- Email: [email protected]
Survival and Mortality in Alzheimer’s Disease
Chapter 8
Survival and Mortality in Alzheimer’s Disease As early as 1976, Dr Robert Katzman noted that the late-onset form of Alzheimer’s disease (AD) “may rank as the fourth or fifth most common cause of death in the US … yet the U.S. vital statistics tables do not list ‘Alzheimer disease,’ ‘senile dementia,’ or ‘senility’ as a cause of death, even in the extended list of 263 causes of death” (Katzman, 1976). The death certificates in the United States and many other developed countries have been poor sources of information regarding death from AD. Dementia and AD cause many deaths indirectly. For example, death may result from aspiration pneumonia in advanced AD patients who are unable to eat or decubitus ulcers among patients who are bedbound. It is therefore important that all causes of death—underlying, immediate, associated or contributing—be analyzed on the death certificates. Even then, the death certificate is a poor source of information due to the frequent lack of information available to the physician completing it. In the first study of US mortality data taking into account all causes of death, Chandra, Bharucha, and Schoenberg (1986) examined causes of death in 1971 and 1973–1978 using International Classification of Diseases Adapted for Use in the United States (ICDA) (US Department of Health Education and Welfare, 1967) rubrics for “senile or presenile brain disease” (ICDA rubric 290) and “senility” (ICDA rubric 794). Data were analyzed separately for deaths due to dementia and those with dementia. The overall annual age-adjusted mortality rate was 0.41 and 2.19/100,000 for mortality due to and with senile and presenile dementia, respectively, and 0.52 and 12.21/100,000 for mortality due to and with senility, respectively. In this analysis, rates were higher among men and for whites for deaths due to and with senile and presenile dementia, and for nonwhites, the rates were higher for deaths due to and with senility. Among those who died with senile and presenile dementia, the most common underlying causes of death were coded as diseases of the heart (33.4%), senile and presenile dementia (18.2%), cerebrovascular diseases (12.9%), and influenza and pneumonia (4.5%). For those dying with senility, the top five underlying
Alzheimer’s Disease. DOI: http://dx.doi.org/10.1016/B978-0-12-804538-1.00008-0 © 2016 2013 Elsevier Inc. All rights reserved.
89
90 SECTION | II Descriptive Epidemiology
causes of death were diseases of the heart (41.7%), cerebrovascular diseases (14.6%), arteriosclerosis (9.3%), influenza and pneumonia (7.9%), and malignant neoplasms (4.4%). Only 18.2% of the death certificates listed senile and presenile dementia as the underlying cause of death, and 4.3% listed senility as the underlying cause among those who had died with these conditions (Chandra et al., 1986). In a study using NCHS mortality data, age-adjusted annual mortality rates per 100,000 for AD for ages 65 and over increased from 2.5 in 1979 to 37.5 in 1995 (Hoyert & Rosenberg, 1997). This large increase likely reflects changing codes and diagnostic criteria, as well as increased awareness of the disease. The study by Hoyert and Rosenberg reported that 20,000 death certificates had AD as the underlying cause of death and another 20,000 as another cause of death. In a commentary, White (1997) estimated a conservative number of 100,000 deaths in the United States around this time from end-stage dementia, based on a prevalence of AD at the time between one and two million. Therefore, it would appear that in only 40% of cases were dementia diagnoses included on death certificates. Other studies have produced higher estimates. Among patients enrolled in studies reporting their findings to the Consortium to Establish a Registry for Alzheimer’s disease (CERAD) (Raiford, Anton-Johnson, Haycox, et al., 1994), dementia was reported on about two-thirds of death certificates. It must be kept in mind that this group of patients was highly selected and that participation in a research study on dementia could increase the likelihood of dementia appearing on the death certificate. In a UK study (Martyn & Pippard, 1988), only 22% of demented patients had an ICD code of presenile or senile dementia on their death certificate. However, 58% of these death certificates mentioned the word “dementia” somewhere on the certificate. Frecker, PrysePhillips, and Strong (1995) reported their experience from Newfoundland, Canada, where they noted that 51/59 (86.2%) death certificates for probable AD cases by NINCDS-ADRDA criteria listed AD on the certificate. The place of death may make a difference in the degree of accuracy, because, for example, the clinical histories of patients dying in hospital are less known to physicians than those dying in nursing homes. Data derived from the Chicago Health and Aging Project (CHAP) and extrapolated to the US population (Weuve, Hebert, Scherr, & Evans, 2014) estimated that 600,000 people with AD died in 2010 in the United States, including people who received medical care and those who did not. By 2030, these estimates are expected to increase to 900,000, and in 2050, to 1.6 million. In addition, due to the aging of the US population, the proportion of deaths with AD will increase rapidly. By 2050, 43% of all older adult deaths will occur among people with AD, up from 32% in 2010 (Weuve et al., 2014). While there is a trend for death certificates to increasingly include AD as a cause, there is consensus that only about half of existing AD cases are ever diagnosed (Thies & Bleiler, 2011).
