PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of ageingLink to Publisher's site
 
Age Ageing. 2011 November; 40(6): 684–689.
Published online 2011 September 2. doi:  10.1093/ageing/afr101
PMCID: PMC3199215

Cognitive decline in the elderly: an analysis of population heterogeneity

Abstract

Background: studies of cognitive ageing at the group level suggest that age is associated with cognitive decline; however, there may be individual differences such that not all older adults will experience cognitive decline.

Objective: to evaluate patterns of cognitive decline in a cohort of older adults initially free of dementia.

Design, setting and subjects: elderly Catholic clergy members participating in the Religious Orders Study were followed for up to 15 years. Cognitive performance was assessed annually.

Methods: performance on a composite global measure of cognition was analysed using random effects models for baseline performance and change over time. A profile mixture component was used to identify subgroups with different cognitive trajectories over the study period.

Results: from a sample of 1,049 participants (mean age 75 years), three subgroups were identified based on the distribution of baseline performance and change over time. The majority (65%) of participants belonged to a slow decline class that did not experience substantial cognitive decline over the observation period [−0.04 baseline total sample standard deviation (SD) units/year]. About 27% experienced moderate decline (−0.19 SD/year), and 8% belonged to a class experiencing rapid decline (−0.57 SD/year). A subsample analysis revealed that when substantial cognitive decline does occur, the magnitude and rate of decline is correlated with neuropathological processes.

Conclusions: in this sample, the most common pattern of cognitive decline is extremely slow, perceptible on a time scale measured by decades, not years. While in need of cross validation, these findings suggest that cognitive changes associated with ageing may be minimal and emphasise the importance of understanding the full range of age-related pathologies that may diminish brain function.

Keywords: aged, 80 and over, cognition disorders, longitudinal study, elderly

Introduction

The average trajectory of cognitive ability over the last three to four decades of life is one of accelerating decline [1, 2]. This trend is seen in cross-sectional cohort comparisons and averages of longitudinal trajectories. Individual trajectories, however, reveal tremendous heterogeneity [3]. The degree to which cognitive decline is intrinsic to ageing and the clinical relevance of age-related changes represent long studied and controversial questions.

In this study, we evaluated changes in cognitive trajectories in a cohort of priests, nuns and brothers from the Religious Orders Study [4]. Our aim was to use an analysis framework that models individual heterogeneity directly and tests whether or not subgroups, with different rates of change, are present in the sample. Based on prior work in this cohort [3] and rates of cognitive decline noted by others [1, 5, 6], we hypothesised the analysis would reveal three subgroups; one with little change, one that declined rapidly and one with intermediate rates of change.

Methods

Study sample

The Religious Orders Study is an ongoing prospective clinicopathological study of ageing in a cohort of Catholic priests, nuns and brothers from about 40 locations across the USA. By signing an informed consent and an anatomic gift act, participants agree to donate their brains at death for neuropathological studies of ageing and dementia. The Institutional Review Board of Rush University Medical Center approved all protocols. As described previously [4], participants undergo a baseline standardised clinical evaluation including neurological examination and neuropsychological testing. Clinical diagnoses of dementia and Alzheimer's disease (AD) are made based on clinical evaluations and in-person examinations using the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer Disease and Related Disorders Association criteria [7]. Apolipoprotein E (APOE) genotyping was performed according to standardised methods [8]. The study has rolling admission and annual follow-up evaluations.

Global cognition

A global composite cognitive score (composite score) was derived from a battery of 19 annually administered neuropsychological tests [3, 9]. The tests, many of which are commonly used in clinical practice, include: Episodic Memory (Word List Memory, Word List Recall and Word List Recognition from the Consortium to Establish a Registry for Alzheimer's Disease (CERAD); immediate and delayed recall of Story A of the Logical Memory subtest of the Wechsler Memory Scale-Revised (WMS-R) and immediate and delayed recall of the East Boston story); Semantic memory (Verbal Fluency from CERAD, a 15-item version of the Boston Naming Test, a 15-item version of the Extended Range Vocabulary Test and a 20-item version of the National Adult Reading Test); Working memory (Digit Span subtests forward and backward of the WMS-R, Digit Ordering, Alpha Span); two tests of perceptual speed (the oral version of the Symbol Digit Modalities Test, and Number Comparison) and Visuospatial ability (a 15-item version of the Judgment of Line Orientation Test and a 17-item version of Standard Progressive Matrices). Using the baseline mean and standard deviation (SD) for each test, z-scores were created and averaged across measures to derive the composite score [4]. Valid completion of at least six measures was required to derive a score for any given evaluation; this allowed us to include telephone assessments for individuals who missed their in-person evaluation [10].

