| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Geriatric Research, Education, and Clinical Center,
b Normative Aging Study, VA Boston Healthcare System, Boston, Massachusetts
c Massachusetts Veterans Epidemiology Research and Information Center, VA Boston Healthcare System, Boston, Massachusetts
d Department of Psychiatry, Harvard Medical School, Boston, Massachusetts
e Department of Epidemiology and Biostatistics, Boston University School of Public Health, Boston, Massachusetts
Christopher B. Brady, Geriatric Research, Education, and Clinical Center (182), VA Boston Healthcare System, 150 South Huntington Ave., Boston, MA 02130 E-mail: christopher_brady{at}hms.harvard.edu.
Decision Editor: Elizabeth A. L. Stine-Morrow, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionAppendix ENDIXReferences |
|---|
RISK factors for stroke (e.g., age, hypertension, diabetes mellitus, cardiac disease) predispose even relatively healthy older adults to experience ischemic changes in the brain, predominantly in the periventricular white matter, basal ganglia, and pons prior to the development of frank cerebrovascular disease (CVD; e.g., McPherson and Cummings 1997
; Pantoni and Garcia 1997; Raz, Gunning-Dixon, Head, Dupuis, and Acker 1998). A large proportion of this white-matter pathology appears to occur in the ascending and descending pathways between the frontal cortex and the basal ganglia and brain stem, as well as in connections between frontal and other cortical regions (e.g., Chui and Willis 1999; Goldberg and Bilder 1987). Furthermore, it has been suggested that this white-matter pathology interrupts the integrity of frontal-subcortical circuits that are thought to be important to frontally mediated cognitive functions (Cummings 1993; McPherson and Cummings 1997). Therefore, stroke risk factors may have greater or specific deleterious effects on frontally mediated cognitive functions, compared with cognitive functions largely mediated by other brain regions (e.g., memory, visuospatial functions).
A commonly used behavioral index of frontally mediated cognitive function is verbal fluency. On verbal fluency tests, persons are asked to generate as many unique words as possible in a given time period (usually 1 min). For example, letter fluency requires the person to say words that begin with a particular letter (e.g., F), whereas category fluency requires the person to say words from a particular semantic category (e.g., animals). Evidence that verbal fluency indexes frontal lobe function comes from neuropsychological studies showing that persons with frontal lobe lesions exhibit deficits on this test to a relatively greater extent than do persons with lesions to other brain regions (e.g., Miceli, Caltagirone, Gainotti, Masullo, and Silveri 1981; Perret 1974; Ramier and Hecaen 1970; see however, Joanette and Goulet 1986; Martin, Loring, Meador, and Lee 1990). Additionally, neuroimaging studies of healthy individuals (e.g., Frith, Friston, Liddle, and Frackowiak 1991; Mummery, Patterson, Hodges, and Wise 1996) have consistently demonstrated physiological activation of the frontal lobes during performance of these tests; however, several studies have also noted activation occurring in other brain regions (e.g., Frith et al. 1991; Mummery et al. 1996). Although some researchers have suggested that letter fluency is more sensitive to frontal dysfunction than is category fluency (e.g., Moscovitch 1994), direct evidence of this has been equivocal (e.g., Baldo and Shimamura 1998; Gourovitch et al. 2000; Mummery et al. 1996). Regardless, there is a preponderance of converging neuropsychological and neuroimaging evidence showing that performance on both letter and category fluency tests is dependent on frontal lobe function to a greater degree than on other brain functions.
There is evidence (e.g., Lafosse et al. 1997; Starkstein et al. 1996; Wolfe, Linn, Babikian, and Albert 1990) that verbal fluency performance is disproportionately impaired relative to tests of other cognitive functions (e.g., memory) in persons with stroke-related cognitive dysfunction (e.g., vascular dementia); however, little is known about whether there is an insidious decline in verbal fluency performance prior to the onset of frank CVD as a function of increasing stroke risk. This is important to consider because such a finding would suggest that "preclinical" cognitive impairment may begin prior to the onset of CVD and, further, may suggest that a specific pattern of cognitive decline is occurring. Evidence also suggests that individual stroke risk factors such as age (e.g., Boone, Miller, Lesser, Hill, and D'Elia 1990; Whelihan 1985; see however, Parkin and Walter 1991) and diabetes mellitus (e.g., Knopman et al. 2001; Strachan, Ewing, Deary, and Frier 1997) are associated with reduced verbal fluency performance. No study, however, has used multiple stroke risk factors to address the association between increasing risk and fluency performance from a multivariate perspective (see, however, Desmond, Tatemichi, Paik, and Stern 1993, for a multivariate analysis of relations between individual stroke risk factors and other cognitive abilities).
