Hippocampal formation size predicts declining memory performance in normal aging

Subjects.

Of all medically healthy and cognitively normal older adults participating in a study of memory impairment at the New York University Aging and Dementia Research Center, 101 individuals received clinical and psychometric follow-up evaluations between 2 and 6 years subsequent to baseline. These subjects consisted of community-dwelling research volunteers over 55 years of age who at baseline received a score of 28 or greater on the Mini-Mental Status Examination ^[6] and a rating of 1 or 2 on the Global Deterioration Scale ^[7] (GDS). GDS ratings of 1 and 2 identify subjects who manifest no evidence for intellectual impairment on the basis of a comprehensive psychiatric interview. In clinical studies of dementia, this GDS classification is used to define samples of cognitively normal control subjects. All subjects received a comprehensive battery of medical, neurologic, psychiatric, neuroradiologic, and laboratory examinations (blood and urine) and were excluded if any of these evaluations suggested conditions that could adversely affect brain functioning.

MRI evaluations.

Of the 101 longitudinally evaluated subjects, 44 received MRI scans at baseline that were used to derive head-size adjusted measurements of the HF, the superior temporal gyrus (STG), and the subarachnoid CSF (to estimate generalized cerebral atrophy). The STG was chosen as a control region because its neuropsychologic role is believed to be relatively independent of memory function and because it could be easily measured on the same set of cross-sectional images used to sample the HF. These measurements were made at baseline as part of a previously published crossectional study using techniques that were described in detail. ^[5]

Briefly, using a 1.5-T Philips Gyroscan (Shelton, CT), 4-mm-thick, T sub 1-weighted spin-echo images through the temporal lobe were acquired in a plane oriented 90 degrees to the long axis of the hippocampus. Slices depicting the body of the hippocampus from immediately posterior to the hippocampal head to the level of the posterior pulvinar were used to measure the size of the HF, the STG, and the subarachnoid CSF compartment (SCSF). These measurements were made directly from the digital images using proprietary region analysis software. HF and STG values were combined for the right and left hemisphere and, together with SCSF, regressed on a measurement of supratentorial intracranial volume to adjust for intersubject head size variability. The resulting residualized scores (rHF, rSTG, and rSCSF) were used in subsequent statistical analyses.

Psychometric evaluations.

For the 44 subjects with quantitative MRI data, the only neuropsychologic instrument measuring memory performance to be administered consistently both at baseline and at follow-up was the Guild Memory Test. ^[8] This examination consists of immediate (initial) and delayed recall of paragraphs and paired associates and a nonverbal test of memory for designs. Forward digit span is used to assess primary memory.

The Guild subtests selected for the statistical analyses were those we found sensitive to longitudinal decline in normal aging. This was investigated by means of a separate analysis of the larger parent sample (n = 101). The demographic characteristics of these subjects as well as the 44 with baseline HF, STG, and SCSF measurements are presented in Table 1.

Guild Memory Test stability in normal aging.

The six Guild subtests were examined for longitudinal change in the parent sample (n = 101). Only the paragraph recall subtests but not the paired associate, design, or digit span subtests demonstrated significant decline Figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Baseline and follow-up Guild Memory Test raw scores for cognitively normal elderly persons without baseline evidence for cognitive impairment (GDS 1-2, MMSE >or=to 28). The demographic characteristics of this sample is provided in Table 1. Longitudinal decline was evident only on the paragraph recall subtests (paired t^delay = 5.96, p < 0.001; paired t^initial = 4.62, p < 0.001).

HF size and longitudinal changes in paragraph recall.

Hierarchical multiple linear-regression models ^[9] were used to determine whether hippocampal formation size at baseline was significantly related to change in the paragraph recall scores over the follow-up interval. Using the paragraph delayed recall follow-up score as the dependent variable, the baseline score was entered at step 1. Beyond this point, additional independent variables will serve as predictors of longitudinal paragraph score change. In step 2, age at baseline, gender, years of education, and length of time to follow-up were entered as covariates. Entering hippocampal size (rHF) at the third step significantly increased the R² of the model. In a parallel analysis with the initial recall paragraph score as the dependent variable, a significant Delta R² at step 3 was also observed Table 2. When entered at the third step, rSTG did not produce a significant Delta R² in either the delayed or initial recall model.

To determine whether the predictive effect of hippocampal size on paragraph recall change was independent of generalized cerebral atrophy, rSCSF was entered at the third step. In the delayed recall model, rHF still produced a significant increment in R² when added to the equation as a fourth step (Delta R² = 0.07, p < 0.01). In the initial recall model, entering rHF after rSCSF produced a Delta R² of 0.06 (p = 0.07).

HF size and longitudinal changes in GDS score.

After covarying for the effects of age and gender, only baseline rHF, not rSTG or rSCSF, was significantly reduced in the 14 subjects who received a GDS score greater than 2 at follow-up (ANCOVA: F^HF = 6.44, p < 0.05; F^STG = 0.01, NS; F^SCSF = 0.64, NS).