Rate of medial temporal lobe atrophy in typical aging and Alzheimer's disease

The ability of MRI to depict in vivo neuroanatomy precisely has led investigators to employ this technique in evaluating brain morphometrics in aging and dementia.^1-12 Particular interest has arisen in MRI-based quantitative measures of the medial temporal lobe limbic structures because of the central role they play in memory function, and because these areas are involved first and most severely by the neurofibrillary pathology of AD.¹³ Although a number of pathologic and imaging studies have documented a decline in brain volume in typical aging, and an accelerated rate of volume loss in AD, with a few exceptions these studies have been cross-sectional in design. To assess the effects of aging on the medial temporal lobe directly, serial measurements in the same individuals are required. The goals of this project were to determine the annual rate of volumetric change of the hippocampus and temporal horn in cognitively normal elderly control subjects who represent typical aging individuals, to determine the annual rates of volumetric change in a group of individually matched patients with AD and to test the hypothesis that this rate was different from that in control subjects, and to assess the influence of several clinical variables such as age, gender, presence of hypertension, cardiac ischemic disease, diabetes, estrogen replacement, and apolipoprotein E (APOE) genotype on these rates of volumetric change.

Methods. Recruitment and characterization of subjects. Forty-eight subjects are included in this report: 24 AD patients and 24 control subjects who were individually matched with the patients for gender and age (±4 years; table 1). Control subjects and patients ranged in age from 70 to 89 years. Patients with AD and the cognitively normal control subjects for this study were recruited from the Mayo Alzheimer's Disease Center/Alzheimer's Disease Patient Registry. Informed consent was obtained for participation from the subjects or an appropriate proxy.

Control subjects were recruited from the pool of patients coming to Mayo primary care physicians for a general medical examination. The two criteria for cognitively normal control subjects were no active neurologic or psychiatric disorders. Some had ongoing medical problems; however, the illnesses or their treatments did not interfere with cognitive function. A control subject was identified for each AD patient and matched by gender and age ±4 years.

The diagnosis of AD was made according to the National Institute of Neurological and Cognitive Disorders/Alzheimer's Disease and Related Disorders Association (NINCDS/ADRDA) criteria.^14,15 The severity of AD was classified on the basis of the clinical dementia rating (CDR) score.¹⁶ Eleven patients had very mild disease (CDR = 0.5), 11 patients had mild disease (CDR = 1), and two patients had moderate disease severity (CDR = 2). Patients and control subjects were well matched on education and, by virtue of the study design, on age and gender as well (see table 1). The number of men in both the patient and the control groups was eight.

APOE genotyping was performed in all subjects. DNA was extracted from peripheral leukocytes and amplified by PCR.¹⁷ PCR products were digested with Hha I, and the fragments were separated by electrophoresis on an 8% polyacrylamide nondenaturing gel. The gel was then treated with ethidium bromide for 30 minutes, and DNA fragments were visualized by ultraviolet illumination.

The presence or absence of three vascular risk factors-hypertension, ischemic cardiac disease, and diabetes-was assessed by review of the medical records. Subjects were recorded as positive for hypertension if hypertension or its treatment was identified at any point in time in the medical record. The same criteria were applied to the diagnosis of diabetes. Subjects were considered to have coronary ischemic disease if any of the following diagnoses were identified: angina pectoris, myocardial infarction, coronary bypass surgery, or coronary angioplasty.

The presence or absence of estrogen replacement therapy was documented in all women. The age of menopause was established in each subject and subsequent estrogen replacement therapy was recorded as either present or absent through review of the medical records.

All subjects underwent an MRI examination protocol of the brain within 4 months of their initial clinical assessment. An identical MRI study was repeated 12 or more months after the initial MRI in all subjects, and this was linked with a second clinical assessment. Potential subjects were excluded if either of the MRI studies were of unacceptable diagnostic quality, demonstrated a focal structural abnormality, or if their clinical status changed between the serial MRI examinations (e.g., a control subject who developed cognitive impairment).

