Acta Psychiatr Scand 2015: 132: 257–269 All rights reserved DOI: 10.1111/acps.12454
© 2015 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd ACTA PSYCHIATRICA SCANDINAVICA
Mortality in incident dementia – results from the German Study on Aging, Cognition, and Dementia in Primary Care Patients Roehr S, Luck T, Bickel H, Brettschneider C, Ernst A, Fuchs A, Heser K, K€ onig H-H, Jessen F, Lange C, M€ osch E, Pentzek M, Steinmann S, Weyerer S, Werle J, Wiese B, Scherer M, Maier W, Riedel-Heller SG, for the AgeCoDe Study Group. Mortality in incident dementia – results from the German Study on Aging, Cognition, and Dementia in Primary Care Patients. Objective: Dementia is known to increase mortality, but the relative loss of life years and contributing factors are not well established. Thus, we aimed to investigate mortality in incident dementia from disease onset. Method: Data were derived from the prospective longitudinal German AgeCoDe study. We used proportional hazards models to assess the impact of sociodemographic and health characteristics on mortality after dementia onset, Kaplan–Meier method for median survival times. Results: Of 3214 subjects at risk, 523 (16.3%) developed incident dementia during a 9-year follow-up period. Median survival time after onset was 3.2 years (95% CI = 2.8–3.7) at a mean age of 85.0 (SD = 4.0) years (≥2.6 life years lost compared with the general German population). Survival was shorter in older age, males other dementias than Alzheimer’s, and in the absence of subjective memory complaints (SMC). Conclusion: Our findings emphasize that dementia substantially shortens life expectancy. Future studies should further investigate the potential impact of SMC on mortality in dementia.
S. Roehr1, T. Luck1,2,
H. Bickel3, C. Brettschneider4, A. Ernst5, A. Fuchs6, K. Heser7, H.-H. K€onig4, F. Jessen8,9, C. Lange5, E. M€osch3, M. Pentzek6, S. Steinmann10, S. Weyerer11, J. Werle11, B. Wiese10, M. Scherer5, W. Maier7,9,*, S. G. Riedel-Heller1,*, for the AgeCoDe Study Group 1
Institute of Social Medicine, Occupational Health and Public Health (ISAP), University of Leipzig, Leipzig, 2LIFE – Leipzig Research Center for Civilization Diseases, University of Leipzig, Leipzig, 3Department of Psychiatry, Klinikum rechts der Isar, Technical University of Munich, Munich, 4Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center HamburgEppendorf, Hamburg, 5Department of Primary Medical Care, Center for Psychosocial Medicine, University Medical Center, Hamburg-Eppendorf, 6Institute of General Practice, Medical Faculty, Heinrich-HeineUniversity D€usseldorf, D€usseldorf, 7Department of Psychiatry, University of Bonn, Bonn, 8Department of Psychiatry, University of Cologne, Medical Faculty, Cologne, 9German Center for Neurodegenerative Diseases, DZNE, Bonn, 10Work Group Medical Statistics and IT-Infrastructure, Institute for General Practice, Hannover Medical School, Hannover and 11Central Institute of Mental Health, Medical Faculty Mannheim/ Heidelberg University, Mannheim, Germany Key words: dementia; mortality; prognosis; risk factors; subjective memory complaints Susanne Roehr, Institute of Social Medicine, Occupational Health and Public Health (ISAP), University of Leipzig, Philipp-Rosenthal-Str. 55, 04103 Leipzig, Germany. E-mail: [email protected] *These authors contributed equally to the manuscript.
Accepted for publication May 11, 2015
257
Roehr et al.
Significant outcomes
• Dementia • •
substantially shortens life expectancy. At a mean age of 85.0 years, life expectancy in dementia was at least 2.6 years shorter compared with the German general population. Besides older age, male gender, and type of dementia, subjective memory complaints had a significant impact on mortality after dementia onset. Subjective memory complaints have not much been taken into account in survival analysis yet. Its potential influence on mortality in dementia should be further investigated in future studies, and it also should be taken seriously in clinical practice when expressed by patients.
Limitations
• Generalizability of the results may be limited due to a moderate response rate to the study. • The sample might not be representative, because participants were recruited by general practitioners (GP), even though more than 90% of the German elderly population consult a GP regularly.
• Dementia diagnosis was based on brief assessment instruments. Those might lack sensitivity or specificity compared with thorough clinical investigations.
Introduction
Dementia is known to increase mortality, but the relative loss of life years and factors contributing to a shorter life expectancy in Alzheimer’s disease (AD) and other types of dementia are not well established (1, 2). Estimates of survival times in dementia and knowledge of risk factors are necessary though to provide information on prognosis. Reliable prognoses support patients, relatives, caregivers, and health experts to decide for appropriate services, such as palliative care or nursing home admission (3, 4). Moreover, accurate prognoses allow calculations of anticipated expenses in health care and social systems. The costs in dementia care are intensive as the progressive cognitive impairment, which is distinctive for dementia, leads successively into a substantial need for care (5, 6). The world-wide expenses will increase in line with the ever-growing number of people affected cumulating in an expected financial burden of US$ 1117 billion per year in 2030 (6). This is due to increasing life expectancies and rapid population aging. The global estimated number of dementia patients is expected to triple from 44.4 million in 2013 to 135.5 million in 2050 (7). Numerous factors have been analyzed regarding their impact on mortality in dementia, including age and gender (8–12), sociodemographic characteristics [e.g. education (9, 10, 12–14), marital status, and accommodation (11, 14)], type (12, 14–17) and severity of dementia (8, 9, 11, 18), ethnicity (19–21), comorbidity (8, 9, 14, 19), and genetic characteristics (16, 17, 19) with often inconsistent results across studies (1, 2). To date, 258
the best established factor associated with shorter survival in dementia is older age. Male gender also seems to contribute to increased mortality, although the results are less clear (1). Most studies investigating mortality in dementia so far focused on prevalent cases, which comprise different stages of disease progression at study entry (1, 2, 19). Hence, survival might be underestimated (22). Likewise, survival times could be overestimated if patients who rapidly die after diagnosis fall short of being recruited into studies at all, which is referred to as length bias (23). As a consequence, estimates of median survival times reported in dementia range widely from 3.2 to 11.7 years (1). Moreover, results on survival also depend on whether they are calculated from date of disease onset or date of diagnosis (2). Prognoses from date of onset in incident cases avoid (i) the variability of disease severeness that is given in points of diagnoses and (ii) the inaccuracy of retrospective estimates of onset in prevalent cases (1, 23). Studies that actively and regularly screen for new – incident – dementia cases can elude these restraints in defining disease onset more precisely, but, to our knowledge, few studies have proceeded likewise before (11, 16, 17, 24). Aims of the study
Thus, this study aimed to investigate survival times and predictors of mortality from disease onset in incident dementia cases from a large German sample of primary care patients aged 75+ over a 9-year follow-up period. We intended to look at differences in sociodemographic and health
Mortality in incident dementia characteristics at dementia onset by survival status, before we assessed the effects of these factors on death. Finally, we estimated survival times in dementia for relevant factors.