Post-mortem data

Autopsies were performed according to standard procedures at sites across the USA to accommodate participants' geographic distribution. Brains were sent to the Rush Alzheimer's Disease Center for evaluation by a board-certified neuropathologist blinded to all clinical data [1113]. Amyloid-β load was quantified [14] and a composite measure of amyloid-β deposition was derived. Neurofibrillary tangle density was characterised within and across regions of interest [14].

Analytic methods

We analysed repeated measures of performance with the composite score using random effects regression with a profile mixture model for random effects [15, 16]. The profile mixture portion of the model specifies that variability observed at baseline and change over time is partially due to the presence of distinct subgroups having different means and variances for the random effects [17]. Change over time was modelled with group centered years of age and age group indicator variables. This strategy enables modelling of change over discrete age periods with non-linear trajectories across a broad age range using a piecewise approach. Centered age group indicators permit interpretation of model intercepts as age-adjusted means. A fixed effect with observation basis was used to capture retest effects [18]. Parameters were estimated with Mplus software (version 6.0, Muthén and Muthén, Los Angeles, CA, USA). Command syntax is available upon request (R.N.J.).

We hypothesised that the cohort would be represented by three subgroups [3]. Starting values were based on means and variances of baseline scores and slope parameters for a preliminary single class model. Model fit was evaluated with standard approaches for profile mixture modelling, considering absolute fit of model-implied values to observed values (Empirical r2) [19], quality of assignment (entropy: range 0–1; higher values signal better models) [20], the probability of class membership (posterior probability) and relative fit to less restrictive models (i.e. one or two class models) with log-likelihood, sample size adjusted Bayesian Information Criterion (aBIC) [21] and residual plots. We classified respondents according to their most likely class and investigated across class differences in demographic characteristics, amyloid load and tangle density in the autopsy subsample.

Results

Our analyses included the first 1,135 consecutive participants. We excluded 86 persons clinically rated as having dementia at baseline. Demographic characteristics are reported in Table 1. The average baseline age was 75 years, the sample was predominantly white and over 90% completed greater than 12 years of education. The number of evaluations ranged from 1 to 16 and the median length of follow-up was 8 years. The overall follow-up rate exceeded 95%. At the time of these analyses, 422 persons had died, of whom 395 had a brain autopsy. Summary data on amyloid and tangle pathology was available for 281 participants.

Table 1.
Participant characteristics at first observation (n = 1,049)

The pace of change per year was −0.15 SD units [confidence interval (CI): −0.17 to −0.14] in the single class model. Given the baseline total sample SD, participants would be expected to experience a half SD decline in performance in about 3.3 years, likely representing a noticeable difference in cognitive performance [22].

Parameter estimates for the three class model are shown in Table 2. The first class (slow decline) included 678 participants and was characterised by a mean composite score at baseline of 0.17, implying that this class had a baseline age-adjusted mean composite score slightly above the total sample baseline average. Linear change over age was slow (slope = −0.04 SD units per year of age, CI: −0.05 to −0.03). A second class (moderate decline; n = 284) had a baseline composite score of −0.17 (about 0.2 SD below the overall sample baseline average). This class had a faster pace of change with age (slope = −0.19, CI: −0.22 to −0.16), declining about five times as fast as the slow decliners and experiencing a half SD decrement in cognitive performance in about 2.6 years. A third class (fast decline, n = 87) had low baseline scores (initial level = −0.32) and a fast rate of annual decline (slope = −0.57, CI: −0.65 to −0.48). On average, members of this class would experience a half SD decline in about 11 months. Retest effects differed by trajectory class. The fast decline class had a higher retest effect between their first and second observation, an average of 0.74 SD units, relative to 0.40 units in the moderate class and 0.27 units in the slow decline class. The Supplementary data available in Age and Ageing online, Appendix 2 figure shows plots of model-implied cognitive trajectories over age by class and age groups.

Table 2.
Parameter estimates from random effects mixture model of change of a composite global cognitive functioning score over time (up to 15 years; n = 1,049)

The overall empirical r2 for the three class model was 0.94. Fit was better at baseline for the slow and moderate classes. Posterior probabilities summarise the probability of membership in each class and were highest for the fast class (0.96). Entropy was 0.73; generally, values greater than 0.80 indicate good classification. Therefore, this model may constitute an over-extraction of classes.