Imaging methods. All subjects were studied at 1.5 T using a standardized imaging protocol. A T1-weighted sagittal set of spin-echo images was used to measure total intracranial volume. A three-dimensional (3D) volumetric spoiled gradient echo sequence with a repetition time of 27 msec, an echo time of 9 msec, 124 contiguous partitions, a 1.6-mm slice thickness, a 22 × 16.5-cm field of view, 192 views, and a 45-deg flip angle was used to measure the volumes of the hippocampus and temporal horn.

All image processing steps (including boundary tracing) in every subject were performed by the same trained research assistant who was masked to all clinical information (i.e., age, gender, and clinical status). The date of each MR image was also masked in the image file so that image processing was done without knowledge of the chronologic ordering of the images in each pair. This ensured rigorous quality control, unbiased data generation, and uniformity in the subjective aspects of image processing across all the subjects in this study.

The 3D MRI data for both MR images were interpolated in the slice select dimension to give cubic voxels.¹⁸ An automated image registration program was employed to coregister the 3D image data set of the first image to that of the second image. This program was developed in-house and was based on the principle of minimization of interscan signal intensity differences across all voxels. The data of both image 1 and image 2 were then interpolated in plane to the equivalent of a 512 × 512 matrix and magnified times two. The voxel size of the fully processed image data was 0.316 mm³. The images of the whole brain were then "subvolumed" to include the temporal lobes. An intensity inhomogeneity correction algorithm developed in-house was then applied to both MR images. After the boundaries of the hippocampi and temporal horns had been delineated on each anatomic slice, the number of voxels in each structure was calculated automatically with a summing region-of-interest function. These were multiplied by voxel volume to give a numeric value in cubic millimeters.

The borders of the right and left hippocampi were traced manually with a mouse-driven cursor for each slice sequentially from posterior to anterior.¹⁸ In-plane hippocampal anatomic boundaries were defined to include the CA1 through CA4 sectors of the hippocampus proper, the dentate gyrus, and the subiculum (figure 1). The posterior boundary of the hippocampus was determined by the oblique coronal anatomic section on which the crura of the fornices were identified in full profile. Thus, essentially the entire hippocampus-from tail through head-was included in these measurements. The entire hippocampal tracing process takes approximately 2 hours per patient. Subdivision of the hippocampus along its anteroposterior axis into three segments labeled head, body, and tail was accomplished as follows: The hippocampal head was defined to encompass those imaging slices extending from the intralimbic gyrus forward to the anterior termination of the hippocampal formation. The posterior margin of the hippocampal head was labeled imaging slice x, and the volume of the hippocampal tail was determined by summing the area of the hippocampus on successive slices beginning from the fornix crura to slice (x - 1)/2. The volume of the body consisted of the sum of the areas of successive slices beginning with slice ([x - 1]/2) + 1 and extending to slice x - 1. This method allowed for assignment of fractional slice areas in the event of an odd number of slices posterior to the hippocampal head.

A region-growing autotrace algorithm was employed to define the boundaries of the temporal horns bilaterally (see figure 1). The signal intensities of temporal lobe white matter adjacent to the temporal horn, as well as CSF in the temporal horn, were sampled in multiple places. The signal intensity threshold employed to define the temporal horn boundary was half of the maximal temporal lobe parenchymal signal intensity above background, where background is defined as CSF signal intensity.¹⁹ The posterior extent of the temporal horn was defined as the same imaging slice used to demarcate the posterior boundary of the hippocampal formation. The anterior boundary of the temporal horn was determined by its full anterior anatomic extent.

Statistical analysis. The primary end point in these analyses was the annual percent change in hippocampal and temporal horn volume. This was computed as the volume in cubic millimeters of image 2 minus that of image 1 divided by structure volume on image 1, divided by the duration between the two images (in years). To compare the annual percent change between AD patients and control subjects, the rank sum test was employed. Stepwise regression (stepping up) was employed to identify explanatory variables that might be associated with the volumetric end points. The regression analyses were performed separately for each group and also with the groups combined, using group as a variable. After identifying significant main effects we looked for two-way interactions. The variables included in the regression analysis were age, gender, presence of hypertension, cardiac ischemic disease, diabetes, estrogen replacement, APOE genotype, and duration of clinical follow-up. Differences in the annual rate of change between sides (right versus left) and among hippocampal segments (head, body, tail) were assessed with paired t-tests.

