Corpus Callosum Atrophy Is a Possible Indicator of Region– and Cell Type–Specific Neuronal Degeneration in Alzheimer Disease: A Magnetic Resonance Imaging Analysis

Background  Pathological studies in Alzheimer disease indicate the specific loss of layer III and V large pyramidal neurons in association cortex. These neurons give rise to long corticocortical connections within and between the cerebral hemispheres.

Objective  To evaluate the corpus callosum as an in vivo marker for cortical neuronal loss.

Method  Using a new imaging technique, we measured region-specific corpus callosum atrophy in patients with Alzheimer disease and correlated the changes with neuropsychological functioning. Total cross-sectional area of the corpus callosum and areas of 5 callosal subregions were measured on midsagittal magnetic resonance imaging scans of 14 patients with Alzheimer disease (mean age, 64.4 years; Mini-Mental State Examination score, 11.4) and 22 healthy age- and sex-matched control subjects (mean age, 66.6 years; Mini-Mental State Examination score, 29.8). All subjects had minimal white matter changes.

Results  The total callosal area was significantly reduced in the patients with Alzheimer disease, with the greatest changes in the rostrum and splenium and relative sparing of the callosal body. Regional callosal atrophy correlated significantly with cognitive impairment in the patients with Alzheimer disease, but not with age or the white matter hyperintensities score.

Conclusions  Callosal atrophy in patients with Alzheimer disease with only minimal white matter changes may indicate loss of callosal efferent neurons in corresponding regions of the cortex. Because these neurons are a subset of corticocortical projecting neurons, region-specific callosal atrophy may serve as a marker of progressive neocortical disconnection in Alzheimer disease.

THE PATHOLOGIC substrate of the dementing process in Alzheimer disease (AD) has been described as a progressive cortical disconnection syndrome^1-3 due to the loss of cortical layer III and V large pyramidal neurons, particularly in association regions.^3-6 These neurons give rise to long corticocortical connections within and between the cerebral hemispheres.^7-10 The corpus callosum (CC) is the major commissure of the brain, and its fibers arise predominantly from large pyramidal neurons in cortical layer III.^9-13 Consistent with the loss of these pyramidal cells, neuropathological evidence of callosal atrophy, substantially exceeding age-related changes in nondemented elderly volunteers, has been demonstrated in AD.¹⁴ Several in vivo studies using magnetic resonance imaging (MRI) also have shown a notable decrease in the total cross-sectional callosal area, measured in the midsagittal slice, in patients with AD compared with healthy age-matched volunteers.^15-19 Several procedures to define CC subregions on the midsagittal MRI slice have been applied in patients with AD^15,17,18; reports of regional patterns of callosal atrophy in patients with AD, however, have not been consistent (Table 1).

Pathologic changes in the subcortical white matter can affect CC size. For example, callosal atrophy has been reported in patients with multiple sclerosis^20-22 or vascular dementia,²³ both disorders involving primary white matter degeneration. A recent study¹⁹ showed a correlation between the total callosal area and the extent of MRI white matter hyperintensities (WMH) in AD. Because WMH seen on T₂-weighted MRI scans can reflect degenerative processes that are affecting cerebral white matter,^24-27 primary white matter degeneration may contribute to callosal atrophy in AD. Callosal atrophy, however, also has been found in patients with AD with a small amount of WMH.^15,18 Furthermore, 2 studies^17,18 reported a correlation of callosal atrophy with the severity of cognitive decline in AD. Differences among previous studies in the regional distribution of callosal atrophy may be related to several factors, including differences in the prevalence of white matter degeneration, measurement techniques, and the severity of dementia.

In this study, we measured callosal atrophy in 14 patients with AD compared with 22 age- and sex-matched healthy volunteers using a new method for defining callosal subregions that is accurate and reproducible. We also quantitated the extent of WMH in our patients with AD and control subjects with a semiquantitative rating scale reported by Scheltens et al²⁸ in selected subjects with only minimal WMH. We hypothesized that the regional pattern of CC atrophy in a subgroup of patients with AD with a lack of primary white matter degeneration would reflect region- and cell type−specific cortical neuronal degeneration that leads to neocortical disconnection in AD.

