Abstract
Objectives: To characterize the incidence of AD among the elderly population of Cache County, UT, noted for its longevity and high response rates; to explore sex differences; and to examine whether AD incidence plateaus or declines in extreme old age.
Methods: Using a multistage screening process in 1998 and 1999, and reexamining 122 individuals who had been identified 3 years earlier as cognitively compromised but not demented, the authors found 185 individuals with incident dementia (123 with AD) among 3,308 participants who contributed 10,541 person-years of observation. Adjusting for nonresponse and screening sensitivity, the authors estimated the incidence of dementia and of AD for men and women in 3-year age intervals. Multivariate discrete time survival analysis was used to examine influences of age, sex, education, and genotype at APOE, as well as interactions of these factors.
Results: The incidence of both dementia and AD increased almost exponentially until ages 85 to 90, but appeared to decline after age 93 for men and 97 for women. A statistical interaction between age and the presence of two APOE-ε4 alleles indicated acceleration in onset of AD with this genotype; the interaction of age and one ε4 suggested more modest acceleration. A statistical interaction of sex and age indicated greater incidence of AD in women than in men after age 85.
Conclusions: The incidence of AD in the Cache County population increased with advancing age, but then peaked and declined among the extremely old. The presence of APOE-ε4 alleles accelerated onset of AD, but did not appreciably alter lifetime incidence apparent over a span of 100 years.
Additional material related to this article can be found on the Neurology Web site. Go to www.neurology.org and scroll down the Table of Contents for the January 22 issue to find the title link for this article.
See the Appendix on page 217 for a list of the Cache County Study Group members.
As the baby boom generation ages, more information is needed about the incidence of AD in late life, both to inform policy and to illuminate the etiology of AD. An exponential increase in the incidence of AD with age, as has been suggested by several studies,1-3⇓⇓ will have policy implications that differ from a pattern that grows less dramatically, plateaus, or even declines in extreme old age.4 An exponential increase in the incidence of AD also would suggest that the disease is an inevitable consequence of aging, whereas convergence to a fixed value or a decline could suggest that an element of the population has reduced vulnerability, owing perhaps to genetic or environmental factors. If such factors could be identified, one might develop effective prevention strategies by promoting or simulating them.
We investigated the incidence of AD extending to age 100 and beyond in the elderly population of Cache County, UT, known to contain a high proportion of very old individuals. In doing so, we examined three questions: 1) does the incidence of AD grow exponentially through the tenth decade, or does it plateau or even decline in late old age? 2) how does polymorphism at APOE modify the association between age and the incidence of AD? and 3) how does gender influence the association between age and incidence of AD?
Subjects and methods.
The sample.
We studied the permanent resident population of Cache County, Utah, who were aged 65 years or older on January 1, 1995. This population is advantageous for the study of AD for several reasons. Its longevity is exceptional, especially for males whose conditional life expectancy at age 65 is the highest in the United States5 and exceeds national averages by almost 10 years.6 The low population mortality probably reflects a lifestyle promoted by the Church of Jesus Christ of Latter-Day Saints (the religion of 91% of the sample), which prohibits intake of alcohol and tobacco—risk factors for such common killers as hypertensive or atherosclerotic cardiovascular disease and several prevalent cancers.7 The corresponding low prevalence of chronic diseases also simplifies the differential diagnosis of dementia, especially in late old age. The local culture is supportive of research, as was evidenced by an initial participation rate in 1995 to 1996 of 90%. Finally, follow-up investigations in this population are facilitated by its low rates of in- and out-migration, and by large families and a close-knit culture.
Broad approach to screening and assessment.
Identification and differential diagnosis of participants with incident dementia in 1998 to 1999 (“Wave II”) followed procedures similar to those used to identify instances of prevalent dementia in 1995 to 1996 (“Wave I”).4 We first used two screening procedures to identify participants with suspected dementia. We then asked these individuals to undergo a detailed clinical assessment, described below. Regardless of their results on prior measures, we also prescribed the same assessment for a stratified probability subsample from the Cache County population that was enriched for older participants and those with one or (especially) two APOE-ε4 alleles. Information from this fully examined subsample enabled us to estimate and adjust for the sensitivity of our screening methods in the remaining population. Because it yielded a substantial proportion of all AD cases, the subsample’s inclusion in the study design also assured an improvement in the overall sensitivity of case detection.
