| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
4 and Risk of Mortality in African American and White Older Community Residents
a Center for the Study of Aging and Human Development, Duke University Medical Center, Durham, NC
b Department of Psychiatry and Behavioral Sciences, Center for the Study of Aging and Human Development, Duke University Medical Center, Durham, NC
c Division of Neurology, Department of Medicine, Duke University Medical Center, Durham, NC
d Center for Health Policy/Law and Management, Duke University Medical Center, Durham, NC
Correspondence: Gerda G. Fillenbaum, PhD, Center for the Study of Aging and Human Development, Duke University Medical Center, Box 3003, Durham, NC 27710. E-mail: ggf{at}geri.duke.edu.
Decision Editor: Laurence G. Branch, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
4 allele of apolipoprotein E is predictive of mortality in a community-based sample. Design and Methods: Of the stratified random household sample of 4,162 participants age 65 years and older enrolled in the Duke site of the Established Populations for Epidemiologic Studies of the Elderly, those included in the present study were the 1,998 who were genotyped for apolipoprotein E (alleles 2, 3, and 4) six years after baseline, and for whom survival status eight years later was known by search of the National Death Index. Information on demographic characteristics, physical and mental health status, functional status, and health services use was determined by structured questionnaires administered in person in the home. Results: The 4 allele did not predict mortality for the group as a whole, or for those who were cognitively impaired. It did predict mortality for those who reported having had a heart attack or stroke. Implications: The apolipoprotein 4 allele—although a risk factor for Alzheimer's disease, heart disease, and stroke—was only found to be a risk factor for mortality for those community residents who had had a heart attack or stroke. Otherwise, for this community-based sample, 71 years of age and older, it did not predict time to death and was not a risk factor for mortality.
Key Words: Apolipoprotein E • Mortality • Race • African American • White
Apolipoprotein E (APOE) has three allelic forms: 2, 3, and 4. Of these, 4 (APOE4) is recognized as a susceptibility gene for Alzheimer's disease (AD; Corder et al. 1993
; Saunders et al. 1996; Strittmatter et al. 1993), the most common of the dementing disorders. 4 is also recognized as a risk factor for coronary heart disease (Laakso et al. 1991; Stengard, Weiss, and Sing 1998; van Bockxmeer and Mamotte 1992; Wilson, Schaefer, Larson, and Ordovas 1996), as well as increasing the risk for stroke (McCarron, Delong, and Alberts 1999). These conditions are among the most prevalent causes of death in the population 65 years of age and older. Diseases of the heart were listed as the leading cause of death in 1980 and 1998, whereas AD was listed as the ninth leading cause of death (National Center for Health Statistics [NCHS], 2000, Table 33), although it may be as high as third (Ewbank 1999).
Given that APOE4 is implicated in two of the most common causes of death, and that the prevalence of APOE4 appears to be underrepresented in those of a very old age (see review in Smith 2000), the question naturally arises as to whether APOE4 is a risk factor for all-cause mortality. To examine this issue, we have analyzed data from the Duke site of the Established Populations for Epidemiologic Studies of the Elderly project (Cornoni-Huntley et al. 1990), a representative community-based sample of African American and White participants who were 65 years of age and older at the start of the study. These participants were interviewed annually for 6 years, with a final in-person interview 10 years after the study began. Information on the presence of the 4 allele was obtained six years after the start of the study. We examine survival for the following eight years to determine if those with the 4 allele are more likely to die within this time interval or sooner than those in whom this allele is absent.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; Cornoni-Huntley et al. 1990). A four-stage stratified random household selection procedure, carried out in a five-county area (one urban county and four rural counties), resulted in the selection of 5,223 community-residing persons 65 years of age and older, of whom 4,162 (80%) agreed to participate. To increase statistical precision, African Americans were oversampled and represented 54% of the participants, compared with their actual representation in the geographical area of 35%. Only 26 of the participants were neither African American nor White. They have been dropped from the present analysis. This study was performed with permission of the Institutional Review Board of Duke University Medical Center. Informed consent was obtained from all participants. In 1986–87, 1989–90, 1992–93, and 1996–97, data were gathered in person in the home by trained interviewers using a structured questionnaire. Information was gathered annually by telephone in each of the two intermediate years between the first and second in-person interviews, and the second and third in-person interviews.
