Terminal Decline and Markers of Cerebro

Journal of Gerontology: PSYCHOLOGICAL SCIENCES
2002, Vol. 57B, No. 3, P268–P276
Copyright 2002 by The Gerontological Society of America
Terminal Decline and Markers of Cerebroand Cardiovascular Disease: Findings From a
Longitudinal Study of the Oldest Old
Linda B. Hassing,1,2 Boo Johansson,1,2 Stig Berg,1 Sven E. Nilsson,1 Nancy L. Pedersen,3
Scott M. Hofer,4 and Gerald McClearn5
of Gerontology, University College of Health Sciences, Jönköping, Sweden.
2Department of Psychology, Göteborg University, Sweden.
3Department of Medical Epidemiology, Karolinska Institute, Stockholm, Sweden.
4Department of Human Development and Family Studies, Pennsylvania State University, University Park.
5Center for Developmental and Health Genetics, Pennsylvania State University, University Park.
The purpose of this study was to examine the cognition–survival relationship among nondemented individuals in
late life. The longitudinal design included three examinations at 2-year intervals. At baseline, 466 individuals (age
range 5 80–98) were examined. During the 6 years of follow-up, 206 individuals died. Four survival groups were
defined on the basis of mortality prior to the subsequent measurement occasion. Tests of cognitive functioning encompassed the domains of crystallized knowledge, inductive reasoning, visuospatial ability, short-term memory,
episodic memory, and speed. Significant associations were found between cognitive performance at baseline and
subsequent survival. After adjusting for stroke and markers of cardiovascular disease, the authors found that
only three out of six cognitive domains remained significant predictors of survival. The longitudinal analyses revealed limited evidence for an accelerated decline prior to death. The main results suggest that level of cognitive
performance in late life is associated with proximity to death, that this relationship is longstanding, and that it is
partially influenced by compromised cardio- and cerebrovascular functioning.
T
HERE is evidence that poor cognitive performance
among elderly individuals is associated with impending
death (for reviews see Berg, 1996; Siegler, 1975; Small &
Bäckman, 1999). However, several issues of the cognition–
survival relationship remain to be clarified. For example, is
the relationship specific or pervasive across groups differing
in age and across cognitive tasks? Is there an accelerated cognitive decline or drop before death? Is the association between
cognition and survival mediated by diseases and health-related
changes? The present study addresses some of these questions
in a longitudinal study of nondemented very old individuals.
It has been suggested that only crystallized knowledge
(e.g., verbal ability, semantic memory) or abilities that are
unaffected by aging are indicative of impending death (Bosworth & Schaie, 1999; White & Cunningham, 1988), whereas
others hold that all cognitive abilities are affected (Ljungquist,
Berg, & Steen, 1996; Maier & Smith, 1999). Mortality effects have been reported in age-maintained abilities including verbal abilities (Bosworth & Schaie, 1999; Reimanis &
Green, 1971; Siegler, McCarty, & Logue, 1982; White &
Cunningham, 1988) and short-term memory functioning
(Johansson & Berg, 1989), as well as in age-sensitive abilities such as psychomotor speed, verbal fluency, and episodicmemory functioning (Bosworth & Schaie, 1999; Botwinick,
West, & Storandt, 1978; Hassing, Small, von Strauss,
Fratiglioni, & Bäckman, 2002; Maier & Smith, 1999; Siegler et al., 1982; Small & Bäckman, 1997). Thus, the evidence remains inconclusive as to whether the cognitive correlates of mortality are task specific or pervasive.
Another issue concerns the time frame of the predictive
P268
period. Although there is evidence that differences in level
of cognitive functioning are associated with subsequent survival, it is not clear if there is a pronounced decline in function closer to time of death (i.e., terminal drop), or if there is
a continuous decline for several years preceding death (Palmore & Cleveland, 1976). To examine this we need longitudinal studies of individual trajectories that can be related to
distance to death. Thus, it is important to distinguish between findings based on differences in level of performance
at one point in time and subsequent mortality, and differences in change across two or more occasions.