Survival and Mortality in Alzheimer’s Disease Chapter | 8 91
AD IN THE TOP 10 LEADING CAUSES OF DEATH In 2010, the top 10 leading causes of death among people age 65 and over listed AD as sixth, with 83,494 deaths (Murphy, Xu, & Kochanek, 2013). If there truly are 600,000 deaths per year from AD, AD would be regarded as the #1 cause of death in the United States, instead of heart disease (which, in 2010, was estimated to claim 477,338 lives). It is important to note that in national statistics, the numbers of deaths reported for AD are those for which the underlying cause of death is AD. Many patients with AD die from heart disease, stroke, pneumonia, and influenza, which would frequently be considered as the underlying causes of death. The degree to which AD contributes to such deaths is unknown. For example, would the person have died from stroke if they didn’t have AD? This issue highlights the difficulty in pinpointing the individual contributions of various chronic diseases as well as acute illnesses, such as influenza, in causing death in the oldest segment of our population. Given that about 50% of cases of dementia are not diagnosed before death, screening for incident dementia or AD in community-based studies and following cases over time is necessary to accurately enumerate deaths and to appreciate the importance of AD and dementia in increasing the risk for death. It also is important to start with incident cases, so that the time between onset and death can be counted properly. In addition, studies need to be relatively large so that there is a sufficient number of incident and mortality outcomes. Such data are available from a number of studies, and a few are selected here for illustrative purposes. In the PAQUID study, begun in 1988, a cohort of 2,923 community-based elders age 65 and over living in southwestern France (Gironde and Dordogne) were randomly sampled from electoral rolls and followed for up to 8 years (Helmer, Joly, Letenneur, Commenges, & Dartigues, 2001). During this time, 281 incident cases were diagnosed with dementia (67% AD), with a mean age at onset of 80 ± 5.9 among men and 83.5 ± 6.1 among women. Mortality was analyzed using Cox proportional hazards models with delayed entry, adjusting for sex, education, comorbidities, and dependence in ADL/IADL at baseline. In the population, the RR for mortality associated with incident dementia was 1.80 (95% CI = 1.46–2.21), and about the same, 1.72 (95% CI = 1.34–2.21) for AD. The analysis also showed effect-modification by age, such that dementia had less of an impact on mortality with increasing age (RR= 0.95, 95% CI = 0.92–0.98 per year). For participants dying with dementia above age 85, the RR was elevated but not statistically significant (RR = 1.37, 95% CI = 0.91–2.04). In this study, there were no sex differences in the RR for dying with dementia. There was an increased risk for dying with respiratory disease among individuals with AD; and the risk for dying with cerebrovascular disease was statistically significant for dementia, but not for AD. The median survival time from onset of dementia to death was 4.5 years for all ages, being longer among women (7.3 years at age 75 compared with 4.5 for men; and 4.4 at age 85 compared with
92 SECTION | II Descriptive Epidemiology
3.5 for men), likely reflecting the survival advantage of women over men in the general population. Survival time also was modified by education, with those with more education having shorter survival times than those with less education (Amieva et al., 2010). Modification of survival time by education was also reported in the WHICAP cohort. The RR for dying among AD patients was 1.9 (95% CI = 1.2–3.0) for those with 9 or more years of education in comparison with those with 8 or fewer years. Those with higher occupational attainment also experienced a higher risk for mortality compared with those with lower occupational attainment (RR = 2.2, 95% CI = 1.1–4.3) (Stern, Tang, Denaro, & Mayeux, 1995). This would be predicted by the threshold hypothesis, since AD patients with higher education and occupation would experience a later clinical onset of their disease, when AD-related neuropathology was more severe given their higher brain reserve. Such individuals would be expected to decline more rapidly and have a shorter duration of disease from onset to death (Stern, Alexander, Prohovnik, & Mayeux, 1992). The WHICAP study also found a differential survival in AD by race/ ethnicity (Helzner et al., 2008). Over a mean follow-up time of about 4 years, 323 