Author Affiliations: Rush Alzheimer's Disease Center and the Departments of Neurological (Drs Wilson, Leurgans, and Bennett) and Behavioral Sciences (Drs Wilson and Boyle), Rush University Medical Center, Chicago, Illinois.
To characterize the course of cognitive decline during the prodromal phase of Alzheimer disease.
Longitudinal cohort study with up to 16 years of observation.
Older persons from 2 studies underwent annual clinical evaluations that included cognitive function testing and clinical classification of mild cognitive impairment, dementia, and Alzheimer disease. At baseline, there were 2071 individuals without dementia and 1511 without cognitive impairment.
During follow-up, 462 persons developed Alzheimer disease (20 with dementia solely due to another condition were excluded). Five to six years before diagnosis, the rate of global cognitive decline accelerated more than 15-fold. The acceleration in cognitive decline occurred slightly earlier for semantic memory (76 months before diagnosis) and working memory (75 months) than other cognitive functions. Mild cognitive impairment was also preceded by years of cognitive decline that began earlier (80 months before diagnosis) and proceeded more rapidly (annual loss of 0.102 unit) in the amnestic than in the nonamnestic (62 months, 0.072 unit) subtype.
Dementia due to Alzheimer disease is preceded by about 5 to 6 years of accelerated decline in multiple cognitive functions. By contrast, little decline is evident in persons who do not develop Alzheimer disease.
Dementia due to Alzheimer disease (AD), by definition, is preceded by a minimum of 6 months of cognitive decline. With recognition of the importance of developing strategies to delay the onset of AD symptoms,1,2 scientific interest in the prodromal cognitive changes has increased. There have been relatively few prospective longitudinal studies of this prodromal phase of the disease,3- 9 however, and estimates of its duration range from about 1 year6 to more than 10.5 Several factors probably contribute to this variability. One is that most studies have had fewer than 100 subjects with incident disease.3,7- 9 In addition, prodromal changes may occur earlier in some cognitive domains than others.6,8 Studies also likely differ in the cutoff point between cognitive normality and dementia.
Here we examine change in cognitive function during the prodromal phase of AD and before the onset of mild cognitive impairment (MCI), a common precursor to dementia in AD. The data are from 2 longitudinal cohort studies, the Religious Orders Study10 and the Rush Memory and Aging Project.11 For up to 16 years, more than 2000 participants underwent annual clinical evaluations that included detailed testing of cognitive function and clinical classification of MCI, dementia, and AD. In mixed-effects models that allowed rate of cognitive change to shift, we characterized the change in global and specific measures of cognition in the years preceding and following dementia onset in AD and conducted similar analyses of cognitive change before and after incidence of MCI.
The subjects are from 2 ongoing longitudinal studies, the Religious Orders Study, which began in 1994, and the Rush Memory and Aging Project, which began in 1997. Both studies involve annual clinical evaluation and brain donation at death and were approved by the institutional review board of Rush University Medical Center.
At the time of the analyses, 2323 individuals had completed the baseline evaluations and been found to be without dementia. Of these, 62 died before follow-up and 40 had been in the study less than 1 year. This left 2221 individuals eligible for follow-up; 2071 (93.2%) had follow-up data. The first set of analyses are based on this group. They completed a mean (SD) of 7.4 (3.8) annual evaluations. At baseline, they had a mean (SD) age of 77.4 (7.7) years, a mean (SD) of 16.2 (3.8) years of education, and a mean (SD) score of 28.2 (1.9) on the Mini-Mental State Examination; 72.0% were women; and 87.9% were white and not Hispanic. A second set of analyses focused on a subset of this group: 1511 persons without MCI or dementia at baseline. They completed a mean (SD) of 7.8 (3.9) annual evaluations and had a mean (SD) baseline age of 76.3 (7.4) years and mean (SD) of 16.4 (3.8) years of education; 72.8% were women; and 90.7% were white and not Hispanic.