Cross-sectional findings have shown that decedents perform at a lower level than survivors several years before
dying. This has been shown for 2 to 3 years (Hassing et al.,
2002; Small & Bäckman, 1997; White & Cunningham, 1988),
7 years (Maier & Smith, 1999), and up to 14 years before
death (Bosworth & Schaie, 1999). Findings from longitudinal studies are somewhat mixed. Johansson and Berg (1989)
found that decline in primary memory across three measurements over a period of 9 years was strongly associated with
subsequent survival time. Similar findings were reported for
episodic memory by Deeg, Hofman, and van Zonneveld
(1990). However, in a study by Bosworth and Schaie (1999)
based on two assessments over a 7-year period, no differences were found in rate of change in 11 of 13 cognitive
measurements. In further analyses of the same data set it
was concluded that distance to death was significantly predicted by level of cognitive performance but not by change
in cognitive performance (Bosworth, Schaie, Willis, & Siegler, 1999).
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
1Institute
COGNITIVE PERFORMANCE AND SURVIVAL
METHODS
Sample
The participants were selected from the ongoing population-based longitudinal study Origins of Variance in the
Old-Old (OCTO-Twin Study; McClearn et al., 1997). The
OCTO-Twin Study began with a sample of 702 individuals
(351 twin pairs), aged 80 years and older, drawn from the
Swedish Twin Registry (Cederlöf & Lorich, 1978). The sample was found to be similar to a population of Swedish singletons of the same age (Simmons et al., 1997). The participants were assessed three times at 2-year intervals. The first
wave (T1) proceeded between 1991 and 1993, the second
wave (T2) between 1993 and 1995, and the third wave (T3)
between 1995 and 1997.
All individuals were examined in their home by experienced registered nurses specifically trained for the study and
continuously supervised. A complete testing session, including
rest periods, took about 3.5 to 4.0 hr. Covariates of interest
for the present study included demographic information
(age, gender, and education), physical health (see Medical
Information), as well as cognitive functioning (see Cognitive Measures).
The following individuals were excluded: People with
dementia at the start of the study (n 5 81) according to the
Diagnostic and Statistical Manual of Mental Disorders (3rd
ed., rev.; American Psychiatric Association, 1987), those
who became demented under the follow-up period (n 5 68),
persons who declined participation in the cognitive testing
(n 5 26), and persons who discontinued participation (dropout for reason other than death) in T2 and T3 (n 5 49). The
total sample that was included in this study comprised 466
persons.
The group that declined participation at follow-up (n 5
49) did not differ from the total sample included in the study
(N 5 466) concerning age, gender, education, or performance
on the cognitive tasks (ps . .05). However, stroke and heart
failure were more frequent among the dropouts (ps , .05).
Mortality surveillance based on information from the
Swedish Death Registry was gathered continuously throughout the study period (1991–1999). Four survival groups
were defined on the basis of subsequent mortality (see Table
1). Sixty-three persons were deceased by the time of the
second examination, another 67 persons had died before the
third examination, and at the end of the study period (September, 1999) another 76 individuals were deceased. Thus,
a total of 260 survivors and 206 decedents for whom the
death dates were known had been tested during at least one
of the three waves. The primary causes of death, gathered
from death certificates, were vascular disease (62%), malignant tumors (19%), infectious diseases (10%), infirmities of
old age (6%), and other causes (3%; e.g., suicide, accidents,
unknown cause). Frequency of primary causes of death for
those who declined participation at follow-up did not differ
from those who continued participation.
Measures
Medical information.—A physician (Sven E. Nilsson)
made the final diagnoses on the basis of (a) medical records
containing information dating back at least to 1985, (b)
medicine use, and (c) self-reported information (from T1
through T3) concerning health and diseases. Diagnoses
were classified and based on ICD-10 criteria. The conditions reported on here are completed stroke, transient cerebral ischemic attack (TIA), myocardial infarction, congestive heart failure, angina pectoris, and arterial hypertension.