There are, however, several studies (M. F. Elias, Elias, D'Agostino, Silbershatz, and Wolf 1998; M. F. Elias, Elias, Robbins, Wolf, and D'Agostino 2001; P. K. Elias, Elias, D'Agostino, and Cupples 1997) that have examined relations between verbal fluency performance and cardiovascular risk from a multivariate perspective. Although there are some differences between the risk-factor profiles for cardiovascular disease (Wilson et al. 1998) and stroke (Wolf, D'Agostino, Belanger, and Kannel 1991), such as the inclusion of cholesterol in the former but not the latter, the studies by Elias and colleagues used similar risk factors to those for stroke (e.g., age, hypertension, diabetes mellitus, cigarette smoking). Two of these studies (M. F. Elias et al. 2001; P. K. Elias et al. 1997) examined the association of multiple risk factors with fluency concurrently, whereas M. F. Elias and colleages (1998) used cumulative risk factors (i.e., the number of risk factors) to assess the relation between the severity of cardiovascular risk and fluency performance. Regarding this latter study, Elias and colleagues found that increasing cardiovascular risk (i.e., a greater number of risk factors) was associated with decreased verbal fluency performance, as well as decreased performance on several memory tests. Therefore, this evidence of a relation between cardiovascular risk and verbal fluency performance partially supports the possibility of a similar relation between stroke risk and verbal fluency.
These studies provide valuable information about potential relations among individual stroke risk factors and verbal fluency and suggest that increasing stroke risk may be associated with greater deficits in fluency performance. Accordingly, these results are also consistent with the suggestion, stated above, that risk factors for stroke may have greater or specific deleterious effects on frontally mediated cognitive functions compared with cognitive functions largely mediated by other brain regions.
A logical question that arises from these findings is whether people who are at greater risk for stroke exhibit a greater longitudinal decline in fluency performance relative to other cognitive functions, compared with those at lower risk; however, no longitudinal study has addressed this possibility. Therefore, the present investigation sought to examine the relation between stroke risk and decline in fluency performance by analyzing longitudinal neuropsychological and medical variables in a sample of older men participating in an ongoing study of aging. Rather than use an arbitrarily designed index of stroke risk (e.g., number of stroke risk factors), we used a risk profile from the Framingham Heart Study (Wolf et al. 1991). We were interested in whether stroke risk affected only verbal fluency performance or performance in other cognitive domains as well. We hypothesized that (a) stroke risk would be associated with decline in verbal fluency performance and (b) stroke risk would exhibit a greater relation with fluency decline than with decline in other cognitive abilities such as memory.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
). It is important to note that all participants initially recruited into the study were optimally healthy, meaning that no participant had cardiac disease, hypertension, cataracts, loss of hearing, or abnormal laboratory tests (e.g., liver function) on entering the study.
Beginning in 1993, a battery of neuropsychological tests (see Payton, Riggs, Spiro, Weiss, and Hu 1998; and Riggs, Spiro, Tucker, and Rush 1996, for a description of this battery) was administered to continuing participants at each examination to assess the effects of aging and disease on cognition. The administration of several tests was subsequently discontinued to reduce respondent burden. From April 1993 through December 1997, 1,033 (87%) of 1,184 men who reported for exams were tested; of these, 321 were tested twice at a 3-year interval. We omitted 6 men who had had a previous cerebrovascular event and selected participants with complete data on the cognitive measures described below who also had received concurrent medical examinations. This yielded a sample of 235 stroke- and dementia-free men.
Procedure
Measures.
During each visit, NAS participants were given a standardized clinical exam by a board-certified internist that included a medical history, review of systems, electrocardiogram, physical and neurological exams, and chest X-ray. The presence of dementia was determined either through the clinical judgment of the internist, the person administering the neuropsychological battery, or the report of the participant's family. Laboratory tests for fasting and postchallenge (100 g) glucose levels were conducted on blood samples drawn after an overnight fast. Blood pressure was measured with the participant seated using a standard mercury sphygmomanometer with a 14-cm cuff. Systolic (SBP) and fifth-phase diastolic blood pressure (DBP) were measured to the nearest 2-mm Hg. The average of measures taken in both arms was used.
