| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
SPECIAL SECTION |
Department of Gerontology and Care for the Elderly, Lund University, Sweden.
Address correspondence to Bo Hagberg, PhD, Hanö, S-29407 Solvesborg, Sweden. E-mail: bo.lennart{at}mbox306.swipnet.se
 |
A |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
Methods. The Swedish Centenarian Study followed 100 centenarians from the age of 100 to death of the entire cohort, by age 111 years. A biomedical, psychological, and social multivariate survival analysis was performed based on factors identified as important in earlier studies of older adults. Latent Variable Partial Least Square Estimation (LVPLS) Soft Modeling was used to test the hypothesized predictions of survival in centenarians.
Results. Fewer predictors for survival were found in centenarians than were observed in studies of younger elderly persons. Survival after age 100 was dependent mainly on better baseline physical reserve, as measured by body mass index and body weight, and better baseline physical and cognitive function, as measured by activities of daily living and verbal ability/spatial orientation, respectively.
Conclusions. Individual characteristics such as physiological reserve, present health and functional status, as well as chance appear important for centenarian survival. Hereditary factors, social relationships, marital status, and personality did not contribute to survival prediction in this exceptional age group. From a theoretical point of view, our data suggest that, in very old age, stochastic determinants may dominate over programmed factors (e.g., family longevity) in determining survival. More research is needed to assess survival factors at exceptional ages.
Key Words: Centenarian • Longevity • Survival • Prediction • Risk factor
Classical survival studies such as the Bonn Study (3–6), the Duke Study (7), the Dalby 67+ Study (8–10), and the Lund 80+ Study (11,12) have baseline ages of as young as 60 years and as old as 80 years. Most studies use a variety of biological, psychological, and social predictors. For example, in our own study, the Dalby 67+ Study, 67-year-olds were followed for survival until 87 years of age, and gender-specific predictors were found (8). For women, early-death age of mother, low social rigidity (i.e., flexible lifestyle), smoking, field dependency (dependent personality), and need for social help predicted early death. For men, early death was predicted by positive attitude to old age home, high social class, comorbidity at baseline age, low life satisfaction, and high number of self-reported diseases earlier in life. Furthermore, when followed until 92 years of age, dementia was found to be a strong risk factor for early death (13).
However, longitudinal studies of coronary heart disease mortality in monozygotic and dyzygotic twins show that the relative risk of either genetic or environmental factors is highest in middle age and declines to parity at about age 90 years (14). This change in predictive capacity of midlife risk factors can also be shown for other diseases such as Alzheimer's disease (15). Thus, mortality selection can be demonstrated on a number of health dimensions, and risk factors change with age.
Little is known, however, about risk factors for mortality in exceptional survivors. Elderly persons who survive to exceptional ages, such as centenarians, may be in better health compared to younger populations when they are younger and may be more resistant to traditional risk factors from midlife or young elderly age (1,2,16). In contrast, after onset of a serious disease there may be a shorter remaining lifetime, which reduces disease prevalence at later ages. Hence, the evidence suggests that centenarians may exemplify compression of the morbidity model suggested by Fries (17).
There is little work to date, however, to assess whether survival predictors differ at exceptional ages. A study of French centenarians focused on biomedical survival factors from age 100 (18) in 800 French centenarians. Continued living in the community, good physical and functional health, no cognitive impairment, somewhat overweight, above-normal cholesterol, and regular diastolic blood pressure predicted survival. Poon and colleagues (19) found in the Georgia Centenarian Study (N = 137) a higher risk for mortality for men, whites, and persons with inability to perform personal activities of daily living (ADLs), such as dressing and bathing.
Furthermore, persons with greater triceps skin fold thickness and lower IQs had a reduced risk of mortality. Poon and colleagues concluded that, even in extreme old age, individual characteristics exert systematic influences on survival.
The findings in these studies have some limitations. First, representativeness can be questioned because both study samples are drawn from patient lists of practicing physicians. In the case of the Georgia Study, there is also a cutoff limit with regard to cognitive functioning, excluding those centenarians with a score < 24 on the Mini-Mental State Examination (MMSE). Second, the studies focused on medical or psychosocial variables at baseline rather than on an interdisciplinary approach. Therefore, the present study attempts to remedy some of these shortcomings using biomedical, psychological, and sociological predictors in a multivariate analysis.
|
METHODS |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
For this study, 164 centenarians were available. This is the total population of centenarians born between 1887 and 1891 living in the southwest county of Sweden. Within 2 months of their 100th birthday, 21 centenarians were lost (16 dead; 5 lost records), leaving 143 eligible individuals. Of these, 100 individuals (82 women, 18 men) agreed to participate and were examined (see Figure 1).
|
Sociological data (n = 77–100) were gathered concerning marital status, type of housing, socioeconomic status (SES) in terms of main occupation, satisfaction with professional life, education, social network (quantitative and qualitative contact frequencies), feelings of loneliness according to the UCLA scale (21), and formal/informal support (20). The percentage of missing information in the sociological part of the study varied from 0% to 14%, apart from some questions regarding loneliness, where the missing information was 26%.
