Abstract
Objective: To examine the extent to which persons with mild cognitive impairment (MCI) have an increased risk of Alzheimer disease (AD) and a more rapid rate of decline in cognitive function compared to similar persons without cognitive impairment.
Method: Participants were 786 community-based persons (221 with MCI and 565 without cognitive impairment) from the Rush Memory and Aging Project, an ongoing longitudinal clinical-pathologic study of common chronic conditions of old age. All participants underwent detailed annual clinical and neuropsychological evaluations. The authors examined the risk of incident AD and rate of change in global cognitive function among persons with MCI and those without cognitive impairment; all statistical models controlled for age, sex, and education.
Results: Over an average of 2.5 years of follow-up, 57 persons with MCI (25.8%) developed AD, a rate 6.7 times higher than those without cognitive impairment. In addition, persons with MCI declined considerably more rapidly each year on a measure of global cognitive function than those without cognitive impairment.
Conclusions: Mild cognitive impairment is associated with a greatly increased risk of incident Alzheimer disease and a more rapid rate of decline in cognitive function.
Although mild cognitive impairment (MCI) is thought to represent a transitional state between normality and dementia,1–4 the extent to which persons with MCI have an increased risk of developing Alzheimer disease (AD) remains controversial. Several studies have examined rates of conversion from MCI to AD,2,3,5–13 but estimates vary widely and are difficult to interpret because few studies14–16 have included comparison groups or adjusted for age and other demographic factors related to the development of AD. Moreover, limited data are available regarding the rate of cognitive decline in MCI.2,14,17–19 We are aware of only one study14 that compared the rate of change in cognitive function in more than 100 community-based persons each with MCI and without cognitive impairment, and participants were highly educated Catholic clergy. Whether persons with MCI exhibit a more rapid rate of cognitive decline, in addition to a greater level of cognitive impairment, has been difficult to establish.
We used data from the Rush Memory and Aging Project,20 a large longitudinal clinical-pathologic investigation of common chronic conditions of old age, to compare the risk of incident AD and the rate of change in global cognitive function in more than 200 community-based persons with MCI and nearly 600 without cognitive impairment. Participants underwent detailed annual clinical and neuropsychological evaluations for up to 8 years.
Methods.
Participants were 786 individuals from the Rush Memory and Aging Project,20 an ongoing longitudinal clinical-pathologic study of common chronic conditions of old age. Study participants are residents of approximately 40 senior housing facilities in the Chicago metropolitan area, including subsidized housing facilities, retirement communities, and retirement homes. Participants in the Rush Memory and Aging Project undergo risk factor assessment, detailed annual clinical evaluations (see below), and organ donation at the time of death. The study was approved by the Institutional Review Board of Rush University Medical Center, and informed consent and an anatomic gift act were obtained from each participant following a detailed presentation of the risks and benefits associated with study participation.
At the time of these analyses, 1,047 participants had completed the baseline evaluation. Eligibility for these analyses required that persons not have dementia at baseline and complete at least one follow-up evaluation; in order to capture the heterogeneity of community-based older persons, those with common chronic conditions (e.g., history of stroke, head injury) were not excluded. Of the 1,047 participants who had completed the baseline evaluation, 65 met criteria for dementia and were excluded from this study. Of the 982 without dementia, 24 died before the first follow-up evaluation and 172 had not yet reached their first follow-up date at the time of these analyses. This resulted in a final group of 786 participants (206 men and 580 women) who completed at least 1 follow-up evaluation and an average of 3.5 clinical evaluations (SD = 1.8, range: 1 to 8) over an average of 2.5 follow-up years (SD = 1.9, range: 0 to 7). All analyses were based on this group. The mean age was 80.5 years (SD = 6.9; range: 55 to 100), the mean education was 14.6 years (SD = 3.0; range: 1 to 28), and the mean score on the Mini-Mental State Examination (MMSE)21 was 27.9 (SD = 2.1; range: 18 to 30) at baseline; 73.4% were women and 94.8% were white and non-Hispanic.
Clinical and neuropsychological evaluation.
Details of the clinical evaluation have been described previously.20 Briefly, each participant underwent a uniform structured baseline evaluation, including medical history interviews, complete neurologic evaluations, and neuropsychological examinations. Follow-up evaluations, identical to the baseline evaluation in all essential details, were conducted annually by examiners blinded to all previous data. Cognitive function was assessed at each evaluation via a battery of 21 tests, including the MMSE,21 but MMSE scores were used only to describe the cohort. Scores on 19 tests were used to create a composite measure of global cognitive function: immediate and delayed recall of story A from Logical Memory,22 immediate and delayed recall of the East Boston Story,23,24 Word List Memory, Word List Recall, Word List Recognition,25 a 15-item version of the Boston Naming Test,26 Verbal Fluency,25 a 15-item reading test,24 Digit Span Forward, Digit Span Backward,22 Digit Ordering,27 Symbol Digit Modalities Test,28 Number Comparison,29 two indices from a modified version of the Stroop Neuropsychological Screening Test,30 a 15-item version of Judgment of Line Orientation,31 and a 16-item version of Standard Progressive Matrices.32 One additional test, Complex Ideational Material,33 was used for diagnostic classification purposes only.