Reproducibility. To assess the reproducibility of the method, 10 young adult volunteers underwent the MRI protocol described earlier on two separate occasions, separated by 2 to 4 weeks. The volumes of the hippocampi and temporal horns in both images of each of the volunteers were then measured as described. Test/retest measurement reproducibility in 10 young adult volunteers was defined in terms of the coefficient of variation.

Results. Reproducibility. For the sum of the right and left hippocampus, the median coefficient of variation was 0.28% (range, 0.02 to 0.70%). For the temporal horn, the median coefficient of variation was 0.95% (range, 0.08 to 3.36%).

Control subjects. The average annual rate of hippocampal volume loss among control subjects was 75 ± 60 mm³, or -1.55 ± 1.38%. The negative sign in table 2 indicates a decline in volume from image 1 to image 2. The temporal horns increased in volume by 167 ± 200 mm³/year, or 6.15 ± 7.69% per year. Hippocampal volume measurements increased between the first and second MRI studies in one control subject, and temporal horn volumes decreased in four control subjects (figure 2).

The annual percent volume loss was significantly greater for the hippocampal head than for either the hippocampal body (p = 0.001) or tail (p = 0.014). The rates of hippocampal and temporal horn volume change were not different between the right and left side, nor were they associated with the interval between the two MRI studies. None of the clinical variables analyzed-age, gender, APOE genotype, estrogen replacement, hypertension, cardiac ischemic disease, or diabetes-were associated significantly with the rate of hippocampal or temporal horn volume change.

AD patients. The mean annual rate of hippocampal volume loss among AD patients was 150 ± 73 mm³, or -3.98 ± 1.92% (see table 2). The mean annual increase in temporal horn volume among AD patients was 660 ± 439 mm³, or 14.16 ± 8.47%. Measured hippocampal volumes decreased over time in all but one AD patient, and temporal horn volumes increased in all AD patients (see figure 2). The annual rates of volume change for the hippocampus and temporal horn were not associated with the interval between MR images, position within the hippocampus (head, body, tail), or side (right, left). None of the clinical variables analyzed-age, gender, APOE genotype, estrogen replacement, hypertension, cardiac ischemic disease, or diabetes-were associated significantly with the annual rate of hippocampal or temporal horn volume change. No difference in the annual percent change in hippocampal or temporal horn volume was present between AD patients with very mild (CDR = 0.5) versus mild (CDR = 1.0) disease severity.

Comparison of patients and control subjects. No differences in the prevalence of any of the three vascular risk factors-hypertension, cardiac ischemic disease, or diabetes-nor in the prevalence or duration of estrogen replacement were identified between the patient and control groups (see table 1). Six control subjects and 11 patients had APOE genotypes (ϵ3/4 or 4/4), which are known to increase the risk of AD. The difference was significant (p = 0.042). The rates of hippocampal volume loss (p < 0.001) and temporal horn enlargement (p = 0.002) were significantly greater in AD patients than in control subjects (see table 2).

The initial (at image 1) hippocampal and temporal horn volumes for patients and control subjects are found in table 3. Values are reported in absolute terms (in cubic millimeters) and have also been normalized for intersubject variation in head size by dividing by total intracranial volume. All volumes were significantly different between control subjects and patients (p < 0.001).

Discussion. Reproducibility. In performing a study that involves serial MRI on the same individuals, a potential source of test/retest measurement variability is variation due to changes in the MRI procedure itself. In addition to the subjective aspects of image tracing, it is possible that instrument drift of the MRI equipment or differences in subject head position may introduce study-to-study variation that would affect serial volume measures. For that reason we assessed the stability of MRI measures in a serial fashion in healthy young volunteers in whom no substantial biological change in hippocampal or temporal horn volumes would be expected over the course of a 2- to 4-week period. The median test/retest coefficients of variation were quite small for both the hippocampi (0.28%) and the temporal horns (0.95%), indicating excellent stability of serial measurements of these structures.