Screening and assessment methods.
Figure 1 portrays the screening and assessment processes. We first identified 4,614 respondents who were living without dementia at the end of the Wave I procedures. At the start of Wave II, we found that 495 of these individuals had died in the interim, and another 708 of the remaining 4,119 refused to participate or had left the area. Thus, 3,411 individuals underwent Wave II screening measures (82.8% response rate). As is indicated by parentheses in figure 1, 121 of these initial respondents (including 27 who died after screening) did not complete the subsequent assessment measures needed for a dementia diagnosis.
Figure 1. Flow chart for screening and diagnostic procedures. Study members who refused to participate, died, or moved out of the study area after the initial screening did not receive a dementia diagnosis if their score on the initial screen indicated possible dementia (the number of individuals without diagnoses are indicated in parentheses). DQ = Dementia Questionnaire.
To screen for dementia, we first administered an adaptation of the 100-point modified Mini-Mental State Examination (3MS),8 adjusting scores for sensory deficits by discarding wrong or missing answers attributed to sensory difficulty, and then calculating the percentage correct among the remaining items. Scores were also normed so that individuals with varying degrees of education obtained roughly equal mean scores.9 When subjects were unwilling or unable to participate in the 3MS, we interviewed collateral informants using Jorm’s10 Informant Questionnaire for Cognitive Decline in the Elderly (IQCODE). Participants with IQCODE scores ≥3.27 or 3MS scores <87 (84 for those aged 80 or older) were studied further with a telephone interview of knowledgeable informants using the Dementia Questionnaire (DQ).11 The DQ was also assigned for other individuals whose 3MS scores had declined by three or more points since their baseline examination, and for all individuals who were aged 90 or older. DQ were scored by a consensus of two or more clinicians as follows: 1) no impairment; 2) isolated dysmnesia or other mild difficulty; 3) moderate cognitive difficulty probably not meeting criteria for dementia (e.g., only one cognitive domain impaired); 4) questionable dementia; and 5) probable dementia. Those with DQ scores ≥3 were asked to undergo a standardized clinical assessment (CA) conducted by a specially trained research nurse and psychometrician.
Clinical assessments.
These assessments were conducted in the presence of a collateral informant at the respondent’s residence (including nursing homes). They included an interval narrative history of cognitive symptoms, an interval family history, the 1994 version of the Neuropsychiatric Inventory,12 and the Dementia Severity Rating Scale13 obtained from informants. Direct assessment of the subjects included a standardized neurologic examination and a 1-hour battery of neuropsychological tests.14 In addition, a 7-minute videotape segment recorded a brief standardized examination of mental status, insight, hand praxis, and gait.
Initial diagnostic conferences.
The results of each CA were reviewed by a neuropsychologist, a board-certified or board-eligible geriatric psychiatrist, and the research nurse and psychometrician who conducted the assessment. The psychiatrist then dictated a chronology of cognitive symptoms (if any) and the medical history (in standardized format, with date of onset for each condition when available). The psychiatrist’s note also recorded current medications, relevant family history information, the results of the nurse’s neurologic examination and the neuropsychological battery, and a modified Hachinski ischemia score.15,16⇓ We then assigned initial working diagnoses of dementia and AD as described elsewhere,4 using the Diagnostic and Statistical Manual of Mental Disorders, 3rd ed, revised (DSM-III-R) criteria17 for dementia, except that we did not insist on a demonstrable deficit in both short-term and long-term memory. AD diagnoses followed the criteria of the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS) and the Alzheimer’s Disease and Related Disorders Association (ADRDA),18 with the exception that a diagnosis of probable AD was deferred pending neuroimaging results if these were forthcoming. Vascular dementia (VaD) diagnoses followed the criteria of the National Institute of Neurological Disorders and Stroke (NINDS) and the Association Internationale pour la Recherche et l’Enseignement en Neurosciences (AIREN)19 as operationalized by Tatemichi.20 Conferees also recorded a rating on the Clinical Dementia Rating scale21 and estimated age at onset, defined as the year when the individual first suffered unambiguous dementia.
Physician examinations.