In the following analyses, myocardial infarction and congestive heart failure are referred to as markers of CVD.
Stroke. A diagnosis of stroke is mainly based on medical
exam information in medical records. Self-reports of stroke
were, however, accepted as a valid diagnosis despite the
lack of record confirmation in a few cases (about 10%).
TIA. TIA refers to a condition with duration shorter than
24 hr. A diagnosis of TIA is based on medical exam as well
as on self-reports.
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
Although several studies have examined the relationship
between cognition and survival, less effort has been given to
the underlying mechanisms for the relationship. In a recent
study by Smits, Deeg, Kriegsman, and Schmand (1999),
several cognitive measures were related to mortality risk, but
when self-rated health, medicine use, physical performance,
and specific chronic diseases were controlled, only the measures of processing speed, problem solving, and episodic
memory remained as significant predictors of mortality.
A distinction made in the study of aging is that between
normative or primary age-related change and change related
to compromised health and disease. Dementia affects a substantial proportion of people in late life and represents the
most devastating disease category for cognitive functioning.
The dramatic effects of the dementing illnesses are well
documented. Also, it is well established that dementia increases the risk of mortality (e.g., Johansson & Zarit, 1997).
Because dementia produces severe cognitive decline, it accounts for a substantial amount of the variance in the overall
cognition–survival relationship. For this reason, many
studies have deliberately excluded individuals with dementia (Bosworth & Schaie, 1999; Bosworth, Schaie, & Willis,
1999; Bosworth, Schaie, Willis, & Siegler, 1999; Hassing et
al., 2002; Small & Bäckman, 1997) to investigate the robustness among nondemented individuals. A potential, but
less investigated, source of cognitive decline and death in
the elderly is that of cardiovascular disease (CVD). Although it is well established that CVD elevates the risk for
stroke and that severe cerebrovascular disease is associated
with lower levels of cognitive performance (Elias, Elias, &
Elias, 1990), the impact of these factors on the cognition–
survival relationship has not been systematically studied.
The purpose of the present study was to examine the level
of cognitive functioning and longitudinal changes in cognitive performance in relation to time to death in individuals
that remain non-demented. More specifically, our aims were
to (a) examine if types of cognitive abilities are differentially
predictive of mortality, (b) study whether markers of cerebrovascular disease and CVD influence the cognition–survival
association, and (c) examine if there is a pronounced decline
in cognitive functioning associated with time to death.
P269
P270
HASSING ET AL.
Table 1. Participant Characteristics by Survival Group
Variable
Deceased Between
T2 and T3
(n 5 67)
Deceased Between
T1 and T2
(n 5 63)
82.66 (2.40)
70
7.22 (2.21)
9 (24)
7 (19)
11 (29)
16 (41)
17 (43)
46 (119)
—
83.57 (2.61)
51*
7.61 (2.86)
22 (17)*
8 (6)
25 (19)*
38 (29)*
32 (24)*
49 (37)
5.73 (1.19)
83.77 (3.52)*
57
7.21 (2.26)
36 (24)*
3 (2)
34 (23)*
42 (28)*
21 (14)
49 (33)
2.98 (0.57)
84.28 (3.80)*
59
7.05 (2.66)
37 (23)*
11 (7)
41 (26)*
27 (17)
24 (15)
38 (24)
1.04 (0.58)
Note: TI 5 Wave 1, T2 5 Wave 2, T3 5 Wave 3; TIA 5 transient cerebral ischemic attack.
*p , .05 as compared with the survivors.
Myocardial infarction. The diagnostic information is
mainly drawn from the records in which a diagnosis requires laboratory evidence of marker enzymes in conjunction with the typical clinical manifestations and/or an ECG
pattern with specific QRS aberrations. Self-reports about
myocardial infarction were accepted in a few cases in which
there was a lack of medical-record information (about
10%). These self-reports include histories with a hospital
observation period for suspect infarction but with an unresolved diagnosis or other instances in which the diagnostic
criteria are not fully met.