The study comprised 14 patients (7 women and 7 men; mean age, 64.4 years [SD, 8.8], range, 54-81 years) with clinically diagnosed AD who were part of a longitudinal study of Alzheimer disease by the Laboratory of Neurosciences of the National Institute on Aging. Probable AD was diagnosed in 11 patients and possible AD in 3 patients according to criteria of the National Institute of Neurological Disorders and Stroke/Alzheimer's Disease and Related Disorders Association.²⁹ One of the patients with possible AD showed predominantly parietal lobe dysfunction. The second patient possibly had a single transient ischemic episode 10 years before the examination, whereas the third patient with possible AD, who presented only with discrete cognitive impairment at the examination, in a clinical follow-up later met criteria for probable AD. The degree of cognitive impairment was assessed using the Mini-Mental State Examination (MMSE).³⁰ Six patients with an MMSE score of less than 10 were classified as severely demented, 4 patients with an MMSE score of 10 or higher to less than 20 were moderately demented, and 4 patients with a score of 20 or higher were mildly demented. The mean (SD) MMSE score was 11.4 (8.6).

For comparison, 22 healthy volunteers, 11 women and 11 men, were selected. The mean (±SD) age of the control group was 66.6±7.8 years, with a range of 52 to 84 years. All control subjects scored 29 or higher on the MMSE (mean [±SD], 29.8±0.4).

To control for the effect of cardiovascular risk factors, subjects with a history or signs of hypertension, diabetes mellitus, and cardiac arrhythmia were excluded. The fasting plasma glucose level was less than 6.7 mmol/L (120 mg/mL) in all subjects included in the study, and electrocardiograms were normal. All selected subjects were normotensive—systolic blood pressures below 140 mm Hg and diastolic blood pressures below 90 mm Hg—except 3 subjects, 2 patients with AD and 1 control subject, who showed borderline systolic hypertension, with a systolic blood pressure of about 165 mm Hg and a diastolic blood pressure of less than 90 mm Hg. Intracerebral pathologic disorders, such as cerebral infarction, neoplasms, and chronic cerebrovascular lesions, were excluded for all subjects by normal MRI scans, other than the overall brain atrophy in the group with AD. Further substantial comorbidity, like focal neurological signs, hypothyroidism, and other pathologic conditions that may directly or indirectly influence cerebral structure and function were excluded in all subjects by history, physical and neurological examinations, psychiatric evaluations, chest x-ray films, electrocardiograms, electroencephalograms, brain MRI, and laboratory tests (complete blood count; sedimentation rate; serum electrolytes, glucose, urea nitrogen, creatinine, liver-associated enzymes, cholesterol, high-density lipoprotein, and triglyceride levels; antinuclear antibodies; rheumatoid factor; VDRL; human immunodeficiency virus test; serum vitamin B₁₂ and folate levels; thyroid function tests; and urinalysis). In particular, no subject included in the study showed periventricular areas of hyperintensity (PVH) extending more than 10 mm in the deep white matter or hyperintense foci in the deep white matter larger than 10 mm in diameter. No hyperintense foci at the rostral and occipital parts of the CC were visible on the axial-sliced MRI scan. Total and regional WMH and PVH scores are shown in Table 2.

All subjects or the holders of their durable powers of attorney signed consent forms to undergo MRI and neuropsychological assessment for clinical investigation and research. The study was approved by the National Institute on Aging's Institutional Review Board.

Each subject was administered the Mattis Dementia Rating Scale (DRS) by trained psychometricians to obtain an overall assessment of cognitive function.³¹ One patient who was assessed as severely demented could not complete the DRS examination. As further measurement of the severity of dementia, the duration of symptoms of AD was measured in years from the onset of symptoms to the examination date. In all cases, this information was obtained from near relatives and caregivers of the patients.

A total of 18 T₂-weighted and 18 proton-weighted axial slices (slice thickness, 6 mm; repetition times, 2000 milliseconds [ms] for both; and echo times, 80 and 20 ms, respectively) and 32 proton-weighted coronal slices (slice thickness, 6 mm; repetition time, 2000 ms; and echo time, 20 ms) were obtained on a 0.5-T tomograph (Picker Instruments, Cleveland, Ohio). Furthermore, 3 patients with AD were administered a 64-slice, T₁-weighted, and sagittally oriented volumetric sequence (slice thickness, 2.5 mm; in-plane resolution, 1×1 mm; repetition time, 36 ms; echo time, 6 ms; and flip angle, 30°). All remaining 33 subjects were investigated with a 90-slice, T₁-weighted, and sagittally oriented volumetric sequence (slice thickness, 2 mm; in-plane resolution, 1×1 mm; repetition time, 20 ms; echo time, 6 ms; and flip angle, 45°).