To compute the composite measure of global cognitive function, raw scores on each of the individual tests were converted to z-scores using the baseline mean and SD of the entire cohort, and the z-scores of all 19 tests were averaged. Computation of the composite measure of global cognitive function required that participants have valid scores on at least half of the component tests; for this study, the composite measure of global cognitive function was based on an average of 18.3 of 19 possible tests (96.3%). Further psychometric information on this composite measure is contained in previous publications.34
Clinical diagnoses.
Clinical diagnoses were performed using a three-stage process, as previously described.20 First, neuropsychological tests were administered by trained technicians, scored by a computer, and ratings of impairment were assigned based on education-adjusted cut-off scores on 11 cognitive tests commonly used in the assessment of AD.14 Second, an experienced neuropsychologist, blinded to subject age, sex, and race, reviewed the results of the cognitive testing including impairment ratings, data on education, sensory, and motor deficits, and rendered a clinical judgment regarding the presence of cognitive impairment. Third, diagnostic classification was performed by an experienced clinician blinded to all previously collected data after a review of all available data from that year's clinical evaluation, including the ratings by the neuropsychologist and the details of the neurologic examination. The clinician then specified whether the participant met clinical criteria for dementia and AD as recommended by the joint working group of the National Institute of Neurologic and Communicative Disorders and Stroke and the AD and Related Disorders Association (NINCDS/ADRDA).35 Dementia required evidence of cognitive decline and impairment in at least two domains of cognition, one of which had to be memory, for a diagnosis of AD. Participants who met these criteria and had another condition thought to be contributing to their cognitive impairment (n = 3, considered possible AD by NINCDS/ADRDA criteria) also were included in primary analyses. The diagnosis of MCI was rendered for individuals who were found to have cognitive impairment by the neuropsychologist but who, in the judgment of the examining clinician, did not meet criteria for dementia20,36; that is, such individuals were judged as not having experienced a meaningful decline from a previous level of functioning, based on the clinician's review of all available data, including the in-home cognitive assessment, educational and occupational attainment, and the neurologic examination. Although there are no well-agreed upon criteria for MCI, these criteria are identical to those used in published research on MCI in the Religious Orders Study.14,37 Persons who did not meet criteria for MCI or dementia, as determined by the clinician's review of all available data, were classified as having no cognitive impairment.
Data analysis.
First, a Cox proportional hazards model38 was used to examine the relation of MCI to the time to initial diagnosis of AD; this model included terms for age, sex, and education. To examine whether the risk of AD varied along demographic lines, we repeated this analysis with additional terms to test for potential interactions with age, sex, and education in separate models. We also repeated this analysis after excluding persons who developed other dementias or possible (vs probable) AD and who had other conditions that could affect cognitive function (e.g., stroke, head injury), so as to examine the potential influence of these persons on our findings.
Next, mixed-effects models39 controlled for age, sex, and education were used to compare the baseline level and rate of change in cognitive function among persons with MCI to those without cognitive impairment. In this approach, we estimated the mean change in each group, conditional on covariates, as in standard fixed-effects repeated measures models. In addition, the mixed-effects models included random coefficients which provided estimates of individual differences from the group. Thus, each person was assumed to follow the average path of the group except for random effects that caused the baseline level of cognition to be lower or higher and the rate of change in cognition to be faster or slower. The variance-covariance matrix for the random coefficients was not assumed to be of a restricted form and we assumed that residual error was identically normally distributed and independent of the random effects. A major strength of this approach is the ability to model all data available for each person, regardless of length of follow-up, number and spacing of evaluations, or missing data at some evaluations.
The core mixed-effects model included terms for time, time-squared, MCI, and MCI × time. The term for time indicates the mean annual rate of linear decline in the composite cognitive measure, and the quadratic term for time (i.e., time-squared) allows the rate of cognitive decline to increase with the passage of time, given that nonlinear decline can occur in older persons.19 The term for MCI indicates the mean difference in the baseline level of cognitive function between persons with MCI and those without cognitive impairment, and the term for MCI × time indicates the mean difference in the annual rate of linear decline between persons with MCI and those without cognitive impairment. In subsequent models, we added terms for the interactions of MCI and MCI × time with age, sex, and education to test whether the baseline level or rate of change in cognitive function in MCI varied along demographic lines. Programming was done in SAS,40 and model validation was carried out using analytic and graphical techniques.