Material and methods Study design and sample
Data were derived from the German study on Aging, Cognition, and Dementia in Primary Care Patients (AgeCoDe1), a prospective longitudinal study on the early detection of mild cognitive impairment and dementia in general practice. The study was conducted at six centers (Bonn, Duesseldorf, Hamburg, Leipzig, Mannheim, and Munich), covering urban areas with populations from 308 000 (Mannheim) to 1.75 million residents (Hamburg). The cohort was recruited between January 1, 2003 and November 30, 2004 by 138 general practitioners (GP). Inclusion criteria at baseline were age 75 years and above, the absence of dementia as judged by the GP, and at least one contact with the GP within the last 12 months. Exclusion criteria were GP consultations at home only, residence in a nursing home, severe illness with an expected fatal outcome within 3 months, German language insufficiency, deaf or blindness, and a lack of ability to provide informed consent. About 91% of the German population are registered at a GP office with even higher numbers in older subjects (25). Study subjects were randomly selected from the medical records by the GPs. At baseline, a total of 3327 GP patients (mean age: 80.1 years, SD = 3.6; 65.5% women) were investigated. Of those, 113 (3.4%) subjects were excluded because of prevalent dementia (n = 70, 2.1%), shortage of the 75 years of age limit (n = 39, 1.2%), or incomplete assessments (n = 4, 0.1%). A total of 3214 subjects constituted the population at risk for developing dementia at follow-up waves (Fig. 1). Study details have been previously described elsewhere (26). Data collection and assessment procedures
Data were collected between January 23, 2003 (begin of baseline) and October 29, 2012 (end of follow-up 5). Follow-up assessments took place on average every 1.5 years during home visits by trained psychologists and physicians, who conducted structured clinical interviews to collect sociodemographic, clinical, and psychometric data. 1 Members of the AgeCoDe Study Group have been given in the Appendix 1.
General practitioners provided further information on medical conditions by completing standardized questionnaires at baseline and each follow-up wave. The main assessment instrument was the structured interview for the diagnosis of dementia of Alzheimer type, multi-infarct dementia, and dementia of other etiology according to DSMIII-R, DSM-IV, and ICD-10 (SIDAM) (27). The SIDAM contains a cognitive test battery and a scale for clinical judgment or proxy information on psychosocial functioning, including a 14-item scale to assess impairment in activities of daily living (ADL, SIDAM-ADL scale). The cognitive battery consists of 55 items (SIDAM score) that cover four main domains of cognitive functioning: orientation, memory, intellectual abilities, and higher cortical functioning. A higher SIDAM score (max. 55) indicates better cognitive functioning. Included in the 55 SIDAM items is also the Mini-Mental State Examination (MMSE), a 30-item screening instrument for cognitive status (28) that is widely used for the prognosis of survival in dementia. A higher score (max. 30) indicates a better cognitive status. The Lawton and Brody scale was used to gather information on the impairment in instrumental activities of daily living (IADL) (29). In this study, we used a modified 5-item version of the scale including IADL that account likewise for men and women: to use the telephone, to handle finances, to use means of public transport, to shop daily supplies, and to responsibly handle medication intake. We excluded three domains from the analysis that have been typically attributed to women (cooking, washing, and housekeeping), as the core items have been shown to be correlated with cognitive impairment independently from age, gender, and education (30). Depressive symptoms were assessed using the short version of the Geriatric Depression Scale (GDS) (31). The GDS consists of 15 questions specific for the elderly such as ‘Have you dropped many of your activities and interests?’. The maximum total score is 15. A score ≥6 indicates the presence of depressive symptoms. Prior to cognitive testing, subjective memory complaints (SMC) were evaluated by asking ‘Do you feel as if your memory is becoming worse?’ (yes/no/I don’t know). In case of a positive answer, we further specified if SMC involved related worry by asking ‘Does this worry you?’ (yes/no/I don’t know). The presence of comorbidity was assessed using a standardized questionnaire that the GP completed at each wave. 259
Roehr et al.