incident AD cases developed, who were on average 87 years old at baseline and who had a median lifespan of 92.3 years. The median survival after AD onset among non-Hispanic whites was 3.7 years, 4.8 years among AfricanAmericans, and 7.6 years among Hispanics. The findings were not explained by differences in overall lifespan by race/ethnicity. However, these differences could be explained by differences in education and occupational attainment, with the African-Americans and Hispanics being diagnosable at earlier ages. Another way to look at mortality is to compare remaining life expectancy for AD patients in comparison to the general population. In the Adult Changes in Thought Study (ACT Study), an AD patient registry was set up in 1987 in the Seattle, WA regional area to identify all persons age 60 and over in the population of a health-maintenance organization of 23,000 members who developed AD (Larson et al., 2004). Between 1987 and 1996, incident cases were invited to participate in the study, and 970 with suspected dementia (521 AD: 431 probable and 90 possible by NINCDS-ADRDA criteria) were followed annually. Survival in AD in this sample did not vary by education and was longer for younger-onset patients and women. Poorer survival with AD was predicted by male sex, lower baseline cognitive score, poorer Dementia Rating Scale score, and by the existence of frontal release and extrapyramidal signs, history of falls, incontinence and history of heart disease and stroke. Survival time medians from diagnosis to death were 7.5 for AD cases ≤75 years old, 5.6 for those 76–80, 4.9 for those 81–85 years old, and 3.2 for those over age 85. A 70-year old woman with AD was expected to live 7.7 fewer years than a woman in the general US population, and a similarly-aged man, 8 fewer years. This gap narrowed with increasing age (3.3 years for an 80-year old woman and 3.1 years for an 80-year old man, and 1.8 and 0.5 years at age 90 for women and men,
Survival and Mortality in Alzheimer’s Disease Chapter | 8 93
respectively). The Hazard Ratios for mortality by MMSE score, using scores of 25–30 as the reference group, were 1.4 (95% CI = 1.01–1.94) for MMSE scores 22–24, 1.69 (95% CI = 1.24–2.31) for MMSE scores 18–21, and 2.67 (95% CI = 1.94–3.66) for MMSE scores ≤17. Two cohort studies that required autopsy consent for enrollment are informative for mortality. Both studies, the Religious Orders Study (ROS) and the Rush Memory and Aging Project (MAP), enrolled nondemented participants. In the ROS, the 1,168 subjects were Catholic nuns, priests, and brothers aged 65 and older from around the United States (Bennett, Schneider, Arvanitakis, & Wilson, 2012). The analysis of mortality in this cohort was conducted between 1994 and 2013 (James et al., 2014). In the Rush MAP, 1,574 people living in retirement communities and senior housing around Illinois (Bennett, Schneider, Buchman, et al., 2012) were enrolled between 1997 and 2013. The studies were conducted by the same research team, all measures were standardized, and the authors pooled the data assuming the populations were comparable. There were a total of 2,566 nondemented subjects 78 years old at baseline who were followed for an average of 8 years, 559 of whom developed incident AD over the study follow-up (mean age of AD diagnosis 86.5 ± 6.5). Six hundred and ninety-two (34.5%) of the nondemented participants died, versus 398 (71.2%) of those with incident AD. The median survival time from diagnosis of AD to death was 3.8 years and depended on the age of the subject at diagnosis: 4.4 years for those age 75–84 and 3.2 for those age 85 and over. After adjusting for age, sex, race, education, and study in the proportional hazards regression model, the HR for mortality associated with incident AD was 3.13 (95% CI = 2.74–3.58). The authors calculated the population attributable fraction (PAF) for mortality associated with AD, meaning the proportion of deaths in this population that were attributable to AD. These estimates were 37% for participants between the ages of 75–84 after adjustments for demographic variables, and 36% for those age 85 and older. When these PAFs were extrapolated to the number of deaths in the United States, the authors estimated that there were 503,400 excess deaths in 2010 associated with AD. This estimate is very close to the estimate of 600,000 obtain by Weuve et al. (2014). Importantly, the authors of the ROS and Rush MAP point out that their estimate of the number of people dying because of AD is 5–6 times higher than that acknowledged