Congestive heart failure. The diagnosis is mainly based
on medical records presenting the diagnosis or explicit symptoms of insufficient cardiac function characterized by edema
and dyspnoe, including treatments with diuretics, digitalis,
and/or angiotensin converting enzyme blockers, providing
evidence for the diagnosis. Self-reports without medicalrecord confirmation were included for about 25%.
Angina pectoris. A diagnosis is based on explicit information in medical records supplemented with information
about the use of prescribed nitrates for suspected angina
pectoris. A myocardial infarction diagnosis is registered in
cases in which angina symptoms were observed in temporal
connection with a myocardial infarct.
memory, episodic memory, and speed. We constructed composite scores for crystallized knowledge, short-term memory, episodic memory, and speed for each participant by
computing a weighted mean based on the observed test
scores in each ability domain.
Crystallized knowledge. To measure crystallized knowledge we used two tasks from the Wechsler Adult Intelligence Scale (WAIS; Wechsler, 1991): the Swedish version
of the Information Task (Jonsson & Molander, 1964), including questions of general knowledge, and the Verbal Meaning
test, which required finding a synonym to match a target word.
Inductive reasoning. The Figure Logic task (Dureman
& Sälde, 1959) requires the person to identify one figure out
of five in a row that is different in concept from the rest.
Visuospatial ability. In the Koh’s Block Design Test
(Dureman & Sälde, 1959) the participant is presented with
red and white blocks and several patterns written on cards.
The task is to reproduce the pattern shown on the cards by
assembling the proper blocks and arranging them to form
the design shown on the card.
Arterial hypertension. A diagnosis of hypertension is
based on medical-record information of specific treatments
for hypertension and a record/in-person examination systolic blood pressure value higher than 160 mm/Hg and/or a
diastolic value above 95 mm/Hg. The recommendation from
the Joint National Committee on Detection, Evaluation, and
Treatment of High Blood Pressure, suggesting limit values
of 140/90, was considered too inclusive for the present sample of very old individuals.
Short-term memory. In short-term memory, incoming
information is registered and retained in accessible form for
a very short period of time after the input (a few seconds).
The capacity of short-term memory is roughly equivalent to
the number of items that a person can reproduce accurately
immediately after hearing them. To measure the passive part
of short-term memory, the Digit Span Subtest (WAIS; Wechsler, 1991) was used. The task is to repeat a sequence of digits in the same order as originally presented (Digit Span Forward). To measure the active part of short-term memory (i.e.,
working memory) we used the Digit Span Backward, in which
the task is to repeat a sequence of digits in the reverse order.
Cognitive measures.—Eleven tests were used to measure cognitive functioning. The tests represent crystallized
knowledge and five abilities associated with fluid abilities;
that is, inductive reasoning, visuospatial ability, short-term
Episodic memory. Episodic memory deals with everyday experiences encoded in a particular time and place. It
enables conscious recollection of personal events and episodes from one’s personal past (Tulving, 1983). The tasks
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
Age [M (SD)]
Gender (% female)
Years of education [M (SD)]
Completed stroke [% (n)]
TIA [% (n)]
Myocardial infarction [% (n)]
Congestive heart failure [% (n)]
Angina pectoris [% (n)]
Arterial hypertension [% (n)]
Years of survival after T1 [M (SD)]
Survivors
(n 5 260)
Deceased Between
T3 and the End of
the Study
(n 5 76)
COGNITIVE PERFORMANCE AND SURVIVAL
reflecting episodic memory included a Swedish Prose Recall task similar to logic memory in WAIS, Thurstone’s Picture Memory (Thurstone & Thurstone, 1949; recognition of
drawn pictures), and Memory-in-Reality Test (Johansson,
1988/1989; free recall of 10 objects).
P271
of longitudinal analyses, data from T1 and T2 are compared
across the three survival groups. In the second set of analyses, data from T1 through T3 are compared across the two
groups that survived all three waves.