A board-certified or board-eligible geriatric psychiatrist examined 135 (83.9%) of the 161 living respondents with initial working diagnoses of dementia. At the participant’s place of residence, and without access to nurses’ notes or psychometric measures, these physicians obtained an interval medical and neuropsychiatric history, repeated the administration of the Neuropsychiatric Inventory and the standardized neurologic examination, and tested the individual with the Mini-Mental State Examination.22
Follow-up laboratory testing and neuroimaging.
As in Wave I, we sought laboratory testing and neuroimaging for individuals with working diagnoses of dementia or other substantial cognitive difficulty. Tests completed on 166 individuals (60.6% of living participants in these categories) included complete blood count, routine chemistries (CHEM-20), serum B12 and folate, thyroid function tests, and urinalysis, as well as standardized brain MRI or (in a few subjects) CT scans.
Consensus clinical diagnoses.
We reviewed all the foregoing information at expert consensus diagnostic conferences that included the study geropsychiatrists, a board-certified neurologist, a senior neuropsychologist, and a senior cognitive neuroscientist. We now sought greater specificity, assigning diagnoses from a list of more than 30 categories (many indicated in figure 2). When making these assignments, we did not consider various dementing illnesses (notably AD and VaD) as exclusive if evidence suggested both were present. Apart from dementia diagnoses, we recorded a diagnosis of prodromal AD in individuals with VaD or other disorders who had shown periods of Alzheimer-like progressive cognitive decline (with no apparent responsible vascular insults). We also applied this term to individuals with progressive mild/ambiguous cognitive syndromes14 thought probably to represent early-stage AD. Similarly, we noted incidental strokes or minor vascular changes (either a history of TIA or MRI evidence of extensive multifocal white matter disease) in individuals with primary diagnoses of AD or other neurodegenerative conditions. When raters’ opinions varied, these were discussed until we reached a consensus.
Figure 2. Differential diagnostic assignment of 185 individuals with incident dementia. The diagnoses of AD and of vascular dementia (VaD) or other specific disorders were not considered exclusive if individuals met criteria for both. Poss = possible; Prob = probable; Def = definite; Dem Unkwn Etiol = dementia of unknown etiology; TBI = traumatic brain injury; PSP = progressive supranuclear palsy.
Postmortem studies.
An autopsy program attempted to pre-enroll participants with dementia and selected others for a post-mortem neuropathologic examination. The 65 autopsy diagnoses available to date, made using criteria from the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD),23,24⇓ showed good agreement with prior clinical diagnoses (see Results). There were another five participants who received autopsy diagnoses of a dementing disorder, although we had not completed their clinical evaluation prior to death. We interviewed their informants with the DQ post mortem and considered the resulting information, along with neuropathologic data, when assigning final diagnoses.
Editing of final diagnoses.
In preparation for analyses, we conducted a final edit of diagnostic categories based upon information available from all phases of the study, including autopsy results. For example, 33 individuals who were not demented at the time of Wave I screening, but who developed dementia during later Wave I procedures, were considered here as incident cases (see figure 1). By contrast, another 15 individuals with dementia discovered at Wave II had clinical histories indicating strongly that they were demented (although not detected) at Wave I. The latter participants were categorized as prevalent cases and removed here from the denominators of at-risk individuals. Finally, we reviewed the prior diagnoses of all autopsied subjects and revised them as necessary.
Analyses.
We first compared attributes of responders vs nonresponders at successive stages of the protocol using Student’s t-test for independent samples, the χ2 test, or Fisher’s Exact Test. To estimate the sensitivity of the 3MS/IQCODE screening procedure and the DQ, we considered the 47 demented members of the fully examined subsample who had completed both screening stages. Of these, 45 had positive 3MS/IQCODE screening results (estimated sensitivity 95.7%). Two of the above 45 individuals lacked a DQ score. Of the remaining 43, 38 had DQ ratings of 3 or higher (conditional sensitivity estimate 88.4%).