by the CDC in 2010 (83,494 deaths). This study has been criticized because no individual data were used to investigate the actual causes of death of cohort members and because the samples were not population-based. In contrast to the PAF published by the Rush group (James et al., 2014), the Alzheimer’s Association report for 2013 states that the PAF for AD on mortality over a 5-year period in people age 65 and over ranges from 5% to 15% (Gaugler et al., 2013). This estimate was based on population cohort studies (in Sweden and in the United States). In a review paper published in 2012 (Brodaty, Seeher, & Gibson, 2012), the authors calculated the number of years of life lost due to dementia by subtracting survival times from life-table data matched by age at diagnosis and sex ratio
94 SECTION | II Descriptive Epidemiology
in countries in which 42 studies were published. They included studies from the United States (49%), Europe (44%), and Australia (7%). In young-onset disease, the absolute number of years of life lost was much higher (range 9.6–19.4 years) compared with late-onset disease (range 1.3–9.2 years). Years of life lost among young-onset cases was 60–94% of their average life expectancy, compared with 16–73% for later-onset patients. Among the eight studies that looked at differences in survival in men and women, women lost more years of life in absolute terms, but the relative loss of years of life was similar by sex, decreasing with increasing age.
COMPETING CAUSES OF DEATH It is important to recognize that deaths from other chronic diseases are decreasing, while deaths due to AD are increasing. In the 2013 Alzheimer’s Disease report (Gaugler et al., 2013), it was noted that deaths due to stroke decreased between 2000 and 2010 by 23%, due to heart disease by 16%, due to human immunodeficiency virus by 42%, and due to prostate cancer by 8%. During the same period, Alzheimer deaths increased by 68% (according to NCHS statistics). The changing picture of mortality suggests that in the future, AD will become a much more important contributor to mortality than it is today, unless effective preventions for this illness are discovered and employed.
Survival and Mortality in Alzheimer’s Disease As early as 1976, Dr Robert Katzman noted that the late-onset form of Alzheimer’s disease (AD) “may rank as the fourth or fifth most common cause of death in the US … yet the U.S. vital statistics tables do not list ‘Alzheimer disease,’ ‘senile dementia,’ or ‘senility’ as a cause of death, even in the extended list of 263 causes of death” (Katzman, 1976). The death certificates in the United States and many other developed countries have been poor sources of information regarding death from AD. Dementia and AD cause many deaths indirectly. For example, death may result from aspiration pneumonia in advanced AD patients who are unable to eat or decubitus ulcers among patients who are bedbound. It is therefore important that all causes of death—underlying, immediate, associated or contributing—be analyzed on the death certificates. Even then, the death certificate is a poor source of information due to the frequent lack of information available to the physician completing it. In the first study of US mortality data taking into account all causes of death, Chandra, Bharucha, and Schoenberg (1986) examined causes of death in 1971 and 1973–1978 using International Classification of Diseases Adapted for Use in the United States (ICDA) (US Department of Health Education and Welfare, 1967) rubrics for “senile or presenile brain disease” (ICDA rubric 290) and “senility” (ICDA rubric 794). Data were analyzed separately for deaths due to dementia and those with dementia. The overall annual age-adjusted mortality rate was 0.41 and 2.19/100,000 for mortality due to and with senile and presenile dementia, respectively, and 0.52 and 12.21/100,000 for mortality due to and with senility, respectively. In this analysis, rates were higher among men and for whites for deaths due to and with senile and presenile dementia, and for nonwhites, the rates were higher for deaths due to and with senility. Among those who died with senile and presenile dementia, the most common underlying causes of death were coded as diseases of the heart (33.4%), senile and presenile dementia (18.2%), cerebrovascular diseases (12.9%), and influenza and pneumonia (4.5%). For those dying with senility, the top five underlying
Alzheimer’s Disease. DOI: http://dx.doi.org/10.1016/B978-0-12-804538-1.00008-0 © 2016 2013 Elsevier Inc. All rights reserved.
