Memory decline in healthy older people

Methods.

Results.

For the entire group, regression estimates of performance on the CERAD WLDR test during the study period yielded slopes that ranged from −1.10 to +0.90 and intercepts that ranged from 5.9 to 10.5. The mean ± SD slope of WLDR performance over five visits for the entire group was 0.03 ± 0.33, and the mean intercept was 8.13 ± 1.24. Thirty-five subjects (34.7%) recorded a slope of less than 0 (negative slope) and were therefore classified as having declining memory. The remaining subjects (n = 66; 65.3%) were placed in the nondeclining memory group. Table 2 shows demographic data for each of these groups. The mean ± SD slope of the declining memory group was −0.31 ± 0.25, and the mean slope of the nondeclining memory group was 0.22 ± 0.20. For the declining memory group, this indicates a decrease in performance of 7.75% per year on the WLDR test (i.e., 0.775 of 10 possible words). For the nondeclining memory group, this represents an improvement in performance of 5.45% per year on the WLDR test (0.545 of 10 possible words). Although the slopes of the two groups were by definition different (t(99) = 57.32, p < 0.001), there were no significant differences between the intercepts of the declining memory (mean = 8.40, SD = 1.51) and nondeclining (mean = 7.98, SD = 1.07; t(99) = 1.522, p = 0.134) memory groups. Analysis of mean performance of both groups at all assessments indicated significant differences on WLDR score at assessments 2, 3, 4, and 5 but not at assessment 1. The magnitude of the difference in WLDR scores between the two groups increased from the first to the last assessment, as indicated by the increasing effect size described in table 3.

Table 4 summarizes the genetic, clinical, and neuropsychological status for the declining and nondeclining memory groups at the fifth assessment. Significant between group differences were observed on the percentage of subjects rated as having a CDR of 0.5. The declining memory group displayed significantly worse performance on the pattern–location paired associative learning test relative to the nondeclining memory group. No other group differences were observed.

Discussion.

During a period of 2 years, a large proportion of healthy older individuals showed progressively worsening episodic memory function as measured by performance on the WLDR test. It is unlikely that the memory decline observed in these individuals was caused by any systemic medical or psychiatric illness because of the strict exclusion criteria used. The average rate of WLDR test decline in these subjects was approximately 0.775 words per year (from a maximum score of 10), although no individual classified as having declining memory met the clinical criteria for probable AD at the final assessment. Despite this decline, group mean performance on the WLDR test always remained within the normal limits established for the test (i.e., >621). The rate of decline was greater than that observed in patients with severe probable AD on the same memory test,27 although floor effects associated with the performance of the AD patients would most likely have reduced estimates of decline in this previous study. No differences were observed here between the declining memory and nondeclining memory groups on WLDR score at baseline (table 3), suggesting that initial level of memory test performance could not predict group membership. Finally, levels of anxiety and depressive symptoms were equivalent in the declining memory and nondeclining memory groups, as was the frequency of complaints of impaired cognitive function. These findings have important implications for the clinical diagnosis of disorders, such as MCI and AD, because they suggest that cognitive decline can be detected in nondemented and high-performing older people using standard neuropsychological tests before these individuals meet conventional clinical criteria for cognitive impairment (i.e., CDR 0.5). However, detection of this decline requires the serial administration of valid and reliable neuropsychological tests over an extended period.

The proportion of individuals classified as having memory decline was larger than expected on the basis of epidemiologic estimates of the prevalence of AD or MCI in individuals older than age 65.36,37⇓ Health-related conditions other than neuronal degeneration (e.g., cardiac illness38) may have adversely affected cognitive performance in some individuals, increasing the estimated rate of memory decline among the current group of subjects. However, the strict exclusion criteria applied makes this explanation unlikely. The high estimate of memory decline may also result from psychometric limitations associated with administering the WLDR test to the same group of people on multiple occasions (e.g., ceiling effects). Alternatively, previous studies examining the prevalence of MCI among community-based cohorts may have underestimated the true rate of cognitive impairment through their inability to classify high functioning individuals who show mild memory decline as impaired. Regardless of the reason for the difference between current and previous estimates, it is unlikely that the cognitive abilities of 35% of people in their 60s will decline to a level indicative of a dementia. However, only further longitudinal analysis of the current cohort will determine the validity of this prediction. The study from which the current data are drawn was not designed to produce epidemiologic estimates of the prevalence of memory impairment or decline in the older community. That is, some components of the study design (e.g., the a priori exclusion of individuals with a history of cardiac or respiratory illness) may mean that the current cohort is not representative of the older community. However, these same components ensure that we are able to detect mild changes in the cognitive function via methods such as those described in this report.