We estimated the incidence of dementia and of AD within 3-year strata of age and gender after adjusting for response rates and for sensitivity of the screening measures. Completion rates through CA (0.90 for the fully examined sub-sample and 0.77 for others) were considered first. No further adjustment was needed for the subsample, but others required correction for the sensitivity of the DQ (which varied from 0.95 for the youngest age group to 0.88 for the oldest) and, among those below age 90, for the sensitivity of the 3MS/IQCODE screening procedures. Adjustment thus followed the formula:
where incidencei denotes incidence in the ith stratum; ci is the number of AD cases observed in the ith stratum of the fully examined subsample; r<90i is the number of cases in the remainder who were less than 90 years of age; r≥90i is the number of non-subsample cases who were aged 90 or older; bi is the number of AD cases in the ith stratum for whom there was no DQ score but who were nonetheless referred for clinical assessment based on their 3MS/IQCODE results; di is the stratum-specific response rate for the DQ; s3ms is the estimated sensitivity for the 3MS/IQCODE; sdq is the estimated sensitivity for the DQ; a1 is the completion rate at CA for the fully examined subsample; a2 is the completion rate through CA for others; and person-yearsi is the number of person years of observation in the ith stratum. We estimated standard errors for incidence in each stratum assuming a hypergeometric distribution for the number of cases in the stratum (from multistage sampling without replacement) and a binomial distribution for the estimated stratum incidence (considering Cache County a sample from a much larger population). We used Cochran’s adaptation of the Mantel–Haenszel procedure or Fisher’s Exact Test for estimation and evaluation of age-, sex-, and genotype-specific incidence ratios and for bivariate comparisons across strata.
Finally, we used discrete-time survival analysis25,26⇓ in multivariate models to examine the incidence of AD as a function of age and other covariates. This method divides the observation period into discrete intervals (in this instance, years) and incorporates covariates as it predicts the hazard of AD; that is, it estimates the probability of developing AD in a given year for subjects who had not developed it previously. We used a logit transformation of the hazard to make it tractable for statistical analysis and estimated coefficients for the survival models using logistic regression. A primary advantage of this method is that the functional form of the relationship between age and the incidence of AD can be modeled using quadratic or other functions that are more flexible than linear regression models; a disadvantage is that discrete-time survival analysis, like other survival methods, does not allow adjustment for factors such as response rate or screening sensitivity.
We restricted our multivariate analyses to subjects who did not have prevalent AD at Wave I, and who were not missing any covariate information (27 of 3,308 eligible respondents lacked some such information). Each term in a given model denotes a variable attribute (e.g., gender) multiplied by an estimated coefficient (e.g., βgender) that indicates the strength and direction of the attribute’s association with the onset of AD. Following the analytic strategy developed for our analyses of AD prevalence,4 we began with simple models and few terms. We then added terms sequentially, testing for improved fit of the model using the likelihood ratio (LR) χ2 statistic with degrees of freedom (df) equal to the number of new parameters added. Finally, we estimated incidence ratios for the several attributes in each model as the natural logarithm of their β coefficients. For ease of interpretation, we exponentiated these coefficients and present them as odds ratios.
Results.
Table 1 shows selected characteristics of 4,092 Cache County elderly who were nondemented at Wave I and alive at the beginning of Wave II (the table excludes 27 participants who died after screening, but before completion of prescribed assessment procedures). Compared with responders, individuals who refused the first stage of Wave II screening (nonresponders) or the subsequent procedures needed for diagnosis (noncompleters) were significantly older and less educated. Noncompleters were significantly more likely to be women. There were no differences among these groups in APOE genotype, although nonresponders and noncompleters were less likely to have provided DNA for analysis.
Selected characteristics of Cache County elders without dementia at Wave I and living at Wave II†
Table 2 shows the same characteristics of the 3,308 respondents who completed assessment procedures sufficient to allow for a diagnosis of incident dementia, categorized by such diagnosis. The table shows a higher crude incidence of dementia (χ2 = 5.58, p < 0.05) and AD (χ2 = 9.40, p < 0.01) in women than in men. As expected, those with dementia were older. They also included disproportionate numbers of individuals with one or (especially) two APOE-ε4 alleles.
Selected characteristics of 3,308 elderly included in the Wave II analysis, categorized by dementia status
Validity of final diagnoses.