At baseline and annually thereafter, subjects in each study underwent a uniform structured clinical evaluation that included a medical history, complete neurological examination, and cognitive performance testing. Based on these data and in-person evaluation of the subjects, an experienced clinician diagnosed dementia and AD using the criteria of the joint working group of the National Institute of Neurologic and Communicative Disorders and Stroke and the Alzheimer Disease and Related Disorders Association.12 The criteria require a history of cognitive decline and impairment in at least 2 cognitive domains, one of which must be memory for a diagnosis of AD. As previously described,13 impairment in 5 cognitive functions (orientation, attention, memory, language, and visuospatial ability) was determined in a 2-step process. First, an algorithm rated impairment in each area of function based on educationally adjusted cutoff scores on 11 individual tests.14 Second, based on all test data and information on education, sensorimotor problems, and effort, a neuropsychologist agreed or disagreed with each rating and supplied a new rating in the event of disagreement. Persons who had cognitive impairment but did not meet criteria for dementia were classified as having MCI and divided into amnestic (ie, with episodic memory impaired) and nonamnestic subtypes, as previously described.15 Individuals who met these criteria for MCI have been found to have intermediate levels of mortality,14,16 cognitive decline,14,17,18 and plaques, tangles, and cerebral infarction19 relative to persons without cognitive impairment and dementia. All clinical classification was done blinded to previously collected data.
Nineteen cognitive performance tests were administered in each study. The Mini-Mental State Examination was used for descriptive purposes and Complex Ideational Material was used only in clinical classification. Analyses were based on the remaining 17 tests. There were 7 measures of episodic memory: Word List Memory, Recall, and Recognition plus immediate and delayed recall of the East Boston story and story A from Logical Memory; 3 measures of semantic memory: verbal fluency, a 15-item form of the Boston Naming Test, and a measure of reading recognition; 3 measures of working memory: Digit Span Forward, Digit Span Backward, and Digit Ordering; 2 measures of perceptual speed: Number Comparison and Symbol Digit Modalities Test; and 2 measures of visuospatial ability: short forms of Judgment of Line Orientation and Standard Progressive Matrices. The primary outcome was a composite measure of global cognition using all 17 individual measures. Raw scores on each of the 17 measures were converted to z scores using the baseline mean and standard deviation of all subjects; z scores were then averaged to yield the composite measure of global cognition. Based in part on previous factor analyses of these cognitive measures in these20- 22 and other23,24 cohorts, individual cognitive measures were grouped into specific cognitive domains. Within each domain, individual z scores were averaged to yield composite scores that were used in secondary analyses: episodic memory (based on 7 individual measures), semantic memory (3 measures), working memory (3 measures), perceptual speed (2 measures), and visuospatial ability (2 measures). Further information on the individual tests and the derivation of the composite scores is described previously.20- 22
We used mixed-effects regression models25 to characterize paths of change in cognitive function. The models include change points that allowed rate of cognitive change to shift at some time before AD (or MCI) was diagnosed and again at the time of diagnosis. To identify the point before diagnosis when rate of cognitive change shifted, we constructed a series of models with change points ranging from a few months to several years before diagnosis and chose the analysis with the highest log-likelihood value, indicating the best model fit. To take advantage of all cognitive data, a composite measure of global cognition was the primary outcome. For AD, we conducted additional analyses excluding those with MCI at baseline and using composite measures of specific cognitive domains as outcomes. For MCI, the core model was followed by separate analyses of MCI subtypes. All analyses controlled for age, sex, and education. Programming was done in SAS (SAS Inc, Cary, North Carolina).26
The 2071 persons without dementia at study enrollment had up to 16 years of annual follow-up (mean [SD], 7.0 [4.0]). During this period, 462 persons developed incident AD. For these analyses, we excluded the 20 people who developed other forms of dementia. Those who developed AD were older than those who did not (81.3 vs 76.3 year; t845 = 14.1; P < .001) but the groups were similar in education (16.4 vs 16.2 years; t2049 = 1.4; P = .16) and sex distribution (72.7% women vs 71.9% men; χ21 = 0.1; P = .74).
We used mixed-effects models to characterize the change in cognitive function before and after AD was clinically diagnosed. To determine if the rate of decline accelerated at some point before the diagnosis, we fit a series of models that allowed for a shift in rate of cognitive change during periods ranging from 6 to 120 months before the diagnosis plus a second shift at the time of diagnosis. We used a composite measure of global cognition in initial analyses. At baseline, scores ranged from −1.910 to 1.466 (mean [SD], 0.094 [0.529]), with higher values indicating better function. The best-fitting model set the point of change at 65 months before diagnosis. In this model (Table 1), the global cognitive score declined a mean of 0.013 units per year before the point of change and 0.209 units per year thereafter, a more than 15-fold increase. After AD was diagnosed, there was a further increase of about one-third in annual rate of global cognitive decline. Figure 1, which is based on this analysis, shows the predicted 10-year paths of change in global cognition for 2 typical participants. Little cognitive decline is evident in the person who remained dementia free. In the person who developed AD, the diagnosis in study year 7 was preceded by more than 5 years of accelerated global cognitive decline.
