Memory and MRI-based hippocampal volumes in aging and AD

Subjects.

The subjects recruited for this study were drawn from the Mayo AD Center/AD Patient Registry, which are longitudinal studies of aging and dementia in the community of Rochester, Minnesota. These projects draw subjects from the community who are receiving their primary care in the Division of Community Internal Medicine of the Mayo Clinic and from regional patients who are referred to the Mayo Clinic for an evaluation of a cognitive impairment. Individuals from both sources receive identical evaluations, which have been described in detail elsewhere.13,14,16,17

The patients suspected of having a cognitive impairment were seen by a neurologist, who took a history from the patient and a collateral source, performed a neurologic examination, and completed several behavioral scales. A research nurse saw the patient and family and obtained an extensive family history and recorded other behavioral data on the patients. Laboratory studies, including a chest radiograph, chemistry group, complete blood count, sedimentation rate, vitamin B₁₂ and folic-acid levels, sensitive thyroid stimulating hormone level, and syphilis serology, were obtained on all patients. Other studies such as an EEG and a CSF analysis were obtained as clinically indicated.

The patients received an extensive neuropsychological evaluation in two sessions, one for diagnostic purposes and one for research purposes. The first set of diagnostic neuropsychological tests included the Wechsler Adult Intelligence Scale–Revised, Wechsler Memory Scale–Revised, Auditory Verbal Learning Test, and Wide Range Achievement Test III.18 The second set of instruments for research included the Mini-Mental State Examination,19 Dementia Rating Scale,20 Free and Cued Selective Reminding Test,21 Boston Naming Test,22 Controlled Oral Word Association Test23 and category fluency procedures.24

Following the completion of this evaluation, a consensus committee consisting of behavioral neurologists, a geriatrician, neuropsychologists, and nurses, was convened to determine the clinical diagnosis. The diagnostic criteria for dementia and AD were derived from the recommendations of the Diagnostic and Statistical Manual for Mental Disorders, edition III–revised, and the National Institute of Neurologic and Communicative Disorders and Strokes/AD and Related Disorders Association Criteria, respectively.25,26 After the diagnosis had been made, the patients were then staged on the Clinical Dementia Rating scale and Global Deterioration Scale.27,28

Control subjects were derived from the same community population as the cognitively impaired subjects.17 These individuals underwent an evaluation similar to that described previously, including a general medical examination, neurologic examination, and battery of neuropsychological test. Subjects qualified as normal controls if, in the opinion of their clinicians, they were functioning normally in the community and did not have a cognitive impairment. They could not have had any active neurologic or psychiatric illnesses and could not be taking any psychotropic medications. It was possible to have comorbid illnesses such as hypertension and coronary artery disease, and these individuals could be taking medications for these disorders, because it was felt that these illnesses and their treatments did not interfere with the patient’s cognitive function. The Mayo AD Center/AD Patient Registry and MRI projects were approved by the Mayo Institutional Review Board.

MRI acquisition.

All subjects were imaged at 1.5 Tesla (Signa, General Electric) using a standardized imaging protocol. Volume measurements of the hippocampus, parahippocampal gyrus, and amygdala were derived from a T1-weighted three-dimensional volumetric spoiled gradient echo sequence with 124 contiguous partitions, 1.6-mm slice thickness, a 22 × 16.5-cm field of view, 192 views, and 45 degree flip angle.

Image processing.

All image-processing steps (including boundary tracing) in every subject were performed by the same trained research assistant, who was blinded to all clinical information in order to ensure that the volumetric data were generated in an unbiased fashion. This ensured rigorous quality control, as well as uniformity in the subjective aspects of image processing across all the subjects in this study. Validation studies show the median intrarater test–retest coefficient of variation of hippocampal volume measurements to be 0.28% (range 0.02–0.07%) with this method.29

The borders of the hippocampus, parahippocampal gyrus, and amygdala were manually traced on each slice. 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. In-plane hippocampal anatomic boundaries were defined to include the CA1 through CA4 sectors of the hippocampus proper, the dentate gyrus, and subiculum.

The superior boundary of the parahippocampal gyrus was defined as the gray–white matter interface between the subiculum and the parahippocampal gyrus white matter. Medially the parahippocampal gyrus was demarcated by CSF in the uncal cistern. Laterally and inferiorly its boundary was the collateral sulcus. In some cases, a clearly identifiable collateral sulcus was not present along the entire anteroposterior extent of the parahippocampal gyrus. For this reason, parahippocampal gyrus measurements were not possible in 16 controls and 14 patients with AD. The posterior boundary of the parahippocampal gyrus was defined in a manner identical to that for the hippocampal formation. The imaging slice immediately preceding that in which the hippocampal intralimbic gyrus first appeared when progressing from posterior to anterior was defined as the anterior boundary of the parahippocampal gyrus.

Depending on the degree of ex vacuo temporal horn dilation, the inferior border of the amygdala was either the uncal recess of the temporal horn or the alveus covering the hippocampal head. The anterior boundary of the amygdala is ill defined in nature; thus we defined it operationally to be the most anterior slice on which the head of the hippocampus was present. The posterior, superior, medial, and lateral boundaries of the amygdala were defined by gray–white matter borders, or if appropriate CSF in the uncal cistern.

Statistical analysis.