Main Independent Variable
The main independent variable is the presence of the 4 allele. At the third in-person interview, blood was drawn, and, among other characteristics, APOE genotype was determined (details have been published in Blazer, Fillenbaum, and Burchett 2001). Of the 2,550 surviving participants, 1,999 were successfully genotyped at this time (one participant was dropped because of incomplete information on date of death). Of the remaining survivors, seven were ineligible on medical grounds, 222 were unable to sign the necessary consent form because of cognitive impairment, 23 had moved outside the personal contact area or had died by the time of blood draw, and the remainder did not consent to the drawing of blood.
Dependent Variables
Survival status of all participants was determined through a search of the National Death Index, which identified deaths and dates of death through December 1999. Of the group of 1,998 genotyped participants present at the third in-person interview (1992–93) for whom date of death was known, 486 (24%) were identified as having died by the end of December 1999. For all study members, duration of survival (days) through December 1999 was calculated.
Covariates
Relevant to examination of all-cause mortality, we selected from the information obtained at each interview data on five health conditions (hypertension, heart attack, stroke, diabetes, and cancer). These conditions were summarized to indicate weighted medical status. The resultant value was dichotomized to indicate better (coded 0) versus poorer (coded 1) medical status (Fillenbaum, Leiss, Pieper, and Cohen 1998). We also selected self-assessed health status (dichotomized as excellent /good [coded 1] vs. fair/poor [0]) (Idler and Benyamini 1997; Wolinsky, Callahan, and Johnson 1994). The items of the available functional status scales (Branch et al. 1984; Fillenbaum 1988; Katz and Akpom 1976) were regrouped into three categories: basic activities of daily living (ADL) (toileting, dressing, eating, transferring, grooming), intermediate ADL (bathing, using transportation, housework, meal preparation, shopping, traveling), and complex ADL (handling money, using the telephone, taking medications) (Thomas, Rockwood, and McDowell 1998). In each of the three categories, the items were summed and recoded to indicate whether a problem was present (coded 1) or not (coded 0). Similarly, the three items of the Rosow-Breslau (1966) (walking half a mile, going up/down stairs, doing heavy housework) were summed and recoded to indicate if a problem was present or not. We also included level of cognitive functioning (Pfeiffer 1975), dichotomized at 0–3 versus 4–10 errors (coded as 0 and 1, respectively); depressive symptomatology as assessed by the Center for Epidemiologic Studies–Depression scale (Radloff 1977), dichotomized as 0–8 versus 9–20 responses indicative of depressive symptomatology; and self-assessed life satisfaction (excellent/good [coded 1] vs. fair/poor [coded 0]). Information on use of health services was also selected (Wolinsky, Stump, and Johnson 1995); specifically, if the participant had a usual health care provider, number of outpatient visits (dichotomized as 0–4 vs. 5+) and emergency room visits in the previous 12 months (yes = 1, no = 0), if hospitalized overnight in the past year (yes, no), and number of prescription and over-the-counter drugs being taken. Because age, gender, race, and socioeconomic status are predictors of mortality (Drever, Whitehead, and Roden 1996; Howard, Anderson, Russell, Howard, and Burke 2000; Kitagawa and Hauser 1973), we included age (in years), sex (male [coded 0], female [coded 1]), race (White [coded 0], African American [coded 1]), years of education (0–8 years [coded 0], 9+ years [coded 1]), income (<$10,000 vs. $10,000+), if married (yes = 1, no = 0), and residence in an urban or rural area (0 vs. 1).