89
90 SECTION | II Descriptive Epidemiology
causes of death were diseases of the heart (41.7%), cerebrovascular diseases (14.6%), arteriosclerosis (9.3%), influenza and pneumonia (7.9%), and malignant neoplasms (4.4%). Only 18.2% of the death certificates listed senile and presenile dementia as the underlying cause of death, and 4.3% listed senility as the underlying cause among those who had died with these conditions (Chandra et al., 1986). In a study using NCHS mortality data, age-adjusted annual mortality rates per 100,000 for AD for ages 65 and over increased from 2.5 in 1979 to 37.5 in 1995 (Hoyert & Rosenberg, 1997). This large increase likely reflects changing codes and diagnostic criteria, as well as increased awareness of the disease. The study by Hoyert and Rosenberg reported that 20,000 death certificates had AD as the underlying cause of death and another 20,000 as another cause of death. In a commentary, White (1997) estimated a conservative number of 100,000 deaths in the United States around this time from end-stage dementia, based on a prevalence of AD at the time between one and two million. Therefore, it would appear that in only 40% of cases were dementia diagnoses included on death certificates. Other studies have produced higher estimates. Among patients enrolled in studies reporting their findings to the Consortium to Establish a Registry for Alzheimer’s disease (CERAD) (Raiford, Anton-Johnson, Haycox, et al., 1994), dementia was reported on about two-thirds of death certificates. It must be kept in mind that this group of patients was highly selected and that participation in a research study on dementia could increase the likelihood of dementia appearing on the death certificate. In a UK study (Martyn & Pippard, 1988), only 22% of demented patients had an ICD code of presenile or senile dementia on their death certificate. However, 58% of these death certificates mentioned the word “dementia” somewhere on the certificate. Frecker, PrysePhillips, and Strong (1995) reported their experience from Newfoundland, Canada, where they noted that 51/59 (86.2%) death certificates for probable AD cases by NINCDS-ADRDA criteria listed AD on the certificate. The place of death may make a difference in the degree of accuracy, because, for example, the clinical histories of patients dying in hospital are less known to physicians than those dying in nursing homes. Data derived from the Chicago Health and Aging Project (CHAP) and extrapolated to the US population (Weuve, Hebert, Scherr, & Evans, 2014) estimated that 600,000 people with AD died in 2010 in the United States, including people who received medical care and those who did not. By 2030, these estimates are expected to increase to 900,000, and in 2050, to 1.6 million. In addition, due to the aging of the US population, the proportion of deaths with AD will increase rapidly. By 2050, 43% of all older adult deaths will occur among people with AD, up from 32% in 2010 (Weuve et al., 2014). While there is a trend for death certificates to increasingly include AD as a cause, there is consensus that only about half of existing AD cases are ever diagnosed (Thies & Bleiler, 2011).
Survival and Mortality in Alzheimer’s Disease Chapter | 8 91
AD IN THE TOP 10 LEADING CAUSES OF DEATH In 2010, the top 10 leading causes of death among people age 65 and over listed AD as sixth, with 83,494 deaths (Murphy, Xu, & Kochanek, 2013). If there truly are 600,000 deaths per year from AD, AD would be regarded as the #1 cause of death in the United States, instead of heart disease (which, in 2010, was estimated to claim 477,338 lives). It is important to note that in national statistics, the numbers of deaths reported for AD are those for which the underlying cause of death is AD. Many patients with AD die from heart disease, stroke, pneumonia, and influenza, which would frequently be considered as the underlying causes of death. The degree to which AD contributes to such deaths is unknown. For example, would the person have died from stroke if they didn’t have AD? This issue highlights the difficulty in pinpointing the individual contributions of various chronic diseases as well as acute illnesses, such as influenza, in causing death in the oldest segment of our population. Given that about 50% of cases of dementia are not diagnosed before death, screening for incident dementia or AD in community-based studies and following cases over time is necessary to accurately enumerate deaths and to appreciate the importance of AD and dementia in increasing the risk for death. It also is important to start with incident cases, so that the time between onset and death can be counted properly. In addition, studies need to be relatively large so that there is a sufficient number of incident and mortality outcomes. Such data are available