On the basis of data gathered in the first medical exam during or after 1993, stroke risk was calculated for each participant using the Framingham stroke risk profile (FSRP; Wolf et al. 1991). Wolf and colleagues developed the FSRP on a sample of 5,734 initially stroke-free individuals from the Framingham Heart Study. This sample was followed for a decade to determine stroke incidence. On the basis of this information, Wolf and colleagues developed the FSRP, a 40-point index that quantified 10-year stroke probability based on the independent contribution of stroke risk factors that were identified from 36 years of longitudinal data collected by the Framingham study. The FSRP is used to calculate a gender-specific risk score (see Table 1 for the index for men) that is the sum of points that are assigned to a person on the basis of the following risk factors: age, SBP, antihypertensive therapy, diabetes mellitus, current cigarette smoking, cardiovascular disease, atrial fibrillation, and left ventricular hypertrophy (LVH). The more severe the risk factor (e.g., LVH), the greater the number of points associated with that risk factor. Because age is a risk factor for both stroke and cognitive decline, and, furthermore, because we were interested in evaluating the independent effects of age and medical stroke risk on cognition, we decided to consider it as a separate variable in the regression equations to better assess age effects in general. Therefore, we recalculated the FSRP omitting age (referred to below as the "modified FSRP"; mFSRP), to provide an assessment of the effects of medical stroke risk factors on cognitive decline.
|
.
To determine whether stroke risk was specifically related to verbal fluency performance rather than to cognitive performance in general, it was important to sample performance on tests of cognitive abilities besides verbal fluency. Therefore, we selected neuropsychological tests of multiple cognitive domains from the NAS battery for which longitudinal data were available; thus, tests of verbal fluency, memory, and visuospatial functions were selected. Category fluency (animal) was the verbal fluency measure. In this test, persons were given 60 s to verbally generate a list of animal names. The dependent variable was the number of unique items generated in 60 s. Digit span backward from the Wechsler Adult Intelligence Scale–Revised (WAIS-R; Wechsler 1981) and a 10-item word-list learning test with both immediate and delayed recall trials were used to measure memory function. For the digit span test, persons were given a series of single-digit numbers (e.g., 4, 2) and were asked to repeat the series in reverse order (e.g., 2, 4). The series starts at two digits and progresses to a maximum of eight. Two trials are given at each series level, and if a person correctly remembers at least one of the two trials, the next level is given. If a person missed two trials of the same series, the test was discontinued. The dependent variable was the total number of trials correctly remembered. In the word-list learning test, persons were auditorally presented with a list of 10 words for immediate recall. Three recall trials of the same word list were given, and the number of items recalled for each trial was recorded. Recall was again tested at a delay of 20 min. We used the number of items recalled from the third trial for an immediate recall measure and the delayed recall measure as dependent variables. Visuospatial function was assessed using a speeded-pattern comparison test presented by computer (Letz 1991). Participants were visually presented with two patterns side-by-side and had to make a speeded response as to whether they were identical. Their response was recorded by a manual keypress. We used the mean response latency (measured in seconds) of correct trials as the dependent variable.
Data analyses.
We examined relations between stroke risk and cognitive change by conducting a hierarchical regression analysis for each neuropsychological test. Thus, for each analysis, the dependent variable was Visit 2 test performance and the independent variables were Visit 1 test performance, age, education, and the mFSRP (entered in that order; see Appendix, Note 1). We entered the independent variables in this order because of our a priori causal assumptions regarding the effects of these variables on cognitive decline. We believed that, after controlling for baseline cognitive performance, age and education effects should be accounted for before assessing the effects of medical stroke risk because the influence of the former on cognitive performance occurred prior to the onset of stroke risk.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
Regarding the health of the 235 men, descriptive statistics for the biomedical indexes at Visit 1 were: SBP, mean (mmHg) = 137.17, standard deviation = 17.22; DBP, mean (mmHg) = 82.41, standard deviation = 9.18; fasting glucose, mean (mg/dL) = 105.20, standard deviation = 23.03; and total cholesterol, mean (mg/dL) = 231.83, standard deviation = 36.81. The percentage of the sample having the following health conditions was as follows: antihypertensive therapy, 36%; diabetes mellitus, 10%; cigarette smoking, 5%; cardiovascular disease, 14%; atrial fibrillation, 3%; and left ventricular hypertrophy, 3%. Thus, the most prevalent medical stroke risk factor was hypertension treatment, which had a similar prevalence rate to the national rate of 35% in older men (National Center for Health Statistics 1998). Furthermore, this group had low rates of both cardiovascular disease and current cigarette smoking. For comparison, the sample in Wolf and colleagues 1991 had rates of 22% and 34%, respectively. Again, this is evidence of the relatively good health of our sample.
The mean FSRP (based on the model of Wolf et al. 1991, including age) at Visit 1 was 9.03 (range 1–21), which represented a 10-year probability of stroke of 8% (based on probability tables in Wolf et al.); the average stroke risk over a similar age range from Wolf and colleagues was 10%. The mean value in the NAS sample for mFSRP, which omitted age, was 5.06 (range 0–13). Given that the mean age of the sample was 66 years, and that an age of 66 receives 4 points in the FSRP (see Table 1 ), this accounts for the 4-point difference between means of the original and modified FSRP scores.