The volumetric measurements of the left and right medial temporal-lobe structures, hippocampal formation, parahippocampal gyrus, and amygdala were normalized to head size by dividing the volume of each structure (in cubic millimeters) by the total intracranial volume (in cubic centimeters) to correct for intersubject variation. Difference scores reflecting the asymmetry between left and right structures for the hippocampal formation, parahippocampal gyrus, and amygdala were calculated also. These structures in addition to age, education, gender, and diagnosis (control or AD where appropriate) were entered into a stepwise regression model (stepping up) to predict each cognitive variable (neuropsychological test measure). Specifically, linear regression models of the form are developed in which y is the cognitive variable and x₁, x₂, … represent the explanatory variables such as age, gender, and imaging variables. At each step, the next variable added to the model is the one with the smallest p value. Addition of explanatory variables to the model ceases if all of the p values for the variables not yet in the model exceed 0.05.

The MR volumes were calculated as percentiles relative to the control subjects adjusted for age, gender, education, and duration of disease.6 Age- and gender-specific volumetric percentiles were determined and converted to W scores (normal deviates) using the inverse of the standard normal distribution (the 95th percentile corresponding to a W score of 1.645, for example). Nonlinear terms and interactions corresponding to significant main effects were evaluated and, if terms were significant, the main effect terms were retained in the model regardless of the corresponding p value. Although p < 0.05 was used as the criterion for developing the final model, only terms that achieved significance at the 0.01 level are reported, to avoid spurious associations resulting from the large number of tests carried out.

The cognitive variables included an index of general cognitive function (Dementia Rating Scale), naming (Boston Naming Test), and multiple memory measures, because memory is often the earliest function to become impaired in AD and is functionally related to medial temporal-lobe structures. The multiple memory measures included list-learning procedures (Auditory Verbal Learning Test [AVLT], Free and Cued Selective Reminding Test [FCSRT]), paragraph recall (Wechsler Memory Scale–Revised–Logical Memory) and nonverbal memory measures (Wechsler Memory Scale–Revised–Visual Reproductions) to determine which type of procedure is most closely associated with the volumes of medial temporal lobe structures.

For the list-learning procedures, AVLT and FCSRT, two measures were calculated. A learning measure (LNG) was calculated by summing the number of words learned across each of the trials (AVLT-LNG for AVLT and FCSRT-LNG for FCSRT). For the FCSRT an additional measure that assessed the ability of the subjects to benefit from the use of semantic cues was calculated. This measure indicated the number of trials, from 0 to 6, in which the subject recalled all 16 words using a combination of free recall and cued recall. This measure has been shown to be sensitive to early impairment and at separating controls from AD subjects.13,14 The other measure of list learning was delayed recall, the ratio of free recall performance after a 30-minute delay to the number of words learned on the last learning trial (trial 5 for the AVLT and trial 6 for the FCSRT) expressed as a percent. The two measures for the Wechsler Memory Scale–Revised were the immediate and delayed recall for paragraphs (Logical Memory I and II) and nonverbal material (Visual Reproductions I and II), respectively.

Subjects.

There were 126 normal control subjects and 94 patients with probable AD. The demographic and clinical data characterizing the groups by dementia severity on the Clinical Dementia Rating scale are presented in table 1. Most of the probable AD patients were mild as noted by scores of 0.5 and 1, with a few of 2. The mild nature also allowed for more detailed neuropsychological assessment of the AD patients.

Combined AD patients and control subjects.

Table 2 presents the significant stepwise regression models for the combined control subjects and AD patients. The data were analyzed initially by combining the groups to elucidate the full spectrum of biologic relations between the volumes and cognitive function. Age, gender, education, and W scores for the volumes of the normalized hippocampus, parahippocampal gyrus, and amygdala were used to predict the various cognitive measures described previously.

The most salient features of the data shown in table 2 demonstrate a strong relationship between the hippocampus and learning and recall. Most notably the combined volumes of the left and right hippocampal formations correlated in a highly significant fashion with the learning and delayed-recall measures of both the FCSRT and the AVLT. A quadratic term (L = 54.87 + 6.09 W − 2.5W²) also appeared, implying that the relationships were not strictly linear as is demonstrated in the figure. Similar, but less prominent correlations were also found between the Logical Memory and Visual Reproduction subtests of the Wechsler Memory Scale–Revised. Lateralized relationships between verbal and nonverbal memory and left and right hippocampi, respectively, did not hold for the combined group of subjects (see below for AD patients).

• Download figure
• Open in new tab
• Download powerpoint

Figure. Curvilinear relationship between total learning score on the Free and Cued Selective Reminding Test and age across control and AD subjects: f(L) = 54.87 + 6.09W − 2.50W².

A slightly different relation was seen with the Boston Naming Test and volumes. The left hippocampus was correlated with naming performance, as was the difference between the left and right parahippocampal gyrus.

A significant relationship involving both linear and quadratic components was found between the Dementia Rating Scale and the hippocampus. To further dissect the structural and functional relationships and the contribution of the disease state, we performed similar analyses for AD patients and control subjects separately.

Control subjects.

In the control subjects, the primary variables accounting for the cognitive measures were age, education, and occasionally gender. The only volumetric variable to enter into the models involved an interaction of gender by difference between left and right hippocampus in predicting delayed recall on the AVLT.

AD patients.

As is shown in table 3, in the AD, the hippocampal formation or, more specifically, the left hippocampal formation correlated most closely with performance on the verbal tasks using delayed-recall measures. For paragraph recall on the Wechsler Memory Scale–Revised–Logical Memory, the left hippocampal formation volume and an age–by–hippocampal volume interaction explained delayed recall performance. For nonverbal memory performance (Wechsler Memory Scale–Revised–Visual Reproductions), the right hippocampal formation predicted delayed recall performance.

For the Boston Naming Test performance, the difference in the volumes between the left and right parahippocampal gyri predicted most of the performance in naming. In particular, the smaller the volume of the left parahippocampal gyrus relative to the right, the poorer the performance on the Boston Naming Test.