from a number of studies, and a few are selected here for illustrative purposes. In the PAQUID study, begun in 1988, a cohort of 2,923 community-based elders age 65 and over living in southwestern France (Gironde and Dordogne) were randomly sampled from electoral rolls and followed for up to 8 years (Helmer, Joly, Letenneur, Commenges, & Dartigues, 2001). During this time, 281 incident cases were diagnosed with dementia (67% AD), with a mean age at onset of 80 ± 5.9 among men and 83.5 ± 6.1 among women. Mortality was analyzed using Cox proportional hazards models with delayed entry, adjusting for sex, education, comorbidities, and dependence in ADL/IADL at baseline. In the population, the RR for mortality associated with incident dementia was 1.80 (95% CI = 1.46–2.21), and about the same, 1.72 (95% CI = 1.34–2.21) for AD. The analysis also showed effect-modification by age, such that dementia had less of an impact on mortality with increasing age (RR= 0.95, 95% CI = 0.92–0.98 per year). For participants dying with dementia above age 85, the RR was elevated but not statistically significant (RR = 1.37, 95% CI = 0.91–2.04). In this study, there were no sex differences in the RR for dying with dementia. There was an increased risk for dying with respiratory disease among individuals with AD; and the risk for dying with cerebrovascular disease was statistically significant for dementia, but not for AD. The median survival time from onset of dementia to death was 4.5 years for all ages, being longer among women (7.3 years at age 75 compared with 4.5 for men; and 4.4 at age 85 compared with
92 SECTION | II Descriptive Epidemiology
3.5 for men), likely reflecting the survival advantage of women over men in the general population. Survival time also was modified by education, with those with more education having shorter survival times than those with less education (Amieva et al., 2010). Modification of survival time by education was also reported in the WHICAP cohort. The RR for dying among AD patients was 1.9 (95% CI = 1.2–3.0) for those with 9 or more years of education in comparison with those with 8 or fewer years. Those with higher occupational attainment also experienced a higher risk for mortality compared with those with lower occupational attainment (RR = 2.2, 95% CI = 1.1–4.3) (Stern, Tang, Denaro, & Mayeux, 1995). This would be predicted by the threshold hypothesis, since AD patients with higher education and occupation would experience a later clinical onset of their disease, when AD-related neuropathology was more severe given their higher brain reserve. Such individuals would be expected to decline more rapidly and have a shorter duration of disease from onset to death (Stern, Alexander, Prohovnik, & Mayeux, 1992). The WHICAP study also found a differential survival in AD by race/ ethnicity (Helzner et al., 2008). Over a mean follow-up time of about 4 years, 323 incident AD cases developed, who were on average 87 years old at baseline and who had a median lifespan of 92.3 years. The median survival after AD onset among non-Hispanic whites was 3.7 years, 4.8 years among AfricanAmericans, and 7.6 years among Hispanics. The findings were not explained by differences in overall lifespan by race/ethnicity. However, these differences could be explained by differences in education and occupational attainment, with the African-Americans and Hispanics being diagnosable at earlier ages. Another way to look at mortality is to compare remaining life expectancy for AD patients in comparison to the general population. In the Adult Changes in Thought Study (ACT Study), an AD patient registry was set up in 1987 in the Seattle, WA regional area to identify all persons age 60 and over in the population of a health-maintenance organization of 23,000 members who developed AD (Larson et al., 2004). Between 1987 and 1996, incident cases were invited to participate in the study, and 970 with suspected dementia (521 AD: 431 probable and 90 possible by NINCDS-ADRDA criteria) were followed annually. Survival in AD in this sample did not vary by education and was longer for younger-onset patients and women. Poorer survival with AD was predicted by male sex, lower baseline cognitive score, poorer Dementia Rating Scale score, and by the existence of frontal release and extrapyramidal signs, history of falls, incontinence and history of heart disease and stroke. Survival time medians from diagnosis to death were 7.5 for AD cases ≤75 years old, 5.6 for those 76–80, 4.9 for those 81–85 years old, and 3.2 for those over age 85. A 70-year old woman with AD was expected to live 7.7 fewer years than a woman in the general US population, and a similarly-aged man, 8 fewer years. This gap narrowed with increasing age (3.3 years for an 80-year old woman and 3.1 years for an 80-year old man, and 1.8 and 0.5 years at age 90 for women and men,