BASELINE
SAMPLING FRAME OF REGISTERED GENERAL PRACTICE POPULATION
NONELIGIBLE n = 11 851
n = 22 701
Irregular patients Only home visits Deceased No ability to consent Severely ill Deaf or blind Language Other reasons Not documented
ELIGIBLE n = 10 850
4792 2477 2075 1107 326 245 226 345 258
RANDOMLY SELECTED SAMPLE SIZE n = 6619
NONPARTICIPANTS n = 3292
INVESTIGATED AT BASELINE
Refused Contacting failed
1775 1517
n = 3327
EXCLUSION n = 113 FOLLOW-UP I
INCLUDED AT BASELINE n = 3214
Dementia 70 Age < 75 years 39 Incomplete assessment 4
EXCLUSION n = 16 INFORMATION ON DEMENTIA STATUS AT FOLLOW-UP WAVES n = 3198
No information on dementia at follow-up waves
EXCLUSION n = 2675 INCIDENT DEMENTIA CASES AT FOLLOW-UP WAVES n = 523
FOLLOW-UP V
DECEASED
SURVIVED
n = 367
n = 156
Died without dementia No dementia by FUV Refused Contacting failed Moved Dropped-out for other reasons
1131 1105 382 15 17 25
Fig. 1. Sample attrition and sample.
At baseline and at the first two follow-up waves, information on smoking status and alcohol consumption was collected by asking whether and how much the participants smoke and/or drink. Alcohol consumption was assessed by rating the amount of alcohol consumed on a regular basis in categories from no drinking to alcohol dependence according to the guidelines of the World Health Organisation (32). Death dates were obtained from relatives, GP, or the local residents’ registration office, at which all persons in Germany must be registered by law. If participants could not be interviewed at follow-up, proxy interviews were conducted which included the Global Deterioration Scale (33) and the subscales ‘Changes on performance of 260
everyday experiences’ and ‘Changes in habits’ of the Blessed Dementia Scale (34). Definition of cases
Dementia was diagnosed in a consensus conference of the interviewers and an experienced geriatrician or geriatric psychiatrist according to DSM-IV criteria as implemented in the SIDAM (27). Type of dementia was assessed by the NINCDS-ADRDA criteria for probable AD (35) and by the NINDSAIREN criteria for vascular dementia (36). Vascular dementia was diagnosed in case of, for example, cerebrovascular events that showed a temporal relationship to cognitive decline as assessed by the Hachinski–Rosen Scale (37) and the medical record. Mixed dementia was assumed if
Mortality in incident dementia cerebrovascular events were not temporally associated with cognitive decline. For all analyses, AD and mixed dementia were combined. Vascular dementia made up a second category. Further etiological diagnoses of dementia were subsummarized as ‘others’, containing such forms as Parkinson’s disease dementia or dementia of Lewy bodies (DLB). If participants could not be interviewed, dementia diagnoses were based on a cutoff score of ≥4 on the Global Deterioration Scale (33) and a total score of ≥9 on the Blessed Dementia Rating (34) subscales as judged by proxies. In this case, type of dementia was only specified if the information provided were sufficient. Statistical analyses
Statistical analyses were performed with STATA/IC, version 13.0 (StataCorp LP, College Station, TX, USA), and SPSS, version 20.0 (IBM, Armonk, NY, USA). The significance level was set at a = 0.05 for all calculations. Group differences in sociodemographic and health characteristics at dementia onset were analyzed in respect to survival status using t-tests or Mann–Whitney U-test for continuous variables and chi-squared test for categorical variables. By convention, dementia onset was assumed the midway point between two follow-up waves, from which the latter constituted the point of incident dementia diagnosis. For variables with potentially time-dependent values, we chose data from the interview closest to dementia onset. Survival in incident dementia was defined as the interval from date of dementia onset to date of death. Subjects who were still alive by the end of follow-up 5 were treated as censored data. Attained age was measured as date of birth to date of death, or, for survivors, date of last contact. Person-years of observation were defined as time between dementia onset and death or last day of contact in survivors. Mortality as case-fatality rates was calculated by dividing the number of deaths by person-years of observation. Multivariable Cox proportional hazards regression models were used to assess the effects of subjects’ health and sociodemographic characteristics at dementia onset on mortality. Age, MMSE score, depressive symptoms as the total score of the GDS, and IADL score were implemented as continuous variables, and gender, comorbidity (nonpresent vs. present), and smoking (non-smoker vs. current and former smoker) as dichotomous variables. Education [low, middle, and high as classified by the Comparative Analysis of Social
Mobility in Industrial Nations, CASMIN (38)], marital status (married/cohabiting, single, divorced, and widowed), accommodation (private home, residential care, nursing home), type of dementia (AD and mixed, vascular dementia, and others), SMC (no, yes without related worry, and yes with related worry), and alcohol consumption (no drinking, normal drinking, risky/harmful drinking, and alcohol dependence) were set up as categorical variables. At first, a Cox regression was performed including all above named variables to evaluate their simultaneous effect on mortality. Then, we applied a second Cox regression using a backward stepwise procedure to identify the most parsimonious model on risk factors for mortality in incident dementia cases. For each variable in the models, we calculated hazard ratios (HR) and Wald 95% confidence intervals (CI). The proportional hazards assumption was tested for both Cox models using Schoenfeld residuals (39). Kaplan–Meier survival analysis was used to determine median survival times for significant prognostic factors that were evaluated in the Cox regressions. Further, log-rank tests were performed to assess survival differences between groups. Finally, plots were generated to graphically illustrate the differences in the cumulative survival probability. Ethical approval
The study was conducted in accordance with the Declaration of Helsinki (40) and has been approved by the ethics committees of the participating centers. All patients and/or their proxies provided written informed consent.