Areas of the total CC and 5 callosal subregions were measured by 1 of us (S.J.T.), who was unaware of the subject's diagnosis, in the sagittal T₁-weighted MRI slice that best represented the midsagittal section. This slice was chosen by using anatomical landmarks in a hierarchical order.³² First, slices were selected that showed no or only minimal white matter in the cortical mantle surrounding the CC. When more than 1 slice fulfilled this criterion, as might happen when a subject's head was not positioned exactly perpendicular in the head coil of the MRI tomograph, the medial thalamic nuclei served as an anatomical landmark of a second order. The selected slice then showed the interthalamic adhesion connecting the left and right medial thalamic nuclei, or only the smallest size of the thalamus of either one or the other side. The transparent septum and the cerebral aqueduct were used in the third step to confirm the selection when 2 slices remained that showed a similar amount of thalamic substance.

After determining the slice that best represented the midsagittal section, the total callosal area was measured. This was done on a Sun workstation (Sun Microsystems, Mountain View, Calif) by manually tracing the outer edge of the CC on this slice using specially designed software (ANALYZE, Biomedical Imaging Resource, Mayo Foundation, Rochester, Minn) for measuring the region of interest. The areas of 5 callosal subregions were outlined in 2 subsequent steps (Figure 1). First, a rectangle was placed over the CC. The lower side of the rectangle cut tangentially the 2 lowest points of the anterior and the posterior parts of the CC. The rectangle's length was determined by 2 lines perpendicular to this lower side that cut the most anterior and the most posterior points of the CC. In the second step, a radial divider with 10 rays equidistant from each other was placed at the midpoint on the lower side of the rectangle. Its 4 upper rays divided the CC into 5 subregions. The number of pixels within each region was summed automatically and multiplied by the pixel size to obtain absolute values (in millimeters squared) for the areas of total CC and the 5 subregions (labeled C1-C5 in a rostral-occipital direction).

Periventricular and deep WMH were graded by 1 of us (S.J.T.) on all T₂- and proton-weighted axial slices for each subject using a slight modification of a rating scale.²⁸ Signal changes on the CC were added as an extra item to this scale. Regional PVH and WMH subscores were summed to obtain total PVH and WMH scores. High levels of inter-rater reliability have been reported for the WMH and PVH scores of this scale.^19,28,33

To assess intrarater reliability, 1 researcher (S.J.T), who was unaware of the diagnosis, twice measured the total callosal area and areas of the callosal subregions in 10 randomly chosen subjects. To evaluate the interrater reliability, 2 independent researchers (H.H. and S.J.T.), who were also unaware of the diagnosis, measured the total callosal area in 10 randomly chosen subjects.

The ages of patient and control groups were compared using the Student t test, and the extent of WMH and PVH were compared using the Wilcoxon rank-sum test. Exploratory data analysis revealed that CC areas were normally distributed in both groups without any outliers, and the variance did not differ between the groups. Differences in the total callosal area between patients with AD and control subjects were assessed with the Student t test. Group differences in the distribution of callosal areas were tested using repeated-measures analysis of variance, with groups as the between-subjects factor and the 5 callosal subregions as the within-subject factor. A significant group-by-subregion interaction was followed up by a pairwise single-effect analysis using the Student t test. Interrater and intrarater reliability of the CC area measurements was assessed with the intraclass correlation coefficient.³⁴

Correlations of total callosal area with the WMH and PVH scores were determined using the Spearman rank-order correlation. Partial correlation was used to control the correlation of the total callosal area with the duration of symptoms, the MMSE score, and the DRS score for age in the group with AD. Because the MMSE and DRS scores in the group with AD were normally distributed without any extreme values and represented a relatively wide range of values, we selected the partial correlation approach for these data to control for age.^17,18 The correlation of the total callosal area with age in the AD and control groups was assessed using the Pearson product-moment correlation. To look for regional differences of callosal atrophy in dependence of the severity of dementia in the group with AD, 2 stepwise multiple regression models were constructed with the 2 callosal subregions as predictor variables for the MMSE score and the DRS score as dependent variables.