RESULTS
Data Analyses
The statistical package used for all analyses was the Statistical Package for Social Sciences (Version 10.0; SPSS, 2000)
for PCs. In the first set of analyses (cross-sectional), level of
cognitive performance at baseline was compared across survival groups using univariate analysis of variance (ANOVA).
Next, cognitive performance at baseline was examined in relation to survival. For this purpose Cox proportional hazard
regression analyses (Cox, 1972) were performed. Cox analyses are well suited to analyze data based on individuals that
are in the study for different lengths of time because they
model the hazard rate rather than mortality status as the outcome variable. As mortality status assessment was discontinued at the end of the study, right censoring was applied (censored cases are those for which the event has not yet occurred).
Two nested Cox regression models were evaluated. In
both models mortality was used as the outcome variable,
with varying number of covariates and cognitive domains as
the predictor variables. The two models were repeated for
each cognitive domain separately. Model 1 included age,
gender, and education as covariates. Model 2 was extended
by adding two variables based on physician evaluation: one
that indicated a history of stroke (0 5 no history of stroke; 1 5
history of stroke) and one that indicated markers of CVD
(0 5 no history of CVD; 1 5 history of CVD).
Parameter estimates obtained from the Cox regression
models were used to assess the associations between cognitive functioning and survival. Odds ratios (OR; obtained by
exponentiating the parameter estimate) are reported with a
95% confidence interval (CI). An OR of 1.00 means that
there is no increased mortality risk, whereas an OR of, for
example, 1.30 indicates a 30% increased mortality risk. In
the Cox regression models, all measures were standardized
(equal means and standard deviations across tasks) prior to
analysis to facilitate comparisons of odds ratios across
tasks. Also, all cognitive scales were reversed such that high
scores indicated low performance.
In the second set of analyses (longitudinal), cognitive
data from T1 through T3 were analyzed to examine if the
three survival groups showed the same pattern of change
across time. For this purpose, mixed ANOVAs were conducted for each cognitive task using a repeated measurements design. The group composed of individuals who died
between T1 and T2 and were assessed at only one occasion
was excluded from the longitudinal analyses. In the first set
Sample Characteristics
The sample characteristics are reported in Table 1. A oneway ANOVA comparing the four survival groups revealed
significant age differences between groups, F(3,462) 5
7.48, MSE 5 8.04, p 5 .000, v2 5 .040. A Tukey’s post hoc
test showed that the group with the longest survival time
was significantly younger than the two groups with the
shorter survival time (ps , .05). There was a difference in
gender distribution, x2(3, N 5 466) 5 10.98, p , .05, reflecting more women among the survivors as compared with
the group who died between T3 and the end of the study.
The groups were similar with respect to education. Concerning the diseases, fewer survivors than decedents experienced a stroke, x2(3, N 5 466) 5 43.39, p , .01, but there
were no differences in TIA. Fewer survivors experienced
myocardial infarction than decedents had, x2(3, N 5 466) 5
38.95, p , .001. Congestive heart failure was less frequent
among the survivors as compared with the groups that died
between T2 and T3 and between T3 and the end of the
study, respectively, x2(3, N 5 466) 5 29.27, p , .001. Angina pectoris was more frequent among the group who died
between T3 and the end of the study as compared with the
survivors, x2(3, N 5 466) 5 8.62, p , .05. No differences
were seen in arterial hypertension across groups.
Cognitive Performance at Baseline
Baseline performance across groups is portrayed in Figure 1 as transformed scores (T scores), to facilitate comparisons across domains. Univariate ANOVAs on the cognitive
data showed significant differences across survival groups
in all cognitive domains (for statistics see Figure 1 caption).
Survivors showed superior performance as compared with
the group with the shortest survival time.
Results from the Cox regression analyses, examining the
relationship between cognition and mortality, are reported
separately for each model (see Table 2).
Model 1.—For all analyses, we found that the mortality
risk was significantly greater with increased age (OR between
1.31 and 1.50) and for men (OR between 1.20 and 1.28),
whereas education was not related to mortality risk. Performance level on all cognitive domains, with the exception of
short-term memory, was associated with impending death.