Statistical Analysis
For the independent variables and for all covariates, we first ran descriptive statistics to determine means and standard deviations, or proportions, as appropriate. For each variable, we calculated unadjusted hazard ratios, 95% confidence intervals (CIs), and level of significance for time to death using survival analysis. Cox proportional hazards analyses, which included age at the genotyping interview and survival status, were then run to determine if, after controlling for covariates, APOE4 made a significant independent contribution to predicting time to death. The censoring date was December 31, 1999; we had no information on participants' subsequent status if they were alive on that date. In these analyses, we first grouped variables into domains (demographic, physical health status, mental health status, functional status, health services use) and treated the set of variables in each domain as a unit. Time to death, or censoring, was regressed on each domain separately. If the domain was a significant predictor, we looked at the constituent variables and selected for the final model those variables that were significant. Although race was not a significant predictor in the demographic domain, it was included in the final model. We also examined the interaction of each variable with 4 in the final model. In addition, we plotted Kaplan-Meier curves to determine whether time to death, in uncontrolled analysis, distinguished those in whom the 4 allele was present from those in whom it was absent. Finally, we ran a controlled logistic regression analysis to determine whether APOE4 predicted death by December 31, 1999.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
), and 16% were cognitively impaired. The proportion with functional status problems ranged from 13% to 52%, depending on the type of area assessed. Nearly all participants had a regular health care provider. One third of the group reported five or more outpatient visits in the previous year, whereas about one fifth had had an emergency room visit or had been hospitalized in the same time interval. The average number of prescription and over-the-counter medications taken was 2.4 and 1.3, respectively. The 4 allele was present in 32%.
|
In unadjusted analyses, the majority of the individual demographic characteristics, and measures of physical, mental, and functional status and health services use, were predictive of death. The only variables not predictive of death were race, area of residence (rural or urban), marital status, consistently using the same health care provider, and the presence of the 4 allele. A Kaplan-Meier curve showed that survival was essentially identical for those with and those without the 4 allele.
Analysis indicated that each of the domains (demographic, physical health, mental health, functional status, health services use) was a statistically significant predictor of death. All significant variables in each chunk were included in the final model (Table 2 ), together with race, which had not reached statistical significance.
|
4 allele were not more likely to be dead by the censor date. This conclusion was unchanged when we used logistic regression to model mortality. Analyses in which the proxy respondents were dropped, so that information on items asked only of sample members could be included (self-rated health, depressive symptomatology, life satisfaction), showed similar findings. Examination of the interaction of APOE4 with each of the variables entered indicated that they were not significantly predictive of mortality, either as a set or individually.
To further explore the relationship of APOE4 to survival, we looked specifically at those conditions for which APOE4 has been shown to be a risk factor: heart disease and stroke, and cognitive impairment. Using the same procedures as before, we found that, among participants who said that their doctor had told them that they had had a heart attack or stroke (or who had reported both of these conditions) (n = 562), the presence of the 4 allele was a risk factor for mortality (odds ratio = 1.64, 95% CI = 1.06, 2.54; p = .0279). Among those with poorer cognitive functioning (n = 305), the presence of the 4 allele was never a significant risk factor for death. For both groups, test of the entire group of interaction terms showed that, as a group, they were not significant; consequently, they were not considered further.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
4 allele is not a risk factor for death or predictive of time to death within the following eight years. Neither is it a risk factor in analyses where adjustments have been made for characteristics consistently predictive of death: demographic variables, such as age, gender, and marital status (Howard et al. 2000; Kitagawa and Hauser 1973); health measures, such as medical status (NCHS 2000), life satisfaction (Idler and Benyamini 1997), and cognitive impairment (Corder et al. 1995, Corder et al. 1996); and ADL (Fillenbaum 1985).
Although studies tend to agree that the prevalence of APOE4 decreases with age, suggesting that APOE4 may be a risk factor for survival (Smith 2000), there is so far little consistent evidence for this. Tilvis, Strandberg, and Juva 1998, using a representative sample of 550 persons (75, 80, and 85 years of age who were followed for five years) found that after controlling for age and gender only, the hazard ratio for all-cause mortality was 1.61 (95% CI = 1.21–2.14). The hazard ratio increased for those diagnosed as demented and further increased for those with AD (2.20, 95% CI = 1.03–4.72; and 3.24, 95% CI = 1.67–6.25, respectively). Unlike other studies, the prevalence of APOE4 did not decline with age. Other studies using population-representative or community-residing groups were unable to replicate these findings for either their samples as a whole, or for those persons who were cognitively impaired (Corder et al. 1996; Lee et al. 2001), or for those who were diagnosed with AD (Koivisto et al. 2000; Stern et al. 1997).