Objective evidence of memory impairment on a single assessment is sufficient to obtain a classification of CDR 0.5 (questionable dementia) or MCI.14,39⇓ Approximately half of all older individuals classified as having MCI are diagnosed with neurodegenerative illness within 2 to 4 years.33,37⇓ In the current study, only 34.3% of individuals with declining memory were rated as impaired, according to CDR criteria at assessment 5. Although clinical scales such as the CDR are often used to stage dementia severity and monitor disease progression in patients with AD,26,27⇓ our results suggest that the sensitivity of CDR criteria for detecting memory decline in healthy older people is poor, because two thirds of older individuals with memory decline were misclassified as having normal cognitive function when this instrument was applied in the current cohort (i.e., false-negative diagnoses). Importantly, the proportion of the nondeclining memory group rated as CDR 0.5 at assessment five was low (i.e., approximately 5%), indicating that older individuals whose memory test performance remains static or improves are unlikely to meet conventional clinical criteria for impairment. Memory decline does not appear to be associated with impairment in other cognitive functions, because there were no differences observed between the performance of the declining memory and nondeclining memory groups on the remaining CERAD tests or the Mini-Mental State Examination at the final assessment. However, the presence of memory impairment in the memory decline group was validated by the observed impairments on the associative learning test. This measure has been shown previously to be sensitive and specific to the mild40 and preclinical33 stages of AD. Taken together, these findings emphasize the potential power of longitudinal analyses of cognitive function for the detection of individuals at risk for AD indicated by memory decline.

The apoE epsilon 4 allele is a genetic risk factor for late-onset sporadic AD and is thought to play a role in the neurodegenerative processes that are the hallmark of the disease.41 A consistent finding from previous research has been that apoE 4 is associated with impaired memory.12,42,43⇓⇓ In the current study, there was no difference in the proportion of the declining memory and nondeclining memory groups carrying the apoE 4 allele, suggesting that there is no association between apoE 4 and memory decline in healthy older people. This finding is consistent with some previous studies that have shown that the apoE genotype does not influence rate of cognitive decline in normal aging,11 mild memory impairment, or AD.44 However, apoE status is also related to cardiovascular health, and the exclusion of patients with cardiac illness from the current study may have influenced our results.45 Alternatively, the small sample size (35 subjects in the declining memory group) in the current study may not have provided the statistical power necessary to detect a subtle relationship between apoE status and memory function.

It has been proposed recently that self-reported cognitive impairment may be one of the earliest behavioral markers of AD.46 However, the association between mild memory decline identified objectively and subjective complaints of memory loss is inconsistently reported.47 In the current study, ratings of cognitive failures did not differ between older individuals with declining and nondeclining memory. This may have arisen because the memory decline observed was subtle and may not manifest as cognitive failures that occur in the individuals’ daily environment. This hypothesis is supported by previous studies that have suggested that older individuals with mild to moderate dementia may have some insight into their own memory deficit.16 However, a negative relationship between severity of cognitive impairment and level of insight seems counterintuitive, and therefore, further investigation of subjective cognitive function in nondemented older people with mild and severe memory impairments are necessary to address this important issue.

Our results suggest that episodic memory decline can be detected among healthy older people before an objective memory deficit is evident using standard clinical criteria. This finding has important implications for the design and implementation of classification systems aimed at identifying individuals at increased risk for AD on the basis of cognitive test performance, because it suggests that the inclusion criteria for such systems should require objective evidence of cognitive decline. The current results also suggest that cognitive decline should be determined through the use of neuropsychological measures of memory that not only are sensitive and specific to the type of impairments observed in preclinical and early stages of AD, but also have psychometric properties adequate to allow the modeling of longitudinal performance. We are currently using more sophisticated statistical methods (e.g., hierarchical linear modelling48) to validate the regression method of measuring decline described in this article. Such statistical modeling may provide further insight into the changes in memory functions that occur with aging and the early stages of neurodegenerative disease. However, estimates of decline based on the regression approach provide an index of performance that can be calculated simply for single subjects and are meaningful clinically because the units of measurement are preserved in the calculations. Finally, our results are consistent with recent research that has suggested that the apoE epsilon 4 allele and subjective complaints of cognitive failures are not associated with memory decline in healthy older people.