For comparison, results from the two class model identified slow (n = 799) and fast decline classes (n = 250). The initial level of the slow class was +0.11 with an annual pace of change of −0.06. The initial level for the faster class was −0.25 with an annual pace of change of −0.38. The two class model had an entropy value of 0.81, but, had a higher aBIC value and a lower log-likelihood than the three class model. Both the two and three class models had an empirical r2 = 0.94. Therefore, the evidence suggested the three class model fit better. We retained the three class model because differences were quantitative rather than qualitative and our analytic goal was description and modelling, rather than development of a classification tool. We investigated model modifications for possible misspecification, including the assumption of homoscedasticity and the common effect of age group indicators on baseline scores and change over time across class. Modifications did not significantly improve model fit.

Study participants differed significantly by class at baseline on age, education and Mini Mental State Examination (MMSE) [23] score (Table 3). Relative to the slow decline class, the fast decline class had higher cumulative risk of death (relative risk (RR) 2.2, CI: 1.9 to 2.6) as did the moderate decline class (RR 1.5, 95% CI: 1.3 to 1.8). The fast decline class had the highest relative frequency of people with one or more APOE ε4 allele(s) (31/87 = 39%), about twice that in the slow decline class (114/435 = 21%).

Table 3.
Participant characteristics at first observation by most likely class membership (n = 1,049)

In the autopsy subsample (n = 281), the slow, moderate and fast decline classes were not significantly different in age, sex or education levels. Moderate and fast decline groups had more visits (χ2 = 6.0, P = 0.05) and longer follow-up (χ2 = 7.4, P = 0.03) than the slow decline group. We found significant differences in amyloid burden and tangle density across classes, amounting to a medium effect size [24] for amyloid load (Cohen's f = 0.25) and large effect (Cohen's f = 0.36) for tangle density (see Supplementary data available in Age and Ageing online).

Conclusions

These results suggest that rapid cognitive decline should not be viewed as a normal, expected consequence of ageing. Two-thirds of this cohort belong to a sub-group experiencing very slow global cognitive decline. Life expectancy at age 75 years in the USA is about 12 years. Over their life expectancy, the slow decline group can expect to decline about a half SD unit, likely just large enough to be perceptible [22]. However, a substantial minority of older adults experience a faster decline. Just over a quarter of our cohort are predicted to decline more than two sample SD units over 12 years, and nearly 10% are predicted to decline about seven SD units over the same time period. Based on empirical post-estimation and including retest effects, the time to decline half a SD on the composite score is 12 years for the slow decline class, 2.6 years for the moderate decline class and about 1 year for the fast decline class. Although Alzheimer's type neuropathological changes are present in all three classes, the burden is significantly greater in those with faster rates of decline. This class also had the highest frequency of APOE ε4 allele(s).

Retest effects differed by class in this sample, with the fast decline class demonstrating the most improvement. Prior work in this cohort showed that retest effects are uncorrelated with age and cognitive status [25]. Although labelled as a ‘retest effect’, this fixed regression parameter will absorb any misspecification in cognitive trajectories over time. Thus, there is a danger of over-interpretation, and differences may simply reflect a poorer fit in the fast decline class. Possible sources of misspecification include linear modelling of the ageing trend and the assumption of homogeneity across class in age-cohort effects. Sensitivity analyses revealed that relaxing the age-cohort homogeneity assumption significantly improved fit, but did not alter class assignments or retest effect differences. Further exploration of retest effects would be better accomplished in a study designed for that purpose.

These findings extend previous studies of cognitive trajectories in ageing. Heterogeneity in trajectories has been described previously and it has been shown that many elderly people experience relatively little change over extended periods [26]. This study helps quantify those observations and provides a framework for describing heterogeneity. Latent class analysis tests the hypothesis that a cohort consists of distinguishable groups. Instead of reporting intra-individual variability, our findings suggest it is useful to understand variability as being captured by patterns. These patterns define the distribution of individuals; quantitative rates of change and link rapid change to more extensive AD pathology. Few studies have attempted to define patterns of cognitive ageing. A recent report from the Cambridge City study using similar models found that a three group solution fit their data well [27]. Consistent with our data, rapid change affected a small minority (5%), although their very slow change group encompassed only about 40% of the sample. Differences may be due to entry criteria, population diversity and their use of the MMSE, which is insensitive to change at higher ability levels. Finally, relatively few studies have examined the relation of changes in cognition to neuropathological indices [2830], and we are not aware of prior reports examining the relation of neuropathology to different classes of decline.