To further explore the presence of APOE4 as a risk factor for death, we examined the effect of age, because in studies of AD, for which 4 has been recognized as a susceptibility gene, the risk of developing the disease decreases as age increases (Farrer et al. 1997). In fact, APOE4 may no longer be a risk factor beyond a certain advanced age (Corder et al. 1993). The current sample has a minimum age of 71 years. No increased risk of mortality was found among participants with the 4 allele as age increased (data not shown). If the 4 allele is a risk factor for mortality, it is possible that its effect is evident only at an age younger than that encompassed by the present sample and those reported on to date.
To explore the possibility that the mortality-related effect of the 4 allele may apply only to conditions for which it is known to be a risk factor, we examined its impact on a group of people who reported that they had been told by their physician that they had had a heart attack or a stroke (or both), and on those with poor cognitive functioning. 4 was predictive of mortality in the former group. No association was found in the group with poor cognitive function, confirming findings by Corder and colleagues 1996.
Our findings suggest that 4 is not predictive of time to death or a risk factor for mortality within the next eight years in a community-based sample of participants 71 years of age and older. Neither, and in agreement with other studies, does it appear to be a risk factor for the cognitively impaired, although in the present study it does seem to be a risk factor for those who report having had a heart attack or stroke.
There are certain limitations to the present study. Perhaps most important, all sample members could not be genotyped. Disproportionately missing from the present study are persons who are older, sicker, and more likely to be cognitively impaired because they could not personally give informed consent. Given the association between cognitive impairment and the 4 allele (Fillenbaum et al. 2001), there may have been a disproportionate loss of persons in whom 4 is present. As should be expected, the odds of death were significantly greater in those who were not genotyped (unadjusted odds ratio = 7.69, p < .0001). We may, therefore, be underestimating the relation of 4 to mortality. We tried to explore this issue by alternately assuming that the missing sample members did, or did not, have the 4 allele. Neither approach made a difference to the findings; there are, however, notable problems with such imputation strategies. Lack of genotype data for a substantial proportion of study participants is a common problem in this area, but not one for which a resolution seems to have been proposed.
Our participants are drawn from the Piedmont section of North Carolina and so may not be representative of other areas. In particular, health care may be more available in this area than in some others and so may influence survival in the present sample. One predictor of survival may be unexpected: whereas nationwide longevity is briefer for African Americans as a whole than for Whites (NCHS 2000), we found that African Americans were less likely to die than were Whites. This may reflect the cross-over effect reported previously (Kestenbaum 1992), for our sample is elderly. No interaction was found between race and 4.
Perhaps the most important limitation of the present study is the age structure of this sample. It is possible that, as with AD, the 4 allele is a risk factor at younger ages than those examined here. To examine the impact of APOE4 on survival, we may need to look at younger people. Those who live to the age of 70 are survivors, for whom the presence of the 4 allele does not appear to have an important effect on survival status over the next eight years. Survival status is not immutable. To continue to increase longevity, we need to continue to address the obvious health concerns of the elderly.
Practice Concepts
The Forum
Book Reviews
|
Acknowledgments |
|---|
Received for publication July 26, 2001. Accepted for publication November 20, 2001.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A. Kulminski, S. V. Ukraintseva, K. G. Arbeev, K. G. Manton, J. Oshima, G. M. Martin, and A. I. Yashin Association between APOE {varepsilon}2/{varepsilon}3/{varepsilon}4 polymorphism and disability severity in a national long-term care survey sample Age Ageing, May 1, 2008; 37(3): 288 - 293. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. C. Ewbank Differences in the Association Between Apolipoprotein E Genotype and Mortality Across Populations J. Gerontol. A Biol. Sci. Med. Sci., August 1, 2007; 62(8): 899 - 907. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. C. Hays, B. M. Burchett, G. G. Fillenbaum, and D. G. Blazer Is the APOE {epsilon}4 Allele a Risk to Person-Environment Fit? Journal of Applied Gerontology, September 1, 2004; 23(3): 247 - 265. [Abstract] [PDF]
|
|
|
|
|
|
D. C. Ewbank The APOE Gene and Differences in Life Expectancy in Europe J. Gerontol. A Biol. Sci. Med. Sci., January 1, 2004; 59(1): B16 - 20. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