In our group of 48 patients and control subjects, hippocampal volume measurements increased by a small amount between the two MRI studies in one control subject (by 1.41%) and one AD patient (by 0.19%; see figure 2). These are clear examples of measurement imprecision, but are reasonably close to the expected test/retest variability found in the reproducibility study. The small decrease in temporal horn volume measured in four control subjects may represent measurement imprecision or may reflect true changes in ventricular size due to changes over time in medication, nutrition, or hydration status.

Typical aging. Studies on aging often identify two subgroups: typical aging and successful aging. Typical aging individuals are those who may have nondementing illnesses that increase in prevalence with advancing age. Of particular interest are several vascular risk factors, most notably hypertension, that increase in prevalence with advancing age and also are associated with impaired cognition in elderly individuals.^20,21 Successful aging refers to individuals who remain free of these comorbidities. Most imaging studies have involved subjects who fall into the typical aging category as normal control subjects. Whether normal aging should be defined as typical aging or successful aging is a contentious topic. The control subjects in this study were not selected on the basis of the presence or absence of vascular risk factors. This control group should therefore be fairly representative of the general elderly population.

Control subjects. The effect of aging on brain morphology has been studied for more than a century. Initial investigations involved the examination of autopsy material. In general, a loss of brain weight with age has been found. Some studies^22,23 indicate that brain weight is stable from roughly age 20 to 60, after which an age-related decline is seen. Other autopsy studies^24,25 found a linear decline with advancing age beginning at approximately age 20.

Imaging studies, both MRI and CT, have generally found a loss of global hemispheric parenchymal volume, and an increase in CSF volume with advancing age in cognitively normal subjects.^10,26-29 Results from recent studies assessing the impact of aging on hippocampal and other temporal lobe structures have been variable. Some have found age-related declines in volume in normal elderly individuals^4,30,31 whereas others have not.^3,32 Variability in the results from different centers are likely due to differences in the age range of the study population, the definition of normal aging, and the MRI methods and neuro-anatomic boundaries employed for measuring structure volume.

Most morphometric studies on aging have been cross-sectional in design. In contrast, our data represent a true longitudinal sample: The initial image served as the reference point and subjects acted as their own control. Therefore, we can be confident that bias was not introduced by such factors as different environmental or socioeconomic conditions experienced by successive age groups.² The rates of volumetric change in the hippocampus (-1.55% annually) and the temporal horn (6.15% annually) that we found seem large in comparison with those reported in prior cross-sectional studies. For example, Miller et al.²³ calculated a loss of brain weight of 2% per decade in autopsied individuals older than 55 years. Coffey et al.³⁰ estimated a rate of volume loss of the amygdala/hippocampus of 0.3% per year. In addition to their cross-sectional nature, another difference between these studies and ours is that the control subjects that we studied were older (range, 70 to 89 years). In a true longitudinal MRI study of the oldest old (mean age, 86.8 years), Kaye et al.³³ found an annual decline in hippocampal volume of -2.09% in cognitively normal control subjects. This is similar to the rate we found of -1.55% per year in slightly younger control subjects (mean, 81.0 years).

In control subjects, the segment of the hippocampus with the greatest annual percent volume loss was the head. This regional variation in the age-related hippocampal volume loss was hypothesized on the basis of a prior cross-sectional study.³⁴ Results of this longitudinal study confirm differential sensitivity of various portions of the hippocampus to age-related volume loss in typical aging.

We found no difference in the rates of volumetric change between men and women. In contrast, Gur et al.²⁹ in a cross-sectional study, found an effect of gender on the rates of regional hemisphere atrophy.

The absence of an association between age and the rates of volumetric change or the interval between imaging suggests a linear decline in volume with age. However, the age range evaluated in this study was fairly narrow-70 to 89 years-and the sample size was modest.