The patients with AD and the control subjects did not differ significantly in WMH scores (P≤.89), but the patients with AD showed significantly higher PVH scores than control subjects (P≤.002). The reliability of CC measurements was excellent. The intraclass correlation coefficient for the total callosal area was 0.96 for the interrater and 0.98 for the intrarater reliability; for regional CC measurements, intrarater reliability ranged from 0.98 for the subregion C1 to 0.75 for the subregion C3.

The absolute mean total callosal area in millimeters squared was significantly smaller in the group with AD than in the control group (P=.008). Furthermore, a repeated-measures ANOVA revealed a significant difference in the distribution of callosal area between patients with AD and the control group (P≤.001). To further elucidate the regional pattern of callosal atrophy, areas of the 5 callosal subregions were compared separately between groups. Areas of the most rostral (C1; P=.007) and the most occipital (C5; P=.006) subregions were significantly decreased in the group with AD. Areas of the 2 callosal subregions representing the posterior part of the callosal body (C3 and C4), however, were not significantly different between the 2 groups (P≤.25) (Table 3). Removing the 3 subjects with possible AD from the analysis of regional callosal atrophy did not essentially change the pattern of callosal atrophy.

In the group with AD, the total callosal area was not correlated significantly with the PVH (Spearman ρ=−0.37; P=.18) or the WMH (Spearman ρ=−0.05; P=.86) score. The total callosal area was not correlated with age in the patients with AD (r=0.1; P=.60), but did correlate with age in control subjects (r=−0.44; P≤.05). The total callosal area was correlated significantly with the MMSE score in the patients with AD, even after controlling for age (partial correlation coefficient, −0.63; P=.03). The correlation between the total callosal area and both the duration of symptoms and the DRS score tended towards significance, even after controlling for age (partial correlation coefficient, −0.58 and 0.53, respectively; P ≤.08). Stepwise multiple regression revealed that the callosal subregion C3 accounted for a significant amount of variance of the MMSE and the DRS scores in the AD group (P≤.01 and P ≤.05, respectively), whereas the remaining subregions did not contribute significantly.

We have demonstrated atrophy of the CC in AD, predominantly in the callosal rostrum and splenium. The mean cross-sectional areas of the total CC in our control group and in the patients with AD was 505 mm²and 415 mm², respectively, which compare with 2 previous MRI studies in which the mean areas of the CC measured 633 mm²and 562 mm², respectively, in healthy elderly subjects and 480 mm² and 469 mm², respectively, in patients with AD.^15,18 Our results support previous studies that show total callosal atrophy in patients with AD. The reduction of the total callosal area was 18% relative to the control group in our study compared with a range of 18% to 38% in previous studies.^14-19 Previous reports on the regional distribution of callosal atrophy have not been consistent. Each of the 3 major parts of the CC, the rostrum or genu, the truncus, and the splenium, was found to be atrophic in at least 1 study (Table 1).^14,15,18 Janowsky et al¹⁷ reported atrophy of all callosal subregions in their patients with AD.

Methodological differences in CC measurements may contribute to these discrepancies. The interrater and intrarater agreement of our CC measurements was high for the total callosal area and for the callosal subregions. Weis et al¹⁴ proposed a region-defining procedure that may have introduced ambiguity in the evaluation of the most anterior CC region because the rostrum of the CC was cut twice by the first of their dividing lines in at least 1 subject. Biegon et al¹⁵ used the most anterior point of the CC as their fixed position for all further steps. The selection of this point depends on the position of the patient's head in the field of view. This random moment affecting the first step may lead to variability in the subsequent evaluations. In our method, we took advantage of the reproducibility of the callosal length measurement introduced by Weis et al¹⁴ and tried to avoid the ambiguities of the outlined region-defining procedures.