A lower performance level was related to an increased risk
of dying.
Model 2.—For all analyses, mortality risk was increased
by stroke (OR between 2.09 and 2.57) and by markers of
CVD (OR between 2.31 and 2.72). However, after adjusting
for these covariates, we found that the association between
mortality and cognitive performance was reduced to a nonsignificant level for crystallized knowledge and inductive
reasoning (see Table 2).
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
Speed. We used two tasks to measure speed—a modified
version of the Symbol–Digit Substitution Test (WAIS;
Wechsler, 1991), which required a verbal rather than a written response that the digit matched or did not match the
symbol, and the Perceptual Speed Task (Dureman & Sälde,
1959). In the Perceptual Speed Task there is one item at the
far left in a row, and the task is to detect the identical item
among five alternatives on the right as fast as possible.
P272
HASSING ET AL.
Longitudinal Comparisons
The outcome of the longitudinal analyses, conducted to
examine whether there is an accelerated decline prior to
death, is presented in Table 3, and the patterns are graphically shown in Figure 2. The first set of longitudinal analyses compared cognitive data from T1 and T2 across the
three survival groups. A significant decline between T1 and
T2 was found in crystallized knowledge, visuospatial ability, episodic memory, and speed. There were no significant
interaction effects between survival group and time in inductive reasoning, visuospatial ability, short-term memory,
and speed. This indicates that those who deceased did not
decline more rapidly across a 2-year period than those who
survived. There was an exception to this in crystallized
knowledge and episodic memory. In these two domains, individuals with a shorter survival (deceased between T2 and
T3) showed a steeper decline compared with those who survived longer.
Table 2. Results From Cox Regression Analyses Predicting Mortality Status
Model 2
Adjusted for Age, Gender, Education, Stroke,
and Markers of CVD
Model 1
Adjusted for Age, Gender, and Education
Ability
Crystallized knowledge
Inductive reasoning
Visuospatial ability
Short-term memory
Episodic memory
Speed
b
OR
95% CI
p
b
OR
95% CI
p
.20
.18
.35
.11
.35
.34
1.22
1.19
1.42
1.12
1.42
1.41
1.04–1.43
1.02–1.40
1.21–1.66
0.97–1.29
1.23–1.64
1.18–1.68
.013
.029
.000
ns
.000
.000
.08
.14
.22
2.03
.29
.30
1.09
1.15
1.25
0.97
1.33
1.35
0.92–1.28
0.97–1.36
1.06–1.47
0.84–1.12
1.15–1.55
1.12–1.63
ns
ns
.007
ns
.000
.002
Notes: Mortality status was coded as either 0 (survivor) or 1 (deceased). Variables are continuous and in z score metric. Cognitive scales are reversed so that a high
value indicates low performance. CVD 5 cardiovascular disease; OR 5 odds ratio; CI 5 confidence interval. ORs represent relative risk of dying over the follow-up
interval per standard deviation difference independent of other variables in the model.
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
Figure 1. Mean cognitive performance (T score) at baseline across survival groups. Error bars indicate standard errors around the means. For
crystallized knowledge, F(3,441) 5 3.35, p 5 .02, for inductive reasoning, F(3,373) 5 3.11, p 5 .03, for visuospatial ability, F(3,413) 5 9.96, p 5
.00, for short-term memory, F(3,454) 5 3.09, p 5 .03, for episodic memory, F(3,450) 5 12.50, p 5 .00, and for speed, F(3,370) 5 6.44, p 5 .00.
T1 5 Wave 1, T2 5 Wave 2, T3 5 Wave 3.