We found no evidence to indicate that the rates of hippocampal and temporal horn volumetric change in control subjects were related to several other clinical variables-estrogen replacement, APOE genotype, presence of hypertension, ischemic cardiac disease, or diabetes. It is possible, however, that the absence of association was due to the modest number of subjects. In contrast, some authors^35,36 have found an association between hemispheric atrophy/ventricular enlargement and hypertension.

AD patients. The temporal horns increased in volume in all AD patients, and the hippocampi declined in volume in all but one patient (see figure 2). The rate of hippocampal atrophy was not different between the right and left sides or among the hippocampal segments (head, body, tail). We found no association between age and the rates of volume change, which suggests that AD is associated with a linear rate of medial temporal lobe atrophy, although the same limitations mentioned earlier for control subjects apply to the AD patients. None of the vascular risk factors-hypertension, ischemic cardiac disease, or diabetes-were associated with the rates of volume change, which (as indicated for control subjects) could be a function of small sample size. The rates of atrophy were not associated with APOE genotype.

The annual rate of hippocampal volume loss in our group of AD patients-3.98 ± 1.92%-was smaller than that reported by Jobst et al.,³⁷ which was 15% per year. There were several methodologic differences between the two studies. Jobst et al.³⁷ used CT, the images were oriented in a tilted axial plane, and a single linear measurement of the minimum thickness of each medial temporal lobe was made. Using MRI-based hippocampal volume measurements, Fox et al.³⁸ found a 1.4 to 8.6% annual volume loss in three symptomatic family members of an autosomal dominant pedigree. These were individuals in their 40s, with an amyloid precursor protein mutation. Nonetheless, the rates of hippocampal atrophy are fairly similar to those observed in our group of older patients with sporadic AD.

Comparison of patients and control subjects. Both indices of the rate of medial temporal lobe atrophy (hippocampal volume loss and increased temporal horn volume) were approximately 2.5 times greater in AD patients than in individually age- and gender-matched control subjects. It seems reasonable to conclude that this is a direct manifestation of the progressive pathology of AD superimposed on that associated with typical aging. In addition, the hippocampi of AD patients were smaller (p < 0.001) and the temporal horns larger (p < 0.001) than those of matched control subjects at the initial baseline image (see table 3). These results are in accord with several CT and MRI studies that demonstrated both cross-sectional and longitudinal differences in global hemispheric brain/CSF measurements between control subjects and AD patients.^39,40

The prevalence of vascular risk factors was not significantly different between AD patients and control subjects, nor was the number of women treated with estrogen replacement therapy or the treatment duration. Therefore the differing rates of atrophy in control subjects versus AD patients could not have been due to the influence of these potentially confounding variables. With regard to APOE genotype, our results are in agreement with those of Growdon et al.,⁴¹ who found that although the prevalence of APOE ϵ4 was greater in AD patients than control subjects, the rate of clinical progression in patients with established AD was not different for ϵ4 carriers versus ϵ4 noncarriers. In our study, the proportion of AD patients with APOE ϵ3/4 or 4/4 was significantly greater than that in control subjects, consistent with APOE ϵ4 as a major risk factor for developing AD. However, we found no association between APOE genotype and the rates of volumetric change in AD patients.

Because the annual rates of hippocampal and temporal horn volume change overlapped between control subjects and AD patients, our data would indicate that it is unlikely that these measures could serve as a useful, stand-alone diagnostic test for AD in individual patients (see figure 2). However, each MRI measurement is associated with an error term that is fairly large relative to the annual volumetric change. As the time interval between serial MR images increases, the absolute volumetric difference between them will also increase and the magnitude of measurement error in proportion to measured difference will decrease. It is possible, therefore, that the rate of volume change over a larger time interval might better discriminate AD patients from control subjects because of improved precision in the difference measurement. The ability to measure with this technique the rate of volumetric change in areas of the brain that are selectively involved early by the pathology of AD might also be useful as a means of assessing the results of therapeutic intervention.

Acknowledgments

The authors thank Brenda Maxwell for her typing assistance and Ruth Cha for aiding in the statistical analysis.