Survival and Mortality in Alzheimer’s Disease Chapter | 8 93
respectively). The Hazard Ratios for mortality by MMSE score, using scores of 25–30 as the reference group, were 1.4 (95% CI = 1.01–1.94) for MMSE scores 22–24, 1.69 (95% CI = 1.24–2.31) for MMSE scores 18–21, and 2.67 (95% CI = 1.94–3.66) for MMSE scores ≤17. Two cohort studies that required autopsy consent for enrollment are informative for mortality. Both studies, the Religious Orders Study (ROS) and the Rush Memory and Aging Project (MAP), enrolled nondemented participants. In the ROS, the 1,168 subjects were Catholic nuns, priests, and brothers aged 65 and older from around the United States (Bennett, Schneider, Arvanitakis, & Wilson, 2012). The analysis of mortality in this cohort was conducted between 1994 and 2013 (James et al., 2014). In the Rush MAP, 1,574 people living in retirement communities and senior housing around Illinois (Bennett, Schneider, Buchman, et al., 2012) were enrolled between 1997 and 2013. The studies were conducted by the same research team, all measures were standardized, and the authors pooled the data assuming the populations were comparable. There were a total of 2,566 nondemented subjects 78 years old at baseline who were followed for an average of 8 years, 559 of whom developed incident AD over the study follow-up (mean age of AD diagnosis 86.5 ± 6.5). Six hundred and ninety-two (34.5%) of the nondemented participants died, versus 398 (71.2%) of those with incident AD. The median survival time from diagnosis of AD to death was 3.8 years and depended on the age of the subject at diagnosis: 4.4 years for those age 75–84 and 3.2 for those age 85 and over. After adjusting for age, sex, race, education, and study in the proportional hazards regression model, the HR for mortality associated with incident AD was 3.13 (95% CI = 2.74–3.58). The authors calculated the population attributable fraction (PAF) for mortality associated with AD, meaning the proportion of deaths in this population that were attributable to AD. These estimates were 37% for participants between the ages of 75–84 after adjustments for demographic variables, and 36% for those age 85 and older. When these PAFs were extrapolated to the number of deaths in the United States, the authors estimated that there were 503,400 excess deaths in 2010 associated with AD. This estimate is very close to the estimate of 600,000 obtain by Weuve et al. (2014). Importantly, the authors of the ROS and Rush MAP point out that their estimate of the number of people dying because of AD is 5–6 times higher than that acknowledged by the CDC in 2010 (83,494 deaths). This study has been criticized because no individual data were used to investigate the actual causes of death of cohort members and because the samples were not population-based. In contrast to the PAF published by the Rush group (James et al., 2014), the Alzheimer’s Association report for 2013 states that the PAF for AD on mortality over a 5-year period in people age 65 and over ranges from 5% to 15% (Gaugler et al., 2013). This estimate was based on population cohort studies (in Sweden and in the United States). In a review paper published in 2012 (Brodaty, Seeher, & Gibson, 2012), the authors calculated the number of years of life lost due to dementia by subtracting survival times from life-table data matched by age at diagnosis and sex ratio
94 SECTION | II Descriptive Epidemiology
in countries in which 42 studies were published. They included studies from the United States (49%), Europe (44%), and Australia (7%). In young-onset disease, the absolute number of years of life lost was much higher (range 9.6–19.4 years) compared with late-onset disease (range 1.3–9.2 years). Years of life lost among young-onset cases was 60–94% of their average life expectancy, compared with 16–73% for later-onset patients. Among the eight studies that looked at differences in survival in men and women, women lost more years of life in absolute terms, but the relative loss of years of life was similar by sex, decreasing with increasing age.
COMPETING CAUSES OF DEATH It is important to recognize that deaths from other chronic diseases are decreasing, while deaths due to AD are increasing. In the 2013 Alzheimer’s Disease report (Gaugler et al., 2013), it was noted that deaths due to stroke decreased between 2000 and 2010 by 23%, due to heart disease by 16%, due to human immunodeficiency virus by 42%, and due to prostate cancer by 8%. During the same period, Alzheimer deaths increased by 68% (according to NCHS statistics). The changing picture of mortality suggests that in the future, AD will become a much more important contributor to mortality than it is today, unless effective preventions for this illness are discovered and employed.