Results.
Of the 786 individuals eligible for these analyses, 221 (28.1%) had MCI and 565 (71.9%) had no cognitive impairment at baseline. As shown in table 1, persons with MCI were older at baseline and performed more poorly on the MMSE and the global measure of cognition than those without cognitive impairment; the poorer performance on the MMSE and the global measure of cognition among those with MCI was expected, given that performance on these tests was used for classification of MCI at baseline.
Table 1 Sample characteristics at baseline
Risk of AD among persons with MCI vs those without cognitive impairment.
We used a Cox proportional hazards model adjusted for age, sex, and education to compare the risk of incident AD among persons with MCI to those without cognitive impairment. Over an average of 2.5 years of follow-up, 57 persons with MCI (25.8% of 221) and 23 persons without cognitive impairment developed AD (4.1% of 565); 32 persons with MCI (14.4% of 221) showed no subsequent evidence of cognitive impairment on follow-up. In the proportional hazards model, persons with MCI were 6.7 times more likely to develop AD than persons without cognitive impairment (table 2). Figure 1, based on this model, shows the substantial difference in the cumulative hazard of developing AD in the two groups.
Table 2 Relative risk of Alzheimer disease (AD) among persons with mild cognitive impairment (MCI) vs those without cognitive impairment
Figure 1. Cumulative hazard of developing Alzheimer disease among persons with mild cognitive impairment vs those without cognitive impairment.
To examine the possibility that the risk of AD may vary along demographic lines, we repeated the above analysis with additional terms to test for potential interactions with age, sex, and education in separate models. No interactions were found (data not shown).
Further, to examine the possibility that the small number of persons who developed dementia of other etiologies (n = 4) or who met inclusion but not exclusion criteria for AD (possible AD, n = 3) may have influenced these analyses, we repeated the core model first excluding those who developed dementia due to other etiologies and then excluding those who met inclusion but not exclusion criteria for AD. Results persisted and were essentially unchanged (data not shown).
Finally, to examine the possibility that persons with stroke (self-report of a history of stroke or evidence of stroke upon examination, n = 111) or head injury (self-report of head injury with loss of consciousness, n = 35) may have influenced these findings, we repeated the core model excluding persons with those conditions. Results persisted and were essentially unchanged (data not shown).
Rate of cognitive decline among persons with MCI vs those without cognitive impairment.
We next used a mixed-effects model to determine whether the increased risk of AD among persons with MCI was due to a lower level of cognitive function at baseline, a more rapid rate of decline, or both. In a model controlled for age, sex, and education, we compared the rate of change in the global measure of cognition among persons with MCI to those without cognitive impairment. This model included terms for time and time-squared; the term for time indicates the mean annual linear rate of change in cognitive function and the quadratic term for time (i.e., time-squared) indicates the mean annual nonlinear (accelerating) rate of change in cognitive function. Persons without cognitive impairment experienced a nonlinear accelerating rate of cognitive decline, as indicated by the quadratic term for time (time-squared, table 3). The global cognitive score among persons with MCI was more than a half standard unit lower than the score for those without cognitive impairment at baseline, as indicated by the term for MCI, and persons with MCI declined at an additional linear rate of 0.033 standard unit per year, as indicated by the term MCI × time. Persons with MCI also experienced a nonlinear accelerating rate of decline similar to those without cognitive impairment, but the presence of MCI did not modify the annual nonlinear rate of change (i.e., the interaction between MCI and time-squared was not significant and is not shown). Therefore, as compared to persons without cognitive impairment, persons with MCI showed an increase of more than 500% in their average rate of decline in the first year of follow-up and an increase of nearly 200% in the second year of follow-up; this difference decreased as the duration of follow-up increased. Figure 2, based on this model, shows the rates of cognitive decline among persons with MCI as compared to those without cognitive impairment for up to 7 years.
Table 3 Summary of mixed-effects model examining the relation of mild cognitive impairment (MCI) to baseline level and annual rate of change in global cognitive function
Figure 2. Rate of change in global cognitive function among persons with MCI vs those without cognitive impairment.
To examine the possibility that the rate of cognitive decline in MCI may vary along demographic lines, we repeated the above analysis with additional terms to test for potential interactions with age, sex, and education in separate models. No interactions were found (data not shown).
Discussion.
In a community-based cohort of nearly 800 individuals, we found that persons with MCI were almost seven times more likely to develop AD than comparable persons without cognitive impairment over an average of 2.5 years of follow-up. In addition, persons with MCI had markedly lower baseline scores and declined considerably more rapidly each year on a measure of global cognitive function than those without cognitive impairment.