Medical examination (n = 97–100) was completed by a geriatrician. The examination included a medical history, questions on health habits (diet, smoking status, alcohol consumption), and a physical examination including blood pressure in arms and legs, midarm circumference, triceps and subscapular skinfolds, body composition [using an impedance technique (22)], hearing assessment, visual acuity, ADL function (23), and general mental function according to the Berger scale (24).
Psychological data included detailed assessment of cognition using tests of verbal ability (25), memory tests (26), reaction time test, personality test (27–29), and a semistructured interview on life quality (30), behavior observation (31), and assessment for dementia (20). The number of persons examined with psychological tests ranged from 43 (reaction time test) to 100 (psychological assessment battery). These partially missing data were due to substantial visual and/or hearing problems.
Structural Model for Survival
The hypothesized conceptual model (see Figure 3) based on literature cited in the introduction was tested with Latent Variable Partial Least Square Estimation (LVPLS) Soft Modeling (32,33). Structural equation modeling with partial least square estimations was developed by Wold (34) for situations in which data do not meet the highly restrictive assumptions of maximum likelihood procedures such as LISREL (32). The advantage of PLS over ordinary multiple regression analyses is that it allows the estimation of latent variables in structural equation modeling. PLS is more appropriate for modeling when the purpose is to predict an outcome (e.g., survival) rather than to assess an overall theoretical model (35).
|
|
RESULTS |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Disease Prevalence
The prevalence of severe diseases was relatively low. In 39%, a disorder of the circulatory system was found. Thirty-nine percent of the women and 11% of the men had suffered at least one hip fracture. Twenty percent had good hearing and vision. Twenty-seven percent suffered from dementia according to Diagnostic and Statistical Manual of Mental Disorders Third Edition, Revised (DSM III-R) criteria (20).
Physical, Cognitive, and Social Function
Fifty-two percent managed ADLs with or without minor assistance according to the Katz ADL index, that is, they were independent (category A), needed help with bathing/showering (category B), or needed help with dressing/undressing (category C). Means on cognitive tests (word list, digit span, learning test, and memory) were lower than the means in the 70- to 80-year-old groups (36). The variation in performance was exceptionally widespread. In general, lower mean cognitive performance was found in the older elderly than in the younger elderly. However, a wide range of performance between individuals was found in semantic or experientially related abilities, which tend to be better maintained over the life span. In contrast, a smaller range of performance was found in fluid or process-related abilities. These abilities generally showed a downward age-related trajectory (37). Personality profiles (Minnesota Multiphasic Personality Inventory [MMPI]) (20) indicated that the centenarians were more responsible, capable, easygoing, and less prone to anxiety than was the population in general (28).
Survival Analysis
An 11-year follow-up for survival showed that all centenarians were dead. The mean survival from age 100 years was 2.03 (standard deviation 1.94) years. The majority was deceased within a period of a few years and with few surviving beyond 105 years (see Figure 2).
|
Decreased mortality risk was predicted by wine consumption, triceps and subscapular skinfold, and lean weight. More specifically, individuals reporting moderate wine consumption (72% of men and 48% of women were presently having wine or beer 1–3 times a week) had an increased survival of 1.30 years. Skinfold thickness (millimeters) showed a small but significant associated with survival, as did body composition (lean weight in kilograms). Altogether there were seven medical predictors found. Noteworthy is that smoking, liquor consumption, fish consumption, body mass index, ADLs, and high-density lipoprotein (HDL) were not independently predictive of survival in this age group. Neither father nor mother's age at death were predictive of survival (Table 1).
|
|
|
The model contains five latent predictor variables summing up relevant markers of survival from three domains: medical, psychological, and social factors that explain 34% of the variation in these domains among the centenarians. The fit of the model is given by the root mean square of the covariance between the manifest variable residuals and the latent variable residuals and is for the present model.08, which is an acceptable fit between data and the proposed model.
"Health" was the strongest predictor of survival, with a path coefficient of.39. Health is defined as taking few drugs, drinking wine to a moderate degree, not having hearing and vision impairments, better (retained) lean body weight (the difference between total body weight and fat weight in kilograms), and high HDL cholesterol.