Discrepant findings on the regional pattern of callosal atrophy also may be related to the different prevalence of primary white matter degeneration in the investigated subjects. In contrast to the results of Vermersch et al,¹⁹ the WMH and the PVH scores in our patients with AD were not correlated with the callosal area. This was expected because our subjects were selected to have a small amount of white matter changes. Callosal atrophy in patients with AD with a small amount of WMH has been previously reported.¹⁸ Because callosal fibers arise predominantly from layer III large pyramidal neurons, callosal atrophy independent of white matter changes suggests that the loss of these cells in the cortex is the primary cause of the CC area reduction.

The functional topographic organization of the CC in the human brain is not yet fully understood. In nonhuman primates, the major part of callosal commissures arises from the cortical association areas.^8,35 In the rhesus monkey, there is an anteroposterior pattern of callosal pathways representing the cortical area of their origin. Thus, the fibers stemming from frontal cortical areas traverse the CC in its rostrum and genu, whereas primary sensorimotor, posterotemporal, parietal, and occipital cortical areas are represented in the body and the splenium.^35,36 Fibers from different cortical areas, however, seem to overlap to a considerable degree. Neuropathological findings suggest a similar pattern of fiber organization in the human CC.³⁷ Intraoperative electrophysiologic stimulation of the CC in patients with epilepsy has shown that the rostrum and genu and the rostral part of the truncus project mainly in the frontal and posterotemporal lobe; the callosal body and splenium have been shown to send their fibers to the parietal and occipital cortices.³⁸ Because the major portion of CC fibers connects homotopic areas of the cerebral cortex,⁸ a limited inference on the organization of incoming projections in the CC may be drawn from this study. The primary sensorimotor areas send their commissures via the truncal CC.^35,36,38

Rostral atrophy of the CC, thus, may reflect mainly the loss of callosal efferent cells in the frontal cortex. The reduction of splenium area, the most prominent change in all subregions, may reflect the loss of layer III large pyramidal neurons in the occipital and parietal association cortices. The relative sparing of the primary sensorimotor areas in AD^5,39,40 may contribute to the lack of substantial atrophy of the CC body.

As the callosal efferent cells in cortical layer III form a subset of cortical interconnective neurons^2,7 and as ipsilateral connections are found parallel to contralateral projections in the neocortex,^8,41 callosal atrophy in AD may not only indicate a loss of interhemispheric connectivity but also serve as an indirect marker for increasing disintegration of ipsilateral association areas.

In this study, the total callosal area was significantly correlated to the MMSE score. Correlation between the total callosal area and the duration of symptoms and the DRS score showed a strong tendency towards significance. Janowsky et al¹⁷ also have shown a significant correlation of the callosal area with the MMSE score, contrary to the findings of previous studies.^15,19 Correlational analyses between the CC and the DRS score in patients with AD have not been reported previously. When we examined regional differences, we found a significant correlation of the C3 area (representing the middle truncus of the CC) with the MMSE score and the DRS score. Our results suggest that callosal atrophy increases with the severity of dementia in patients with AD, independently of age.

In summary, a region-specific reduction of the CC is seen in patients with AD even after excluding primary white matter degeneration as a considerable cofactor of commissural fiber loss. The MRI-based CC measurements, thus, may reflect the specific loss of large pyramidal neurons in cortical layers III and V of frontal and parieto-occipital association areas. Therefore, CC area measurements may serve as an in vivo indicator of the progress of neocortical disintegration in AD. Further studies are required to evaluate the clinical significance of the relation between callosal atrophy and neuropsychological test scores for interhemispheric information processing and the integrity of higher cortical functions in patients with AD.

Accepted for publication July 11, 1997.

This work was supported in part by a grant from the Ernst Jung Foundation, Hamburg, Germany (Dr Hampel).

Part of the material presented here is taken from the doctoral thesis of Mr Teipel (Ludwig-Maximilian University; in preparation).

S. Weis, MD, Department of Neuropathology, Ludwig-Maximilian University, provided helpful discussions. Professors H. Hippius, MD, and H.-J. Möller, MD, Department of Psychiatry, Ludwig-Maximilian University, provided generous support.

Reprints: Harald Hampel, MD, Department of Psychiatry, Geriatric Psychiatry Branch, School of Medicine, Ludwig-Maximilian University, Nussbaumstr 7, 80336 Munich, Germany (e-mail: Hampel@psy.med.uni-muenchen.de).