Table 2 presents descriptive statistics for the neuropsychological tests for each visit. Significant 3-year decline was exhibited on all tests except for digit span backward. Looking at the digit span backward test score distributions, we found no evidence of a ceiling effect to account for the lack of decline over time. Digit span may have remained stable because the test was not sensitive to the memory decline that was detected by the immediate and delayed word-list recall measures. We calculated an effect-size measure for 3-year performance decline by subtracting Visit 2 from Visit 1 performance, and then divided this difference by the pooled standard deviation to express change as a percentage of the standard deviation. The values were as follows: verbal fluency, .14; immediate word-list recall, .20; delayed word-list recall, .29; and pattern comparison, -.50. The effect size for pattern comparison was negative because, given that response latency was the dependent measure, a decrease in performance over time resulted in a slower, or larger, value at Visit 2 relative to Visit 1. For the other measures, a decline in performance resulted in lower scores at Visit 2, and the resulting effect-size measure, although also representing decline in performance, was positive.
|
|
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
These results, along with those of previous studies (e.g., Boone et al. 1990; P. K. Elias et al. 1997; Knopman et al. 2001; Strachan et al. 1997; Whelihan 1985) of stroke risk factors and verbal fluency, suggest that frontally mediated cognitive functions may be especially vulnerable to the pathophysiological processes resulting from stroke risk factors. Furthermore, as stroke risk worsens and leads to the onset of CVD, frontally mediated cognitive functions such as verbal fluency may become especially impaired relative to other cognitive abilities such as memory. In fact, there is a growing consensus suggesting that certain types of CVD-related cognitive impairment such as vascular dementia (VaD) exhibit such a pattern (e.g., Looi and Sachdev 2000; McPherson and Cummings 1997), and that such a pattern of cognitive dysfunction may aid in the early differential diagnosis of VaD and dementia of the Alzheimer's type. For example, Lafosse and colleagues 1997 showed that persons with VaD exhibited poorer verbal fluency, but better memory performance, than did persons with dementia of the Alzheimer's type who were closely matched in terms of age, dementia severity, education, gender, and premorbid verbal IQ.
These results are also consistent with findings in the cognitive aging literature documenting relations between increasing age and cognitive decline over a relatively short time period in nondemented older adults. Decline in performance over 18 months to 6 years has been demonstrated in older adults for verbal fluency (e.g., Sliwinski and Buschke 1999; see however, Small, Basun, and Backman 1998), list recall (e.g., Colsher and Wallace 1991; see, however, Small et al. 1998), and visuospatial tasks (e.g., Small et al. 1998). Furthermore, our finding of no decline in digit-span-backward performance is consistent with Small and colleagues 1998, who also examined performance over a 3-year interval (see, however, Sliwinski and Buschke 1999).
Given the present results, future studies of the effects of aging on verbal fluency and other frontally mediated cognitive functions should also document the degree of stroke risk in their samples. In fact, all studies of the effects of aging on cognition would benefit from the inclusion of multiple objective measures to assess the potential effects of health on age–cognition relationships. Because some stroke risk factors can be treated, cognitive decline related to stroke risk, and furthermore, to VaD, represents an opportunity in which dementia may, at least in theory (e.g., Gorelick 1997), be prevented or its course modified through more aggressive treatment and improved compliance. Therefore, identifying a profile of cognitive deficits related to stroke risk prior to the onset of frank CVD would have profound implications for patient care, health care recommendations, and future research on cognitive aging.
|
Acknowledgments |
|---|
We thank Pantel Vokonas at the Normative Aging Study for providing the well-specified population for this study. Portions of this research were presented at the Eighth Biennial Cognitive Aging Conference in Atlanta, Georgia. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
Received for publication July 20, 2000. Accepted for publication May 31, 2001.
|
Appendix ENDIX |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
1. Please see Cohen and Cohen 1983 for a discussion of why this method for analyzing change in performance over time is preferable to other methods such as using difference, or change, scores (i.e., subtracting Visit 2 from Visit 1 performance). In short, the present method accounts for more test score variance and provides a more accurate estimate of change compared with using difference scores.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  P. Ghisletta Application of a Joint Multivariate Longitudinal-Survival Analysis to Examine the Terminal Decline Hypothesis in the Swiss Interdisciplinary Longitudinal Study on the Oldest Old J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2008; 63(3): P185 - P192. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