ADL functioning is central because it indirectly affects survival through strong loadings on both health (.59) and cognition (.52) but also because it has a direct (although moderate) influence on survival (–.11).
Cognition, as defined by verbal understanding, memory, and learning capability, had a moderate influence (.07) on survival. It is noteworthy that cognition is well measured with strong loadings in the markers and a high degree of explained variance (.45), although with a low but significant effect on survival.
Social experience was the second most important predictor, with a path coefficient of.10. Its markers are economy, education, job satisfaction, marital status, and living conditions (with economy, education, and living conditions as the strongest markers). It is important to note that all markers are life-span-related experiences except present living conditions. More precisely, the social experience predictor shows that centenarians who live the longest after 100 years of age had the following factors in common: similar economy compared to their peers when growing up, education < 7 years and having attended school only sporadically (similar to peers), low job satisfaction during the early working period, being married or cohabiting, and presently not living in an institution.
Much of the variance in centenarians' health (R2 =.41), cognition (R2 =.45), and social experience (R2 =.34) is explained, but that these factors together only explain survival to a modest degree (R2 =.14). Thus, there are other markers not covered in this study that provide significant information on survival after 100 years of age. Most likely this relatively low explanative power is due to the fact that the model is based on findings from survival studies using younger elderly persons as the baseline population. There also appears to be a great deal of hazard determining the outcome at exceptional ages. For instance, a common cold in a frail centenarian with low immunologic defense may be fatal.
Survival promoting (salutogenic) and survival reducing (pathogenic) markers are presented in Figure 3. We found that high lean body weight; high HDL cholesterol; moderate wine consumption; independence in ADLs; good verbal ability, memory, and learning ability; better economy then peers when growing up; being married or cohabiting; and independent home living are survival-promoting factors. Mortality risk factors were higher number of drugs taken (more drugs increased risk), hearing and/or vision impairment, low ADL capability, education > 7 years, low job satisfaction, and living in an institution.
|
DISCUSSION |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Initial analyses of baseline data revealed single-variable predictors of survival until the age of 111 years. In the medical domain, survival-promoting variables such as wine consumption, lean body weight, and HDL were found. Survival-reducing factors such as dementia symptoms, impaired vision and/or hearing, and higher number of drugs taken were also found. In the psychological domain, mainly survival-promoting factors were found. These included verbal capacity, memory, and learning. In the sociological domain, satisfaction with professional life was survival promoting, whereas institutional care was survival reducing. From a personal resource perspective, these factors can be viewed as assets in a continuous attempt to deal with reality through adaptive behavior.
One way to look at adaptation to aging has been the "optimal selection with compensation theory" (38). In this very old age group, we have to assume that most of the individuals are very frail (39–42). Thus, resources for compensation may be limited; therefore, risk factors might be more dangerous and at the same time protective or compensatory factors might be less effective. Given that relatively few factors are predictive of mortality for centenarians compared to younger groups, one might also assume that, because of underlying frailty, there is a great deal of "accidental" death—that is, some centenarians may experience a fatal fracture after an accidental fall, whereas others may "catch" influenza—both potentially lethal and stochastic in nature.
Recognized predictors for survival in the 60s, 70s, 80s, and 100s are functional capacity, as measured by ADLs or institutionalization. This aspect of functional capacity probably reflects the fact that current health and functional status is an important survival predictor at every stage of life. Given the attrition that takes place because of poor health and death from serious diseases, negative predictors at younger ages might become less important at older ages because the healthier "disease-resistant" phenotypes selectively survive at older ages. This phenomenon has been referred to by Perls (43) as "selective survival" or "survival of the fittest."
As increasing selection occurs in older persons, the emphasis may change from risk factors for particular diseases to "longevity-promoting" factors. Several such hypothetical factors may be considered. First, one of these factors may be cognition. In many studies, cognitive ability (intelligence especially), as in the 80+ Study (11,44), maintaining a high level of "fluid intelligence" (i.e., process-related abilities), seems to be important. In the centenarians, cognition is still important but more in terms of crystallized intelligence (i.e., semantic or experience-related abilities) as well as memory and learning capacity. One hypothesis is that differences between exceptionally old persons are caused by enduring aging processes in all physical, social, and psychological functions (45). If this is correct, then it is likely that maintained cognitive capacity has served as a resource for adapting to changing life circumstances (3–5).
Second, given the differential decline in cognitive function during the aging process, it seems natural that a fluid intelligence capacity, being predictive of survival at 80, is replaced by the more decline-resistant crystallized ability, as a predictor of survival in very old persons (46).