Results Descriptive characteristics
Among the 3214 subjects at risk in the AgeCoDe cohort after baseline investigation, 523 (16.3%) incidentally developed dementia throughout five follow-up waves covering a time period of up to 9.3 years (Fig. 1). Of the 523 incident dementia cases, 362 (69.2%) were women and 161 were (30.8%) men. The mean follow-up time was 6.8 years (SD = 2.5), while deceased subjects were followed for M = 5.9 years (SD = 2.2) and survivors for M = 9.1 years (SD = 1.3; U = 4286.0, P < 0.001). Mean age at estimated dementia onset was 85.0 years (SD = 4.0), ranging from 76 to 99 years. Men were significantly younger at dementia onset than 261
Roehr et al. women (M = 84.0, SD = 3.8 vs. M = 85.4, SD = 4.1; t(521) = 3.61, P < 0.001). By the end of follow-up 5, 367 subjects (69.2%) had died by a mean age of 87.6 years (SD = 4.3), ranging from 77 to 103 years. Overall, subjects were under observation for 1640.7 person-years. The casefatality rate cumulated in 223.7 per 1000 personyears (95% CI = 201.9–247.8). Among the 523 incident cases, 328 (67.8%) participants were diagnosed with AD and mixed dementia, 87 (18.0%) with vascular dementia, and 69 (14.3%) with other types of dementia. In 39 (7.5%) cases, type of dementia could not be defined due to insufficient information from proxy interviews. Table 1 shows sociodemographic and health characteristics of the participants by survival status. Subjects who had died were significantly more frequently male than female (P < 0.001), had more frequently vascular dementia or another type of dementia than AD or mixed dementia (P < 0.001), were less impaired on the IADL at dementia onset (P < 0.02) and were less often affected by hypertension at dementia onset (P < 0.01). Prognostic factors
Results of the impact of incident dementia on mortality were based on 431 subjects due to missing values in covariates in 92 (17.6%) cases. This sample reduction did not change the proportion of deceased subjects and survivors in the model sample compared with the initial sample [v²(1, 933) = 0.09, P = 0.76] nor did it lead to differences in other characteristics at dementia onset (results not shown). Both multivariable Cox proportional hazards models, the full and the backward stepwise regression, found older age, male gender, vascular dementia and dementia of other types (reference category: AD and mixed dementia), and the absence of SMC as significantly associated with increased mortality in incident dementia (Table 2). In the parsimonious backward stepwise model, every additional year of age yielded a HR of 1.06 (95% CI = 1.03–1.09, P < 0.001), male gender yielded a HR of 1.4 (95% CI = 1.1–1.8, P < 0.01) as opposed to women, vascular dementia yielded a HR of 1.5 (95% CI = 1.1–2.0, P = 0.01), and other types of dementia yielded a HR of 1.7 (95% CI = 1.2–2.3, P < 0.01) in contrast to AD and mixed dementia. The presence of SMC without related worry resulted in a HR of 0.7 (95% CI = 0.5–0.9, P = 0.01) and SMC with related worry in a HR of 0.7 (95% CI = 0.5–0.9, P < 0.01) in reference to the absence of SMC. 262
Table 1. Sociodemographic and health characteristics of the study sample at onset of incident dementia by mortality status
Variables†
All incident dementia cases (n = 523)
Deceased (n = 367)
Age at onset in years, mean (SD) Total 84.99 (4.03) 84.98 (4.09) Female 85.41 (4.05) 85.54 (4.04) Male 84.04 (3.83) 83.97 (3.99) Age group at onset in years, n (%) 75–79 57 (10.9) 42 (11.4) 80–84 215 (41.1) 151 (41.1) 85–90 186 (35.6) 125 (34.1) ≥90 65 (12.4) 49 (13.4) Gender, n (%) Female 362 (69.2) 236 (65.2) Male 161 (30.8) 131 (81.4) Education, n (%) Low 341 (65.2) 235 (64.0) Middle 128 (24.5) 91 (24.8) High 54 (10.3) 41 (11.2) Marital status, n (%) Single 33 (6.4) 25 (7.0) Married/cohabiting 180 (35.0) 136 (37.9) Divorced 26 (5.1) 18 (5.0) Widowed 275 (53.5) 180 (50.1) Accommodation, n (%) Private home 469 (90.0) 336 (91.8) Residential care 26 (5.0) 16 (4.4) Nursing home 26 (5.0) 14 (3.8) Type of dementia, n (%) Alzheimer’s and mixed 328 (67.8) 207 (62.0) Vascular dementia 87 (18.0) 72 (21.6) Others 69 (14.3) 55 (16.5) Subjective memory complaints (SMC), n (%) No 147 (28.5) 114 (31.7) Yes, without 196 (38.1) 131 (36.4) related worry Yes, with related 172 (33.4) 115 (31.9) worry Mini-Mental State 25.81 (2.24) 25.89 (2.32) Examination (MMSE), total score, mean (SD) IADL‡, total score, 4.04 (1.29) 4.12 (1.25) mean (SD) Comorbidity, n (%) Diabetes mellitus 129 (25.9) 91 (26.0) Hypertension 320 (64.4) 212 (60.6) Cardiac arrhythmias 151 (30.4) 101 (28.9) Coronary heart disease 160 (32.1) 109 (31.1) Myocardial infarction 41 (8.2) 29 (8.3) 65 (13.1) 48 (13.7) Peripheral artery occlusive disease (PAOD) Stenosis (of afferent 17 (3.4) 12 (3.4) brain vessels) Transient ischemic 66 (13.3) 43 (12.3) attack (TIA) Stroke 41 (8.2) 29 (8.3) Hyperlipidemia/ 238 (47.8) 159 (45.4) hypercholesterolemia Kidney insufficiency 80 (16.1) 51 (14.6) Hepatic disease 29 (5.8) 21 (6.0) Depressive symptoms§, 3.43 (2.78) 3.42 (2.79) total score, mean (SD) Smoking, n (%)
Survivors (n = 156)
P value (group difference)
85.00 (3.91) 85.16 (4.08) 84.35 (3.04)
0.95 0.40 0.63
15 (9.6) 64 (41.0) 61 (39.1) 16 (10.3)
0.58
126 (34.8) 30 (18.6)
© 2015 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd ACTA PSYCHIATRICA SCANDINAVICA
Mortality in incident dementia – results from the German Study on Aging, Cognition, and Dementia in Primary Care Patients Roehr S, Luck T, Bickel H, Brettschneider C, Ernst A, Fuchs A, Heser K, K€ onig H-H, Jessen F, Lange C, M€ osch E, Pentzek M, Steinmann S, Weyerer S, Werle J, Wiese B, Scherer M, Maier W, Riedel-Heller SG, for the AgeCoDe Study Group. Mortality in incident dementia – results from the German Study on Aging, Cognition, and Dementia in Primary Care Patients. Objective: Dementia is known to increase mortality, but the relative loss of life years and contributing factors are not well established. Thus, we aimed to investigate mortality in incident dementia from disease onset. Method: Data were derived from the prospective longitudinal German AgeCoDe study. We used proportional hazards models to assess the impact of sociodemographic and health characteristics on mortality after dementia onset, Kaplan–Meier method for median survival times. Results: Of 3214 subjects at risk, 523 (16.3%) developed incident dementia during a 9-year follow-up period. Median survival time after onset was 3.2 years (95% CI = 2.8–3.7) at a mean age of 85.0 (SD = 4.0) years (≥2.6 life years lost compared with the general German population). Survival was shorter in older age, males other dementias than Alzheimer’s, and in the absence of subjective memory complaints (SMC). Conclusion: Our findings emphasize that dementia substantially shortens life expectancy. Future studies should further investigate the potential impact of SMC on mortality in dementia.