Global cognitive decline before and after dementia onset in Alzheimer disease (AD). Predicted 10-year paths of decline in global cognition for typical participants who remained dementia free or developed AD in study year 7, adjusted for age, sex, and education.
At baseline, 552 subjects had MCI. To assess how this subgroup affected results, we excluded them and repeated the analyses in the remaining 1499, of whom 217 developed AD. The best-fitting change point was at 54 months before AD was diagnosed. Global cognition declined a mean of 0.016 units per year before this point (standard error [SE], 0.001; P < .001). The rate of annual decline increased more than 15-fold to 0.262 units (SE, 0.008; P < .001) during the 54-month prodromal period and to 0.314 units (SE, 0.021; P = <.001) after AD was diagnosed.
To see if results varied across cognitive domains, we conducted separate analyses using composite measures of specific cognitive functions (Table 1). As shown in Figure 2, the prodromal period began about 63 months before dementia onset for episodic memory, 76 months for semantic memory, 75 months for working memory, 70 months for perceptual speed, and 65 months for visuospatial ability. Prior to the prodromal period, episodic memory was stable and there was gradual decline in the other domains. Decline in all functions increased sharply during the prodromal period with less marked acceleration after AD was diagnosed.
Decline in specific cognitive functions before and after dementia onset in Alzheimer disease. Predicted 10-year paths of cognitive decline for typical participants who remained dementia free or developed AD in study year 7, adjusted for age, sex, and education.
To further examine prodromal AD, we assessed the change in cognitive function during the development of MCI, widely recognized as a precursor to AD. Of 1511 persons without cognitive impairment at study enrollment, 742 developed cognitive impairment on follow-up. Those who developed MCI were older (78.1 vs 74.6 years; t1509 = 9.4; P < .001) and slightly more educated (16.6 vs 16.2 years; t1509 = 1.9; P = .05) than those who did not, with a comparable sex distribution (71.0% women vs 74.5% men; χ21 = 2.3; P = .13). There was an approximately 4.5-year period of gradual global cognitive decline prior to MCI onset (as shown by the term for time during prodrome in Table 2), with a further doubling in the rate of decline after MCI was diagnosed.
Because memory impairment in MCI has been associated with increased risk of dementia and AD,27,28 we conducted separate analyses for amnestic (observed in 487 patients) and nonamnestic (observed in 442 patients) MCI to examine their developmental histories (Table 2; Figure 2). A mean of 80 months before the diagnosis of amnestic MCI (Figure 3A), the annual rate of global cognitive decline increased more than 10-fold, with a further increase of about 80% following diagnosis. By contrast, in nonamnestic MCI (Figure 3B) the prodromal period began later (62 months before diagnosis) and the acceleration in global cognitive decline was less marked, with an approximately 2.5-fold increase during the prodromal period and a further 50% increase following diagnosis.
Global cognitive decline before and after onset of amnestic (A) or nonamnestic (B) mild cognitive impairment (MCI).
In a group of more than 2000 elderly people followed up annually for up to 16 years, we assessed the change in cognitive function during the prodromal phases of AD and its precursor, MCI. The rate of cognitive decline increased sharply about 5 to 6 years before dementia was diagnosed and showed a modest increase approximately 4 to 6 years before MCI was diagnosed. The results indicate that dementia in AD is preceded by many years of progressively accelerating cognitive decline.
We are aware of only one previous study of cognitive change preceding development of MCI, and the results were comparable with ours, with accelerated cognitive decline occurring 3 to 4 years before MCI diagnosis.9 In addition, we found that the prodromal period for amnestic MCI began 1 to 2 years earlier than the nonamnestic MCI prodrome and was characterized by more rapid cognitive decline. These findings are consistent with neuroimaging29 and neuropathologic19 data suggesting that many people with amnestic MCI actually have mild AD.