Compared against a “gold standard” of 65 diagnoses that included autopsy findings, we found that prior consensus diagnoses of AD showed sensitivity, specificity, and predictive values similar to those reported from university AD clinics27 (Plassman BL, et al., manuscript in preparation). Among participants with dementia, the consensus diagnoses correctly identified 88.2% of those with an autopsy diagnosis of probable or possible AD.23,24⇓ As expected, the specificity of the consensus clinical diagnoses (proportion whose diagnosis correctly noted that they did not have AD) was lower at 53.9%, but the positive predictive value of a consensus AD diagnosis (proportion of those with such diagnoses that were confirmed by autopsy) was 83.3%.
Figure 2 shows the distribution of the final diagnoses of incident dementia. Including those with mixed diagnoses, 123 subjects (66% of those with incident dementia) received final diagnoses of AD, and 36 (19%) were given diagnoses of VaD.
Incidence estimates.
Table 3presents the estimated adjusted incidence of AD (from the formula) per thousand person-years in 3-year age strata beginning at respondents’ age 68 or less. For both men and women, the incidence of AD was lower in the oldest as compared to the penultimate age category, suggesting a decline in incidence with age among the extremely old. Also, the incidence of AD was higher in men than in women for every age group until age 78, at which point the trend was reversed. To test these patterns formally, we turned to discrete time survival analysis.
Estimated incidence (with SE) per 1,000 person-years, by sex and 3-year age groups
Discrete time survival analysis models.
Table 4 beginswith a basic Model 1 that included only the covariates age, gender, and education. These results were as expected, except that education appeared to have little or no association with incidence of AD. This result, which is consistent with our earlier analysis of prevalence data from this sample,28 may reflect either a relative homogeneity in the population’s educational attainment or an unusual disjunction in Cache County between education and healthy lifestyles or cognitive activity. Better fitting, more complex models added terms for the presence of one or two ε4 alleles (Model 2) and for age-squared (Model 3). We next tested for effects of two-way interactions (nonadditive effects) between age and the presence of ε4 alleles (Model 4). Finally, we tested for interaction between gender and age (Model 5).
Results from discrete time survival analyses in five models of increasing complexity
The fit in Models 3, 4, and 5 was substantially improved with addition of a term for age-squared. The resulting models suggested that the typical growth in age-specific incidence was attenuated in late old age until, at the oldest ages, the incidence rate peaked and then declined (see figure 3; confidence bands for these curves may be found in supplementary data by accessing the title link for this article at www.neurology.org). We used four procedures to examine whether such an apparent decline reflected the true behavior of the hazard. First, we compared the residuals (difference at each year of age between modeled and empirical incidence) of models with and without the age-squared term. We found that models with an age-squared term provided a much closer fit to the data (go to www. neurology.org and access the title link to this article for supplementary data). A statistically significant improvement in model fit after inclusion of the quadratic term was evident not only in extreme old age but also when computing the difference in log-likelihoods for respondents younger than age 93 (χ2 = 8.32, df = 1, p < 0.01). Second, we assessed the influence of extreme values and outliers and found that the models were not especially sensitive to any particular data point. Third, we reestimated the quadratic models after removing data for participants aged 93 and older. The resulting incidence estimates continued to peak and then to decline for both men and women, regardless of the presence of APOE-ε4 alleles. These results indicate that the decline in the hazard after age 93 is consistent with trends established earlier in the life span. Finally, we used several nonparametric smoothing procedures, such as robust locally weighted regression (LOWESS),29 to examine directly the shape of the empirical hazard. These curves mimicked the parametric models in showing a nonlinear relationship between age and the log of the hazard, with a peak and decline among the oldest old. Thus, it appears that, in this data set at least, the peak and decline in the hazard of AD cannot easily be explained as a methodological artifact.
Figure 3. Estimated probability by age that subjects will have incident AD, given that they have survived to that age, as predicted from discrete-time survival analysis (Model 5 from table 4). Squares represent men; circles represent women. Blackened symbols indicate no ε4; gray symbols, one ε4; open symbols, two ε4s. Data points connected by lines show values modeled from observed data. Symbols with no connecting lines show values projected (extrapolated) by the model, but for which no data are available. The graphed hazard (incidence) estimates are not corrected for nonresponse or for screening sensitivity and are, therefore, generally lower than the adjusted stratum-specific estimates of table 3.