S. Roehr1, T. Luck1,2,
H. Bickel3, C. Brettschneider4, A. Ernst5, A. Fuchs6, K. Heser7, H.-H. K€onig4, F. Jessen8,9, C. Lange5, E. M€osch3, M. Pentzek6, S. Steinmann10, S. Weyerer11, J. Werle11, B. Wiese10, M. Scherer5, W. Maier7,9,*, S. G. Riedel-Heller1,*, for the AgeCoDe Study Group 1
Institute of Social Medicine, Occupational Health and Public Health (ISAP), University of Leipzig, Leipzig, 2LIFE – Leipzig Research Center for Civilization Diseases, University of Leipzig, Leipzig, 3Department of Psychiatry, Klinikum rechts der Isar, Technical University of Munich, Munich, 4Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center HamburgEppendorf, Hamburg, 5Department of Primary Medical Care, Center for Psychosocial Medicine, University Medical Center, Hamburg-Eppendorf, 6Institute of General Practice, Medical Faculty, Heinrich-HeineUniversity D€usseldorf, D€usseldorf, 7Department of Psychiatry, University of Bonn, Bonn, 8Department of Psychiatry, University of Cologne, Medical Faculty, Cologne, 9German Center for Neurodegenerative Diseases, DZNE, Bonn, 10Work Group Medical Statistics and IT-Infrastructure, Institute for General Practice, Hannover Medical School, Hannover and 11Central Institute of Mental Health, Medical Faculty Mannheim/ Heidelberg University, Mannheim, Germany Key words: dementia; mortality; prognosis; risk factors; subjective memory complaints Susanne Roehr, Institute of Social Medicine, Occupational Health and Public Health (ISAP), University of Leipzig, Philipp-Rosenthal-Str. 55, 04103 Leipzig, Germany. E-mail: [email protected] *These authors contributed equally to the manuscript.
Accepted for publication May 11, 2015
257
Roehr et al.
Significant outcomes
• Dementia • •
substantially shortens life expectancy. At a mean age of 85.0 years, life expectancy in dementia was at least 2.6 years shorter compared with the German general population. Besides older age, male gender, and type of dementia, subjective memory complaints had a significant impact on mortality after dementia onset. Subjective memory complaints have not much been taken into account in survival analysis yet. Its potential influence on mortality in dementia should be further investigated in future studies, and it also should be taken seriously in clinical practice when expressed by patients.
Limitations
• Generalizability of the results may be limited due to a moderate response rate to the study. • The sample might not be representative, because participants were recruited by general practitioners (GP), even though more than 90% of the German elderly population consult a GP regularly.
• Dementia diagnosis was based on brief assessment instruments. Those might lack sensitivity or specificity compared with thorough clinical investigations.
Introduction
Dementia is known to increase mortality, but the relative loss of life years and factors contributing to a shorter life expectancy in Alzheimer’s disease (AD) and other types of dementia are not well established (1, 2). Estimates of survival times in dementia and knowledge of risk factors are necessary though to provide information on prognosis. Reliable prognoses support patients, relatives, caregivers, and health experts to decide for appropriate services, such as palliative care or nursing home admission (3, 4). Moreover, accurate prognoses allow calculations of anticipated expenses in health care and social systems. The costs in dementia care are intensive as the progressive cognitive impairment, which is distinctive for dementia, leads successively into a substantial need for care (5, 6). The world-wide expenses will increase in line with the ever-growing number of people affected cumulating in an expected financial burden of US$ 1117 billion per year in 2030 (6). This is due to increasing life expectancies and rapid population aging. The global estimated number of dementia patients is expected to triple from 44.4 million in 2013 to 135.5 million in 2050 (7). Numerous factors have been analyzed regarding their impact on mortality in dementia, including age and gender (8–12), sociodemographic characteristics [e.g. education (9, 10, 12–14), marital status, and accommodation (11, 14)], type (12, 14–17) and severity of dementia (8, 9, 11, 18), ethnicity (19–21), comorbidity (8, 9, 14, 19), and genetic characteristics (16, 17, 19) with often inconsistent results across studies (1, 2). To date, 258
the best established factor associated with shorter survival in dementia is older age. Male gender also seems to contribute to increased mortality, although the results are less clear (1). Most studies investigating mortality in dementia so far focused on prevalent cases, which comprise different stages of disease progression at study entry (1, 2, 19). Hence, survival might be underestimated (22). Likewise, survival times could be overestimated if patients who rapidly die after diagnosis fall short of being recruited into studies at all, which is referred to as length bias (23). As a consequence, estimates of median survival times reported in dementia range widely from 3.2 to 11.7 years (1). Moreover, results on survival also depend on whether they are calculated from date of disease onset or date of diagnosis (2). Prognoses from date of onset in incident cases avoid (i) the variability of disease severeness that is given in points of diagnoses and (ii) the inaccuracy of retrospective estimates of onset in prevalent cases (1, 23). Studies that actively and regularly screen for new – incident – dementia cases can elude these restraints in defining disease onset more precisely, but, to our knowledge, few studies have proceeded likewise before (11, 16, 17, 24). Aims of the study
Thus, this study aimed to investigate survival times and predictors of mortality from disease onset in incident dementia cases from a large German sample of primary care patients aged 75+ over a 9-year follow-up period. We intended to look at differences in sociodemographic and health
Mortality in incident dementia characteristics at dementia onset by survival status, before we assessed the effects of these factors on death. Finally, we estimated survival times in dementia for relevant factors.
