, higher BDI scores were associated with higher λL at prefrontal (r = 0.255, p = 0.0092) and parietal (r = 0.295, p = 0.0025) bundles. A follow-up ANCOVA, with BDI scores as covariate, was performed for λL at parietal bundle, where group differences met Bonferroni significance criteria (see ). The alcohol group effect persisted [F(1,100) = 4.09, p = 0.05] after accounting for the BDI score.
Contributions from prior substance abuse and current/past cigarette smoking were assessed with separate ANOVAs (substance abuser vs. nonsubstance abuser; smoker vs. nonsmoker) each with an additional factor for sex. There were no main effects for either substance abuse or cigarette smoking and no interactions between either factor and sex. These negative results were also obtained when applied to the restricted sample of alcoholic men and women, matched on length of sobriety and lifetime alcohol consumption.
Performance Associations With DTI Metrics
Alcoholics showed marked deficits on the measure of balance [t(161) = 7.183, p = 0.0001], time to complete the Digit Symbol grid [t(165)=2.943, p = 0.0037], and the traditional Digit Symbol score [t(165) = 3.894, p = 0.0001], but not on the unannounced recall test of the symbols [t(162) = 1.123, p = 0.2633]. See for raw and Z-score means and standard deviations and comparisons. Two-factor ANOVAs with group and sex found that men took longer than women to complete the Digit Symbol grid [F(1,163) = 16.24, p ≤ 0.0001] and completed fewer items within 90 seconds [F(1,163) = 24.39, p ≤ 0.0001]. However, there were no sex differences in balance or Digit Symbol recall and no group-by-sex interactions for any test. Among all alcoholics, men performed worse than the women on the traditional Digit Symbol score [t(81) = 3.3, p = 0.0014] and time to complete the Digit Symbol grid [t(81) = 2.29, p = 0.025] but not on recall or balance. These sex differences persisted only for the Digit Symbol score [t(61) = 2.2, p = 0.03] in the restricted sample of alcoholic men and women, matched for lifetime alcohol consumption and length of sobriety. Among alcoholics, irrespective of sex, greater lifetime alcohol consumption correlated with poorer balance scores (r = −0.334, p = 0.0019) and lower scores on the Digit Symbol Test (r = −0.30, p = 0.0057); the parametric test results were confirmed with a nonparametric test (balance: ρ = −0.328, p = 0.0026; Digit Symbol: ρ = −0.328, p = 0.0026). Both of these alcohol consumption performance associations were observed only in the men (balance: r = −0.367, p = 0.005; Digit Symbol Test: r = 0.29, p = 0.027) and not in the women (balance: r = −0.165, p = 0.41; Digit Symbol Test: r = 0.055, p = 0.78).
Mean (SD) for Uncorrected and Age- and Education-Corrected Z-Scores for Controls and Alcoholics
Two alcoholic men were excluded from correlational analyses involving the ataxia test and time to complete the Digit Symbol grid because they were statistical outliers. Pearson correlations between behavioral tests and each fiber bundle across the alcoholics are summarized in . In the combined group of alcoholic men and women, longer time to complete the Digit Symbol grid was associated with higher radial diffusivity (λT) measures in prefrontal (r = 0.260, p = 0.017) and higher axial (λL) diffusivity measures in temporal (r = 0.302, p = 0.005) fiber bundles. Shorter time maintaining balance was associated with greater axial diffusivity (λL) in both sensory-motor (r=-0.364, p = 0.0007) and parietal (r = −0.321, p = 0.003) bundles. No associations were found between DTI metrics and the Digit Symbol recall test.
Pearson Correlations (r) Between Regional Tractography Metrics and Neuropsychological Test Scores in Alcoholics
To determine whether there was a double dissociation between the contribution of prefrontal and temporal bundles to Digit Symbol grid completion time and the contribution of sensory-motor and parietal bundles to balance, both Digit Symbol grid completion time and balance scores were entered as independent associates in multiple regression analyses for prefrontal λT, sensory-motor, parietal λL, and temporal λL. These associations are plotted in . For prefrontal λT, Digit Symbol grid completion time remained an independent associate (β coeff = 0.281, p = 0.01) after accounting for the association with balance (β coeff = −0.175, p = 0.105). For sensory-motor λL, both Digit Symbol grid completion time (β coeff = 0.271, p = 0.008) and balance (β coeff = −0.404, p = 0.0001) remained independent associates, together accounting for 22% of the variance. For parietal λL, balance remained an independent associate (β coeff = −0.349, p = 0.0013) after accounting for the association with Digit Symbol grid completion time (β coeff = 0.163, p = 0.122). Finally, for temporal λL, Digit Symbol grid completion time remained an independent associate (β coeff = 0.309, p = 0.005) after accounting for the association with balance (β coeff = −0.018, p = 0.866).
Fig. 2 Scatter plots illustrating associations between the two behavioral measures (balance on one foot with eyes closed and time to complete grid) used in multiple regressions with four diffusion tensor imaging metrics (prefrontal radial diffusivity, sensory-motor (more ...)