Third, there is a sociodemographic domain that is predictive of survival in all age groups. For example, social class at age 67 years, SES in both the Bonn and Duke Studies at ages 60–75, educational level at age 80 years, and economy/education/past job satisfaction in the centenarians, can all be related to the individual's social–historical environment. This interpretation agrees with the assumption of Rott and colleagues (6) that longevity is the result of enduring aging processes acting throughout life as a combination of individual resources and environment (47,48). It is also compatible with the assumption of Martin and colleagues (49) that historical events, both micro and macro, have shaped the individual.
How do survival predictors differ between age groups? Forette (50) has drawn attention to gender (in combination with health), body weight, blood pressure, and cholesterol. These risk factors tend to change their significance in the sense that what was a positive risk factor in younger adults becomes a negative risk factor (associated with increased longevity) in the oldest old. Thus men, if long-lived, generally appear healthier than women. This assumption cannot be examined in the present study with so few men.
Body weight-associated mortality risk declines with age for both men and women (41). In the current study, higher body weight was a survival factor as seen in skinfold and lean weight measures. The reasons for this are unclear, but sarcopenia (loss of muscle mass) with old age appears to be linked to increased frailty and mortality, and maintenance of muscle mass suggests better survival in old age (42).
In the current study, the structural model shows that, among the variables defining health, wine consumption and HDL are both included with high factor loadings. This finding is consistent with those from studies that show that "good" cholesterol is protective and is increased by moderate wine consumption. These two variables may interact to contribute to good health in the very old population. This interaction is primarily a reflection of a lifestyle pattern and appears to hold even at 100 years of age.
Cognition was shown to be important both in the bivariate analysis and in the structural model. In the bivariate prediction, dementia, as measured by the Berger scale, was a significant predictor for mortality. However, in the structural model analysis it does not come out as such; rather, it is that cognitive abilities are predictive of survival. They are well measured with high factor loadings in the external model, explaining much of the cognitive variance in the material but with a moderate contribution to survival. One possible explanation for this finding might be that the number of persons with dementia is rather low in this study—only 27% as identified by International Classification of Diseases 10th Revision (ICD-10) or DSM IV criteria. The cognitive domain is defined, not on the few probands at risk, but rather on the 73% not suffering from dementia. In this group, a varying degree of cognitive capacity contributes in a moderate way (.07) to future survival.
From the social domain, there are several lifestyle factors that promote survival beyond 100 years of age. Unexpectedly, it is an education < 8 years and low prior job satisfaction that promote survival. In the structural model, these two factors go together. This finding, which at first seems to be counterintuitive, might be explained by the use of a dichotomous variable for education, i.e., most of the participants scored < 8 years for education but this was at least some years of education, when in the early 20th century in Sweden, school was attended only when the youngsters were not needed for work at home. Most likely, the kind of work they had to do was manual labor and not something they enjoyed, which might explain the low job satisfaction.
Taken together, the social variables emphasize the importance of the life-span perspective perhaps better than the other variables. These variables also support one of the basic assumptions from the survival models, namely that it is the aggregated experiences during the life span that account for the wide variation in the aging pattern, with regard to longevity.
Finally, it should be noted that, in the structural model analysis, we obtained good measurements of health, cognition, and social experiences, with high factor loadings in the outer model. In the inner model, a high degree of variance was explained in each of the latent variables, except for survival. Only 14% of the variation in survival beyond 100 years of age could be explained.
There are two likely explanations for this finding: (i) Other more appropriate variables could have been chosen as independent predictors, or (ii) perhaps "accidental" risk factors for death (stochastic factors) may be more active in this age group. Frailty, with concomitant deficits in muscle function, immune system, and sensory abilities (among other factors) can predispose to fatal events such as fractures, infections, or other minor diseases that can cause serious problems or death in such exceptional survivors (41).
|
A |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
We are indebted to Professor Emeritus Sven-Marten Samuelsson, who completed the medical examinations; Betty Bauer-Alfredsson, PhD, who completed the psychological testing; Vibeke Hortstmann, PhD, for assistance in the statistical analyses; and Cheryl McCamish-Svensson, PhD, and Peter Martin, PhD, for English revisions.
|
F |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Received December 8, 2007
Accepted August 16, 2008
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  B. J. Willcox, D. C. Willcox, and L. Ferrucci Secrets of Healthy Aging and Longevity From Exceptional Survivors Around the Globe: Lessons From Octogenarians to Supercentenarians J. Gerontol. A Biol. Sci. Med. Sci., November 1, 2008; 63(11): 1181 - 1185. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