The strengths of the study are the sample size, duration and completeness of follow-up (clinical evaluations took place proximal to death) and high autopsy rate. However, the specialised nature of the cohort is a limitation. Similarities among participants' lifestyles (i.e. Catholic Orders) and the high education level of the sample (93% have more than 12 years of education) reduce heterogeneity. However, this may improve our ability to detect associations by reducing residual confounding. Thus, these findings should be replicated in other cohorts. Another limitation is the use of a composite score rather than individual tests or domains. While composite scores are better than global measures of cognition in their ability to capture individual variation while reducing floor and ceiling effects [3], they cannot reveal differences in specific cognitive domains. These masked differences may provide valuable clues to the course and aetiology of cognitive decline. Further, the composite scores required a minimum of only six complete tests. A sensitivity analysis was performed in which participants with less than complete data for the composite score were dropped from the analysis and the results were relatively unchanged.

Most individuals in this sample had minimal decline over time. Among those experiencing decline, change correlated with neuropathology. Our findings generally support the notion that substantial cognitive decline is not intrinsic to ageing. Although this optimistic view needs to be validated in other populations, it underscores the importance of investigating the full range of pathologies that may diminish brain function in old age. A potential direction for future research is the development of clinical tools based on large population-based statistical models, which could be used to provide estimates for probable class assignment as a means to improve clinical decision-making.

Key points

  • Performance over time (up to 15 years, mean 8 years) on a global composite measure in a cohort of individuals aged 75 or older at baseline was analysed using random effects modelling with a mixture component.
  • Three groups of individuals with three different average rates of decline were found. A majority of individuals fell into a class characterised by relatively slow decline over time.
  • When a subset of the three groups' trajectories were compared with neuropathological studies, individuals in the slow decline class had the least amount of neuropathology.
  • Individuals in the intermediate and fast decline groups had intermediate and high levels of amyloid plaques and neurofibrillary tangles.
  • Results suggest that cognitive decline is not a normal part of ageing and is associated with some form of neuropathology.

Funding

Data for this study were collected in the context of the Religious Orders Study (P30AG10161, R01AG15819) and supported by the National Institutes of Health, National Institute on Aging. Study participants were older Catholic nuns, monks, and priests from California, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maryland, Minnesota, New York, Pennsylvania, Tennessee, Texas, and Wisconsin. Data were provided by the RELIGIOUS ORDERS STUDY group, David Bennett, MD, Principal Investigator (PI). The research was supported in part by a National Institute on Aging conference grant Conference on Advanced Psychometric Methods in Cognitive Aging Research, (R13AG030995) Dan Mungas, PhD, PI. K.M.H.'s effort was supported by a grant from the National Institute on Aging (K01AG029336); R.N.J.'s effort on this analysis was supported by a grant from the National Institute on Aging, the Harvard Older Americans Independence (Pepper) Center (P60AG008812).

Acknowledgements

The authors thank the study participants and the staff of the Rush Alzheimer's Disease Center.