Material and methods Study design and sample
Data were derived from the German study on Aging, Cognition, and Dementia in Primary Care Patients (AgeCoDe1), a prospective longitudinal study on the early detection of mild cognitive impairment and dementia in general practice. The study was conducted at six centers (Bonn, Duesseldorf, Hamburg, Leipzig, Mannheim, and Munich), covering urban areas with populations from 308 000 (Mannheim) to 1.75 million residents (Hamburg). The cohort was recruited between January 1, 2003 and November 30, 2004 by 138 general practitioners (GP). Inclusion criteria at baseline were age 75 years and above, the absence of dementia as judged by the GP, and at least one contact with the GP within the last 12 months. Exclusion criteria were GP consultations at home only, residence in a nursing home, severe illness with an expected fatal outcome within 3 months, German language insufficiency, deaf or blindness, and a lack of ability to provide informed consent. About 91% of the German population are registered at a GP office with even higher numbers in older subjects (25). Study subjects were randomly selected from the medical records by the GPs. At baseline, a total of 3327 GP patients (mean age: 80.1 years, SD = 3.6; 65.5% women) were investigated. Of those, 113 (3.4%) subjects were excluded because of prevalent dementia (n = 70, 2.1%), shortage of the 75 years of age limit (n = 39, 1.2%), or incomplete assessments (n = 4, 0.1%). A total of 3214 subjects constituted the population at risk for developing dementia at follow-up waves (Fig. 1). Study details have been previously described elsewhere (26). Data collection and assessment procedures
Data were collected between January 23, 2003 (begin of baseline) and October 29, 2012 (end of follow-up 5). Follow-up assessments took place on average every 1.5 years during home visits by trained psychologists and physicians, who conducted structured clinical interviews to collect sociodemographic, clinical, and psychometric data. 1 Members of the AgeCoDe Study Group have been given in the Appendix 1.
General practitioners provided further information on medical conditions by completing standardized questionnaires at baseline and each follow-up wave. The main assessment instrument was the structured interview for the diagnosis of dementia of Alzheimer type, multi-infarct dementia, and dementia of other etiology according to DSMIII-R, DSM-IV, and ICD-10 (SIDAM) (27). The SIDAM contains a cognitive test battery and a scale for clinical judgment or proxy information on psychosocial functioning, including a 14-item scale to assess impairment in activities of daily living (ADL, SIDAM-ADL scale). The cognitive battery consists of 55 items (SIDAM score) that cover four main domains of cognitive functioning: orientation, memory, intellectual abilities, and higher cortical functioning. A higher SIDAM score (max. 55) indicates better cognitive functioning. Included in the 55 SIDAM items is also the Mini-Mental State Examination (MMSE), a 30-item screening instrument for cognitive status (28) that is widely used for the prognosis of survival in dementia. A higher score (max. 30) indicates a better cognitive status. The Lawton and Brody scale was used to gather information on the impairment in instrumental activities of daily living (IADL) (29). In this study, we used a modified 5-item version of the scale including IADL that account likewise for men and women: to use the telephone, to handle finances, to use means of public transport, to shop daily supplies, and to responsibly handle medication intake. We excluded three domains from the analysis that have been typically attributed to women (cooking, washing, and housekeeping), as the core items have been shown to be correlated with cognitive impairment independently from age, gender, and education (30). Depressive symptoms were assessed using the short version of the Geriatric Depression Scale (GDS) (31). The GDS consists of 15 questions specific for the elderly such as ‘Have you dropped many of your activities and interests?’. The maximum total score is 15. A score ≥6 indicates the presence of depressive symptoms. Prior to cognitive testing, subjective memory complaints (SMC) were evaluated by asking ‘Do you feel as if your memory is becoming worse?’ (yes/no/I don’t know). In case of a positive answer, we further specified if SMC involved related worry by asking ‘Does this worry you?’ (yes/no/I don’t know). The presence of comorbidity was assessed using a standardized questionnaire that the GP completed at each wave. 259
Roehr et al.