We also considered whether the observed decline might have reflected nothing more than study clinicians’ reluctance to diagnose dementia as AD in extreme old age. Such an explanation would imply a decline in the proportion of extremely late-onset dementing illness diagnosed as AD, and a corresponding increase in the proportion of very old participants receiving other dementia diagnoses. In fact, however, no such trend was evident. The proportion of AD diagnoses among participants with incident dementia who were 93 or older was approximately 70%, no different from the corresponding proportion at earlier ages.
The interaction of age and presence of two ε4 alleles also improved the fit of Models 4 and 5. This interaction term suggests that the relationship between age and AD incidence is different for respondents with two ε4 alleles as contrasted with those having no ε4. As figure 3 suggests, those with two ε4 alleles tended to develop AD at least 10 years earlier than their counterparts without ε4. Participants with one ε4 also appeared to have accelerated onset of AD, although the acceleration was relatively much smaller and was not statistically significant.
Introduction of an interaction between age and gender further improved the model fit (Model 5; see also figure 3). This term suggests that sex differences in the incidence of AD vary with age. Figure 3 and table 3 suggest only very modest gender differences in AD incidence before age 80. After age 85, however, the hazard (incidence) of AD appeared to be markedly higher in women than in men. In fact, pooling the estimates in figure 3 across APOE genotypes suggested that the incidence of AD in men begins to decline at about age 93, but later at age 97 in women. After age 85, the estimated incidence of dementia was approximately twice as high for women as for men.
Discussion.
We found that the incidence of AD increased with age, peaked, and then declined among extremely old Cache County participants. The observed increase in the incidence of AD up to age 90–doubling approximately every 5 years–is consistent with other studies.30 Similarly, a deceleration in incidence after age 90 has also been noted previously.31 However, few studies have examined the incidence of AD among those in their mid- to late-90s and beyond, and the observed decline for both men and women is a new and provocative finding. The EURODEM project, a pooled sample of incidence surveys from Denmark, France, the Netherlands, and the United Kingdom,32 had shown a decline in the incidence of AD among elderly men, but to our knowledge this is the first study to find a (later) decline among elderly women.
Consistent with previously reported analyses of AD prevalence,4 the presence of APOE-ε4 alleles modified the association between age and incidence of AD, but appeared to exert only modest effects on lifetime incidence. We also found that gender influenced the association between age and incidence of AD. The term for interaction between gender and age in Model 5 indicates that gender differences in the incidence of AD varied with age. Specifically, the incidence appeared to be much higher in women after age 85. These results are also consistent with previous findings: although some studies with relatively few respondents older than 85 reported little gender difference in AD incidence,33-35⇓⇓ others that included substantial numbers over age 90 reported a significant interaction of gender and age.32,36⇓
The adjusted incidence estimates from Cache County are somewhat higher than results from some other samples; for example, among respondents aged 85 to 89, our estimate of AD incidence was 44.90 per thousand person-years, as compared with 38.02 in the combined EURODEM sample.37 As an explanation, we speculate that our unusual success in recruiting elderly participants and our methods of adjustment for underdetection of cases may have helped us consider incident cases that would have been nonparticipants elsewhere. In addition, a relatively high frequency of the APOE-ε4 allele among elderly Cache County residents4 could account for greater occurrence of AD.
Other unusual attributes of the Cache County population may dictate some caution, however, when interpreting these results. The population is highly educated, mostly Caucasian, relatively homogeneous, and typically adheres to a healthy lifestyle that follows the teachings of its predominant religion. Generalization of this study’s results to minorities or populations with less healthy lifestyles should therefore be avoided. Other limitations may also be noted. Although the sample size of the Cache County survey is large by historic standards, the confidence intervals for AD incidence rates at older ages remain wide, and additional studies with larger numbers of elderly would help to illuminate the incidence of AD more definitively among extremely old people. Also, differential diagnosis of dementing illnesses at late ages is a relatively unfamiliar task that is potentially subject to bias (even though we found no evidence to suggest such bias in this study).