Although several previous studies have examined the progression from MCI to AD,2,3,6–13 rates of conversion vary widely across studies, with estimates ranging from about 4% to 40% per year. Methodologic limitations such as the inclusion of small groups of highly selected patients, the lack of similar comparison groups, the use of only brief measures of cognitive function, and failure to control for demographic variables likely account for some of the inconsistencies across studies. We are aware of few large community-based studies that have examined the relative risk of AD among persons with MCI and comparable individuals without cognitive impairment14–16; these studies indicate that the relative risk of AD is increased between three- and eightfold among persons with MCI, with estimates depending in part on the length of follow-up. In the Religious Orders Study,14 the relative risk of developing AD was about three times higher over an average of nearly 5 years of follow-up among Catholic clergy members with MCI. In this study, we found the relative risk to be nearly seven times higher over an average of less than 3 years of follow-up among lay persons with MCI. It is possible that our rate of conversion was higher due to the shorter duration of follow-up; some data suggest that the risk of incident AD decreases as the duration of follow-up increases.37
We also found that persons with MCI declined considerably more rapidly each year on a composite measure of cognition than those without cognitive impairment in analyses that controlled for baseline level of cognitive function. Limited data are available regarding the rate of change in cognitive function among individuals with MCI relative to those without cognitive impairment.2,13,17,18 One study2 examined the rate of decline on the MMSE and reported an average decline of approximately two points per year among persons with MCI compared to a one-point decline per year among those without cognitive impairment. Another reported an increased rate of cognitive decline on a general psychometric factor among persons with MCI or incipient AD,17 but participants were highly selected individuals enrolled in clinic-based research studies. Our finding that community-based persons with MCI declined considerably faster than their non-impaired counterparts is consistent with findings from the Religious Orders Study14 and suggests that persons with MCI develop incident AD at a higher rate than those without cognitive impairment because they are declining faster in addition to already being closer to the threshold of a dementia diagnosis.
Little is known regarding the possibility of a nonlinear or accelerating rate of cognitive decline among persons with MCI compared to those without cognitive impairment. Prior studies have reported that dementia is characterized by a period of accelerated cognitive decline both before19,41 and after42,43 the time of clinical diagnosis. However, previous studies of persons without dementia did not separate those with MCI from those without cognitive impairment, leaving open the possibility that the observed nonlinear decline prior to the onset of dementia was strongly influenced by the inclusion of persons with MCI who already were transitioning from normality to dementia. We explicitly compared the rate of change in cognitive function among persons with MCI to those without cognitive impairment. We found that both persons with MCI and without cognitive impairment exhibited a nonlinear accelerating rate of cognitive decline. Those with MCI also exhibited an additional faster linear rate of decline.
Strengths of this study include the large, well-characterized cohort free of dementia and the availability of a single cohort of community-based persons with and without cognitive impairment, which increases the likelihood that the comparison groups are truly comparable. In addition, the availability of multiple observations per individual, the high rate of follow-up participation, and the use of a composite cognitive outcome measure that accommodated a wide range of function increased our ability to detect nonlinearity in cognitive decline. Limitations include the use of a selected cohort and the relatively short duration of follow-up; it is possible that, with longer follow-up, we may have found that the presence of MCI modified the nonlinear rate of decline in cognitive function. In addition, while the use of a single cognitive battery for diagnostic classification and examination of cognitive decline may be considered a limitation, the use of two batteries would have markedly increased the burden on study participants and possibly impacted overall rates of follow-up; loss of follow-up presents a considerable threat to the validity of longitudinal studies. Moreover, some have questioned the existence of MCI as a distinct condition and suggested that, at least for a subset of persons, MCI represents the earliest stages of AD.44 Our finding that many persons with MCI subsequently developed AD is consistent with this notion. Finally, although the criteria used in this study have been used in several publications,14,36,37 there is a lack of widespread agreement regarding MCI criteria and their operationalization in community-based studies. While some studies require memory impairment for MCI classification, others use more inclusive criteria such as those employed here, and the application of different criteria can influence study findings. One concern about the use of restrictive criteria is that they may prematurely limit investigation of the full spectrum of factors that contribute to cognitive impairment in aging; evidence from community-based studies of MCI suggests that restrictive criteria yield low prevalence estimates and may reduce applicability of findings.16,45 Importantly, more inclusive criteria allow for investigations of the many factors (e.g., stroke, head injury) associated with cognitive impairment and ultimately may provide a more thorough foundation of knowledge from which narrower definitions can be approached and studied. Extension of these findings to longitudinal, population-based studies is needed to better estimate the risk of AD and rate of cognitive decline in persons with MCI vs those without cognitive impairment.
Footnotes
-
Disclosure: The authors report no conflicts of interest.
Received October 7, 2005. Accepted in final form April 4, 2006.
References
Disputes & Debates: Rapid online correspondence
- Reply from the Authors
- Mild cognitive impairment: Risk of Alzheimer disease and rate of cognitive decline
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.