BASELINE
SAMPLING FRAME OF REGISTERED GENERAL PRACTICE POPULATION
NONELIGIBLE n = 11 851
n = 22 701
Irregular patients Only home visits Deceased No ability to consent Severely ill Deaf or blind Language Other reasons Not documented
ELIGIBLE n = 10 850
4792 2477 2075 1107 326 245 226 345 258
RANDOMLY SELECTED SAMPLE SIZE n = 6619
NONPARTICIPANTS n = 3292
INVESTIGATED AT BASELINE
Refused Contacting failed
1775 1517
n = 3327
EXCLUSION n = 113 FOLLOW-UP I
INCLUDED AT BASELINE n = 3214
Dementia 70 Age < 75 years 39 Incomplete assessment 4
EXCLUSION n = 16 INFORMATION ON DEMENTIA STATUS AT FOLLOW-UP WAVES n = 3198
No information on dementia at follow-up waves
EXCLUSION n = 2675 INCIDENT DEMENTIA CASES AT FOLLOW-UP WAVES n = 523
FOLLOW-UP V
DECEASED
SURVIVED
n = 367
n = 156
Died without dementia No dementia by FUV Refused Contacting failed Moved Dropped-out for other reasons
1131 1105 382 15 17 25
Fig. 1. Sample attrition and sample.
At baseline and at the first two follow-up waves, information on smoking status and alcohol consumption was collected by asking whether and how much the participants smoke and/or drink. Alcohol consumption was assessed by rating the amount of alcohol consumed on a regular basis in categories from no drinking to alcohol dependence according to the guidelines of the World Health Organisation (32). Death dates were obtained from relatives, GP, or the local residents’ registration office, at which all persons in Germany must be registered by law. If participants could not be interviewed at follow-up, proxy interviews were conducted which included the Global Deterioration Scale (33) and the subscales ‘Changes on performance of 260
everyday experiences’ and ‘Changes in habits’ of the Blessed Dementia Scale (34). Definition of cases
Dementia was diagnosed in a consensus conference of the interviewers and an experienced geriatrician or geriatric psychiatrist according to DSM-IV criteria as implemented in the SIDAM (27). Type of dementia was assessed by the NINCDS-ADRDA criteria for probable AD (35) and by the NINDSAIREN criteria for vascular dementia (36). Vascular dementia was diagnosed in case of, for example, cerebrovascular events that showed a temporal relationship to cognitive decline as assessed by the Hachinski–Rosen Scale (37) and the medical record. Mixed dementia was assumed if
Mortality in incident dementia cerebrovascular events were not temporally associated with cognitive decline. For all analyses, AD and mixed dementia were combined. Vascular dementia made up a second category. Further etiological diagnoses of dementia were subsummarized as ‘others’, containing such forms as Parkinson’s disease dementia or dementia of Lewy bodies (DLB). If participants could not be interviewed, dementia diagnoses were based on a cutoff score of ≥4 on the Global Deterioration Scale (33) and a total score of ≥9 on the Blessed Dementia Rating (34) subscales as judged by proxies. In this case, type of dementia was only specified if the information provided were sufficient. Statistical analyses
Statistical analyses were performed with STATA/IC, version 13.0 (StataCorp LP, College Station, TX, USA), and SPSS, version 20.0 (IBM, Armonk, NY, USA). The significance level was set at a = 0.05 for all calculations. Group differences in sociodemographic and health characteristics at dementia onset were analyzed in respect to survival status using t-tests or Mann–Whitney U-test for continuous variables and chi-squared test for categorical variables. By convention, dementia onset was assumed the midway point between two follow-up waves, from which the latter constituted the point of incident dementia diagnosis. For variables with potentially time-dependent values, we chose data from the interview closest to dementia onset. Survival in incident dementia was defined as the interval from date of dementia onset to date of death. Subjects who were still alive by the end of follow-up 5 were treated as censored data. Attained age was measured as date of birth to date of death, or, for survivors, date of last contact. Person-years of observation were defined as time between dementia onset and death or last day of contact in survivors. Mortality as case-fatality rates was calculated by dividing the number of deaths by person-years of observation. Multivariable Cox proportional hazards regression models were used to assess the effects of subjects’ health and sociodemographic characteristics at dementia onset on mortality. Age, MMSE score, depressive symptoms as the total score of the GDS, and IADL score were implemented as continuous variables, and gender, comorbidity (nonpresent vs. present), and smoking (non-smoker vs. current and former smoker) as dichotomous variables. Education [low, middle, and high as classified by the Comparative Analysis of Social
Mobility in Industrial Nations, CASMIN (38)], marital status (married/cohabiting, single, divorced, and widowed), accommodation (private home, residential care, nursing home), type of dementia (AD and mixed, vascular dementia, and others), SMC (no, yes without related worry, and yes with related worry), and alcohol consumption (no drinking, normal drinking, risky/harmful drinking, and alcohol dependence) were set up as categorical variables. At first, a Cox regression was performed including all above named variables to evaluate their simultaneous effect on mortality. Then, we applied a second Cox regression using a backward stepwise procedure to identify the most parsimonious model on risk factors for mortality in incident dementia cases. For each variable in the models, we calculated hazard ratios (HR) and Wald 95% confidence intervals (CI). The proportional hazards assumption was tested for both Cox models using Schoenfeld residuals (39). Kaplan–Meier survival analysis was used to determine median survival times for significant prognostic factors that were evaluated in the Cox regressions. Further, log-rank tests were performed to assess survival differences between groups. Finally, plots were generated to graphically illustrate the differences in the cumulative survival probability. Ethical approval
The study was conducted in accordance with the Declaration of Helsinki (40) and has been approved by the ethics committees of the participating centers. All patients and/or their proxies provided written informed consent.