As noted above, estimates of the temporal course of cognitive decline prior to dementia onset in AD have varied widely. The 5 to 6 years of cognitive decline preceding dementia observed in our study is close to the midpoint of comparable studies.3- 8 Several factors probably contribute to this variability. One suggested by our analyses is the extent to which symptomatic individuals are included. When we excluded people with MCI at baseline, the prodromal period was briefer, and the acceleration of cognitive decline was more marked. This may be because individuals with milder disease take longer to transition from no cognitive impairment to dementia and are thus less likely to be identified with observation periods of fewer than 2 or 3 decades. Another factor contributing to variability in findings is study differences in the effective cutoff point for dementia. Thus, in a diagnostic system that regards MCI as mild AD,30 the cognitive prodromal period is relatively brief.6 By contrast, in the study with the longest estimate of the prodromal period,5 the mean Mini-Mental State Examination score in the pre-AD group declined to below 20 before the diagnosis was made, a relatively low dementia cutoff point.
There is also evidence that prodromal cognitive decline in AD varies across cognitive domains. We found that prodromal decline in semantic memory and working memory preceded decline in other cognitive domains, consistent with findings from a previous epidemiologic study.5 However, the differences between domains in this study were not large, and other studies suggest that prodromal decline in AD begins in other domains, including episodic memory8 and visuospatial ability.6 Thus, cognition is clearly globally affected in the prodromal phase of AD, and some of the variability between cognitive outcomes probably reflects metric factors, including retest effects,31 rather than the cognitive domain being assessed. On the other hand, prodromal cognitive decline in amnestic MCI began earlier and progressed more rapidly than in nonamnestic MCI, consistent with neuropathologic32,33 and neuroimaging34,35 data suggesting that medial temporal lobe structures that support episodic memory are among the first brain regions affected by the disease.
These data show that by the time individuals meet clinical criteria for a diagnosis of AD, they have already experienced many years of accelerating cognitive decline. This has important public health implications because it is generally assumed that treatments for AD will be more effective if introduced before this prodromal period begins and cognitive systems are manifestly dysfunctional. Further, an effective early treatment could compress the cognitive morbidity of AD,36 whereas effective treatment after dementia onset might prolong it, underscoring the need for biologic and behavioral markers to aid in early diagnosis.
Little cognitive decline was evident in people who did not develop MCI or AD. This observation is consistent with prior research5,6,37,38 and suggests that cognitive decline may not be an inevitable consequence of old age.
Strengths and limitations of this study should be noted. Clinical classification of MCI, dementia, and AD was based on a uniform clinical evaluation and widely accepted criteria applied by experienced clinicians, thereby minimizing diagnostic error. The large cohort, high rate of follow-up participation, and availability of previously established psychometrically sound measures of cognition allowed us to capture subtle nonlinear changes in function. The principal limitation is that subjects are selected, and so the generalizability of the findings will need to be established.
Correspondence: Robert S. Wilson, PhD, Rush Alzheimer's Disease Center, Rush University Medical Center, 600 S Paulina Ave, Ste 1038, Chicago, IL 60612 (firstname.lastname@example.org).
Accepted for Publication: July 6, 2010.
Author Contributions:Study concept and design: Wilson and Bennett. Acquisition of data: Wilson and Bennett. Analysis and interpretation of data: Wilson, Leurgans, Boyle, and Bennett. Drafting of the manuscript: Wilson. Critical revision of the manuscript for important intellectual content: Wilson, Leurgans, Boyle, and Bennett. Statistical analysis: Leurgans. Obtained funding: Wilson and Bennett. Administrative, technical, and material support: Wilson, Boyle, and Bennett. Study supervision: Bennett.
Financial Disclosure: None reported.
Funding/Support: This study was supported by National Institute on Aging grants R01 AG17917, R01 AG024871, R01 AG15819, and P30 AG10161, and the Illinois Department of Public Health.
Role of the Sponsor: The sponsors had no role in the design or conduct of the study; in the collection management, analysis, or interpretation of the data; or in the preparation, review, or approval of the manuscript.
Additional Contributions: The authors thank the many Illinois residents who have participated in the Rush Memory and Aging Project and the many Catholic nuns, priests, and brothers who have participated in the Religious Orders Study; Traci Colvin, MPH, Julie Bach, MSW, George Hoganson, and Tracy Faulkner for coordinating the studies; Woojeong Bang, MS, for statistical programming; John Gibbons, MS, and Greg Klein for data management; and Priya Patel for preparing the manuscript.
Thank you for submitting a comment on this article. It will be reviewed by JAMA Neurology editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 44
Customize your page view by dragging & repositioning the boxes below.
Users' Guides to the Medical Literature
Users' Guides to the Medical Literature
All results at
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.