To date, there has been limited consideration of the possibility that the incidence of AD might decline in extreme old age. If our findings can be replicated, we suggest at least two explanations. First, the decline may reflect results from a “mixed” population that includes some individuals who are relatively invulnerable to the development of AD. In these circumstances, there must be an eventual decline in incidence as the more susceptible elements of the population are depleted in very old age, while other elements will remain in the denominators of incidence calculations. Genes would seem a likely contributor to such variation in susceptibility, and the current search for “AD genes” should probably look specifically for alleles that confer relative invulnerability to the development of AD (at least within the range of usual human life expectancy), as well as others like APOE that influence timing of disease onset. Environmental factors may also contribute, however, and there are substantial current efforts to identify such factors, including neuroprotective influences.38 Among the latter are nonsteroidal antiinflammatory drugs,39 and hormone replacement therapy in postmenopausal women.40 Our findings here on gender differences in AD incidence are consistent with the last, and other analyses of the Cache County cohort (manuscripts under review) may lend further credence to this view. The coming years may therefore see promising strategies for the prevention of AD.
A second explanation for a late decline in incidence is selective censorship of individuals with risk factors for AD. Atherosclerotic cardiovascular disease (ASCVD) is thought, for example, to be such a risk factor.41 But vascular disease is also an important cause of mortality. Perhaps individuals with the strongest predisposition to ASCVD die earlier and are largely absent from the population in extreme old age. Survivors would then have reduced prevalence of ASCVD and, accordingly, a reduced incidence of AD. Because men tend to show cardiovascular mortality at earlier ages, this explanation might also account in part for the different ages at which the incidence of AD begins to decline (in the present data, at least) for men and women.
Regardless of its causes, a late decline in AD incidence should have important implications for the projected burden of disease over time. The best known projections of such a burden1 rest on the widely held view that the incidence of AD increases exponentially, doubling approximately every 5 years.30 Although that rule-of-thumb is clearly apt for ages through 85 or 90, it may provide a poorer fit to the data thereafter.31 If, as our data and others’ suggest, the incidence of AD does not continue to increase exponentially in late old age, then the future public health burden of AD, while still enormous, might be less than has previously been projected.
Appendix
Other Cache County Study investigators involved in the project: James Burke, MD; Michelle Carlson, PhD; Chris Corcoran, PhD; Marion David, PhD; Robert Green, MD; Andrea Hart, MS; Kathleen M. Hayden, MA; Michael Helms, MS; Carol Leslie, MA; Constantine Lyketsos, MD; Maria Norton, PhD; Brenda Plassman, PhD; Russell Ray, MS; Christine Reagan, MS; Ingmar Skoog, MD; David C. Steffens, MD; Martin Steinberg, MD; JoAnn T. Tschanz, PhD; Jeannette J. Townsend, MD; Kathleen A. Welsh–Bohmer, PhD; Nancy West, MS; and Michael Williams, MD.
Neuropsychological testing procedures were developed by Drs. Tschanz and Welsh–Bohmer. Dr. Tschanz reviewed all individual test results. The board-certified or board-eligible geriatric psychiatrists who examined the study members included Drs. Steinberg, Breitner, Steffens, Lyketsos, and Green. Dr. Williams also examined several subjects and provided expert neurologic consultation. Autopsy examinations were conducted by Dr. Townsend. Ms. Leslie coordinated the autopsy enrollment program. Diagnosticians at the expert consensus conferences included Drs. Burke, Breitner, Lyketsos, Plassman, Steffens, Steinberg, Tschanz, and Welsh–Bohmer. Ms. Hart, Ms. West, and Mr. Ray assisted Dr. Norton with the project’s data management. Drs. Corcoran and David, Mr. Helms, Ms. West and Ms. Reagan assisted with the statistical analyses. Drs. Carlson, Plassman, and Skoog provided expert advice with aspects of study design and interpretation. Dr. Wyse supervised all aspects of the project’s data collection and data management in Cache County.
Acknowledgments
Supported by the NIH grant AG-11380, and by MH-14592 (P.P.Z.).
Acknowledgment
The authors thank the neurogenetics laboratory of the Bryan AD Research Center at Duke University for the APOE genotyping. The authors also thank Tony Calvert, RN, Barb Gau, MSW, Tiffany Newman, BS, Cara Brewer, BS, and Joslin Werstak, BA, for expert technical assistance. Dr. Marshal Folstein suggested the 7-minute videotape procedure used at Clinical Assessments.
- Received March 14, 2001.
- Accepted September 27, 2001.
References
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