Results Descriptive characteristics
Among the 3214 subjects at risk in the AgeCoDe cohort after baseline investigation, 523 (16.3%) incidentally developed dementia throughout five follow-up waves covering a time period of up to 9.3 years (Fig. 1). Of the 523 incident dementia cases, 362 (69.2%) were women and 161 were (30.8%) men. The mean follow-up time was 6.8 years (SD = 2.5), while deceased subjects were followed for M = 5.9 years (SD = 2.2) and survivors for M = 9.1 years (SD = 1.3; U = 4286.0, P < 0.001). Mean age at estimated dementia onset was 85.0 years (SD = 4.0), ranging from 76 to 99 years. Men were significantly younger at dementia onset than 261
Roehr et al. women (M = 84.0, SD = 3.8 vs. M = 85.4, SD = 4.1; t(521) = 3.61, P < 0.001). By the end of follow-up 5, 367 subjects (69.2%) had died by a mean age of 87.6 years (SD = 4.3), ranging from 77 to 103 years. Overall, subjects were under observation for 1640.7 person-years. The casefatality rate cumulated in 223.7 per 1000 personyears (95% CI = 201.9–247.8). Among the 523 incident cases, 328 (67.8%) participants were diagnosed with AD and mixed dementia, 87 (18.0%) with vascular dementia, and 69 (14.3%) with other types of dementia. In 39 (7.5%) cases, type of dementia could not be defined due to insufficient information from proxy interviews. Table 1 shows sociodemographic and health characteristics of the participants by survival status. Subjects who had died were significantly more frequently male than female (P < 0.001), had more frequently vascular dementia or another type of dementia than AD or mixed dementia (P < 0.001), were less impaired on the IADL at dementia onset (P < 0.02) and were less often affected by hypertension at dementia onset (P < 0.01). Prognostic factors
Results of the impact of incident dementia on mortality were based on 431 subjects due to missing values in covariates in 92 (17.6%) cases. This sample reduction did not change the proportion of deceased subjects and survivors in the model sample compared with the initial sample [v²(1, 933) = 0.09, P = 0.76] nor did it lead to differences in other characteristics at dementia onset (results not shown). Both multivariable Cox proportional hazards models, the full and the backward stepwise regression, found older age, male gender, vascular dementia and dementia of other types (reference category: AD and mixed dementia), and the absence of SMC as significantly associated with increased mortality in incident dementia (Table 2). In the parsimonious backward stepwise model, every additional year of age yielded a HR of 1.06 (95% CI = 1.03–1.09, P < 0.001), male gender yielded a HR of 1.4 (95% CI = 1.1–1.8, P < 0.01) as opposed to women, vascular dementia yielded a HR of 1.5 (95% CI = 1.1–2.0, P = 0.01), and other types of dementia yielded a HR of 1.7 (95% CI = 1.2–2.3, P < 0.01) in contrast to AD and mixed dementia. The presence of SMC without related worry resulted in a HR of 0.7 (95% CI = 0.5–0.9, P = 0.01) and SMC with related worry in a HR of 0.7 (95% CI = 0.5–0.9, P < 0.01) in reference to the absence of SMC. 262
Table 1. Sociodemographic and health characteristics of the study sample at onset of incident dementia by mortality status
Variables†
All incident dementia cases (n = 523)
Deceased (n = 367)
Age at onset in years, mean (SD) Total 84.99 (4.03) 84.98 (4.09) Female 85.41 (4.05) 85.54 (4.04) Male 84.04 (3.83) 83.97 (3.99) Age group at onset in years, n (%) 75–79 57 (10.9) 42 (11.4) 80–84 215 (41.1) 151 (41.1) 85–90 186 (35.6) 125 (34.1) ≥90 65 (12.4) 49 (13.4) Gender, n (%) Female 362 (69.2) 236 (65.2) Male 161 (30.8) 131 (81.4) Education, n (%) Low 341 (65.2) 235 (64.0) Middle 128 (24.5) 91 (24.8) High 54 (10.3) 41 (11.2) Marital status, n (%) Single 33 (6.4) 25 (7.0) Married/cohabiting 180 (35.0) 136 (37.9) Divorced 26 (5.1) 18 (5.0) Widowed 275 (53.5) 180 (50.1) Accommodation, n (%) Private home 469 (90.0) 336 (91.8) Residential care 26 (5.0) 16 (4.4) Nursing home 26 (5.0) 14 (3.8) Type of dementia, n (%) Alzheimer’s and mixed 328 (67.8) 207 (62.0) Vascular dementia 87 (18.0) 72 (21.6) Others 69 (14.3) 55 (16.5) Subjective memory complaints (SMC), n (%) No 147 (28.5) 114 (31.7) Yes, without 196 (38.1) 131 (36.4) related worry Yes, with related 172 (33.4) 115 (31.9) worry Mini-Mental State 25.81 (2.24) 25.89 (2.32) Examination (MMSE), total score, mean (SD) IADL‡, total score, 4.04 (1.29) 4.12 (1.25) mean (SD) Comorbidity, n (%) Diabetes mellitus 129 (25.9) 91 (26.0) Hypertension 320 (64.4) 212 (60.6) Cardiac arrhythmias 151 (30.4) 101 (28.9) Coronary heart disease 160 (32.1) 109 (31.1) Myocardial infarction 41 (8.2) 29 (8.3) 65 (13.1) 48 (13.7) Peripheral artery occlusive disease (PAOD) Stenosis (of afferent 17 (3.4) 12 (3.4) brain vessels) Transient ischemic 66 (13.3) 43 (12.3) attack (TIA) Stroke 41 (8.2) 29 (8.3) Hyperlipidemia/ 238 (47.8) 159 (45.4) hypercholesterolemia Kidney insufficiency 80 (16.1) 51 (14.6) Hepatic disease 29 (5.8) 21 (6.0) Depressive symptoms§, 3.43 (2.78) 3.42 (2.79) total score, mean (SD) Smoking, n (%)
Survivors (n = 156)
P value (group difference)
85.00 (3.91) 85.16 (4.08) 84.35 (3.04)
0.95 0.40 0.63
15 (9.6) 64 (41.0) 61 (39.1) 16 (10.3)
0.58
126 (34.8) 30 (18.6)