References

1. Salthouse TA. Memory aging from 18 to 80. Alzheimer Dis Assoc Disord. 2003;17:162–7. [PubMed]
2. Verhaeghen P, Salthouse TA. Meta-analyses of age-cognition relations in adulthood: estimates of linear and nonlinear age effects and structural models. Psychol Bull. 1997;122:231–49. [PubMed]
3. Wilson RS, Beckett LA, Barnes LL, et al. Individual differences in rates of change in cognitive abilities of older persons. Psychol Aging. 2002;17:179–93. [PubMed]
4. Bennett DA, Wilson RS, Schneider JA, et al. Natural history of mild cognitive impairment in older persons. Neurology. 2002;59:198–205. [PubMed]
5. McArdle JJ, Ferrer-Caja E, Hamagami F, Woodcock RW. Comparative longitudinal structural analyses of the growth and decline of multiple intellectual abilities over the life span. Dev Psychol. 2002;38:115–42. [PubMed]
6. Hall CB, Lipton RB, Sliwinski M, Stewart WF. A change point model for estimating the onset of cognitive decline in preclinical Alzheimer's disease. Stat Med. 2000;19:1555–66. [PubMed]
7. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology. 1984;34:939–44. [PubMed]
8. Hixson JE, Vernier DT. Restriction isotyping of human apolipoprotein E by gene amplification and cleavage with HhaI. J Lipid Res. 1990;31:545–8. [PubMed]
9. Wilson RS, Schneider JA, Barnes LL, et al. The apolipoprotein E epsilon 4 allele and decline in different cognitive systems during a 6-year period. Arch Neurol. 2002;59:1154–60. [PubMed]
10. Wilson RS, Bennett DA. Assessment of cognitive decline in old age with brief tests amenable to telephone administration. Neuroepidemiology. 2005;25:19–25. [PubMed]
11. Bennett DA, Schneider JA, Bienias JL, Evans DA, Wilson RS. Mild cognitive impairment is related to Alzheimer disease pathology and cerebral infarctions. Neurology. 2005;64:834–41. [PubMed]
12. Mitchell TW, Mufson EJ, Schneider JA, et al. Parahippocampal tau pathology in healthy aging, mild cognitive impairment, and early Alzheimer's disease. Ann Neurol. 2002;51:182–9. [PubMed]
13. Schneider JA, Arvanitakis Z, Leurgans SE, Bennett DA. The neuropathology of probable Alzheimer disease and mild cognitive impairment. Ann Neurol. 2009;66:200–8. [PMC free article] [PubMed]
14. Bennett DA, Schneider JA, Wilson RS, Bienias JL, Arnold SE. Neurofibrillary tangles mediate the association of amyloid load with clinical alzheimer disease and level of cognitive function. Arch Neurol. 2004;61:378–84. [PubMed]
15. Muthén B. Beyond SEM: general latent variable modeling. Behaviormetrika. 2002;29:81–117.
16. Laird N, Ware H. Random-effects models for longitudinal data. Biometrics. 1982;38:963–74. [PubMed]
17. Xu W, Hedeker D. A random-effects mixture model for classifying treatment response in longitudinal clinical trials. J Biopharm Stat. 2002;11:253–73. [PubMed]
18. Rabbitt P, Diggle P, Smith D, Holland F, Mc-Innes L. Identifying and separating the effects of practice and of cognitive ageing during a large longitudinal study of elderly community residents. Neuropsychologia. 2001;39:532–43. [PubMed]
19. Singer JD, Willett JB. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press; 2003.
20. Celeux G, Soromenho G. An entropy criterion for assessing the number of clusters in a mixture model. J Class. 1996;13:195–212.
21. Nylund KL, Asparouhov T, Muthén BO. Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo Simulation Study. Struct Equ Modeling. 2007;14:535–69.
22. Norman GR, Sloan JA, Wyrwich KW. Interpretation of changes in health-related quality of life: the remarkable universality of half a standard deviation. Med Care. 2003;41:582–92. [PubMed]
23. Folstein MF, Folstein SE, McHugh PR. Mini-mental state'. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–98. [PubMed]
24. Cohen J. Statistical Power Analysis for the Behavioral Sciences. Hillsdale, New Jersey: Lawrence Erlbaum Associates; 1988.
25. Darby D, Maruff P, Collie A, McStephen M. Mild cognitive impairment can be detected by multiple assessments in a single day. Neurology. 2002;59:1042–6. [PubMed]
26. Christensen H, Jorm AF, Henderson AS, Mackinnon AJ, Korten AE, Scott LR. The relationship between health and cognitive functioning in a sample of elderly people in the community. Age Ageing. 1994;23:204–12. [PubMed]
27. Terrera GM, Brayne C, Matthews F. One size fits all? Why we need more sophisticated analytical methods in the explanation of trajectories of cognition in older age and their potential risk factors. Int Psychogeriatr. 2010;22:291–9. [PubMed]
28. Wilson RS, Leurgans SE, Boyle PA, Schneider JA, Bennett DA. Neurodegenerative basis of age-related cognitive decline. Neurology. 2010;75:1070–8. [PMC free article] [PubMed]
29. Green MS, Kaye JA, Ball MJ. The Oregon brain aging study: neuropathology accompanying healthy aging in the oldest old. Neurology. 2000;54:105–13. [PubMed]
30. Galvin JE, Powlishta KK, Wilkins K, et al. Predictors of preclinical Alzheimer disease and dementia: a clinicopathologic study. Arch Neurol. 2005;62:758–65. [PubMed]

Articles from Age and Ageing are provided here courtesy of Oxford University Press