COGNITIVE PERFORMANCE AND SURVIVAL
P273
Table 3. Summary of Statistics From the Longitudinal Analyses
MANOVA Group 3 Time
Group Effect
Group 3 Time
Time Effect
F
p
MSE
F
p
F
p
N
207.78
22.64
78.77
3.39
5.70
135.36
0.93
1.76
8.72
3.45
14.98
6.29
.396
.173
.000
.033
.000
.002
10.44
10.67
11.10
1.14
1.39
23.12
5.90
0.17
7.22
2.77
9.26
8.99
.016
.680
.008
.097
.002
.003
3.53
0.78
2.12
2.42
3.46
1.42
.030
.461
.122
.785
.033
.243
383
286
330
390
387
288
Group 2 3 Time 3
Crystallized knowledge
Inductive reasoning
Visuospatial ability
Short-term memory
Episodic memory
Speed
305.83
25.55
111.31
4.26
8.44
188.27
2.99
0.43
6.40
0.57
16.12
3.41
.085
.514
.012
.451
.000
.066
12.38
12.94
13.29
1.39
1.55
26.69
15.26
0.07
10.87
11.42
21.15
15.46
.000
.790
.001
.001
.000
.000
0.63
1.06
1.75
0.14
0.73
1.78
.430
.305
.188
.705
.393
.184
314
217
268
313
314
225
In the second set of the longitudinal analyses, data from
T1 through T3 were compared across the two groups that
survived all three waves. A significant decline between T1
through T3 (linear trend) was found in all domains, except
for inductive reasoning (see Table 3). There were no significant interaction effects between survival group and time;
that is, those who deceased had not declined more rapidly
across the 4-year period than those who survived had.
DISCUSSION
The aim of this study was to examine potential survival
effects in level and change across a wide range of cognitive
abilities and to study the impact of stroke and markers of
CVD on the cognition–survival association. The main findings from the cross-sectional analyses (based on data from
the T1) indicated a relationship between survival and level
of performance in all cognitive domains except short-term
Figure 2. Mean cognitive performance 3 survival groups across Time. T1 5 Wave 1, T2 5 Wave 2, T3 5 Wave 3.
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
MSE
Group 3 3 Time 2
Crystallized knowledge
Inductive reasoning
Visuospatial ability
Short-term memory
Episodic memory
Speed
P274
HASSING ET AL.
flecting crystallized knowledge may therefore be more indicative of secondary aging processes, like compromised
cerebrovascular supply. Further, certain cognitive abilities
(i.e., visuospatial ability, episodic memory, and speed) are
affected by other factors as well. One suggestion, out of a
number of potential mediating factors, relates to incipient
dementia disorders. It is interesting to note that episodic
memory functioning is the most salient predictor of incident
dementia (e.g., Howieson et al., 1997; Jacobs et al., 1995);
measures of speeded psychomotor performance have also
been shown to predict incident dementia (Masur, Sliwinski,
Lipton, Blau, & Crystal, 1994). Although our sample comprised only nondemented individuals, it is not unlikely that
individuals in a preclinical phase of Alzheimer’s disease
were included, given the long preclinical phase of this disease (e.g., Linn et al., 1995; Small, Herlitz, Fratiglioni,
Almkvist, & Bäckman, 1997).
After having found that level of cognitive performance at
baseline was related to impending death, our next question
of interest was to explore whether there was an accelerated
decline among decedents as compared with the survivors.
Results based on two measurement occasions across a 2year period and three measurement points across a 4-year
period showed decline in all domains with the exception of
inductive reasoning and short-term memory. Concerning the
question of differential rate in decline between survival
groups, an accelerated decline in the group with the shortest
survival time was observed only across a 2-year interval in
crystallized knowledge and in episodic memory. This was,
however, not seen in the other ability domains and across
the 4-year interval. In a study based on change over a 7-year
period, Bosworth, Schaie, Willis, and Siegler (1999) reported no accelerated decline. However, two studies have
reported steeper decline among decedents (e.g., Deeg et al.,
1990; Johansson & Berg, 1989). Thus, the question still remains open as to whether a greater decline is to be expected
closer to death.
Bringing together the results from the cross-sectional and
the longitudinal analyses, our findings suggest that cognitive performance in late life is associated with proximity to
death and that this relationship can be partially accounted
for by compromised cardiovascular and cerebrovascular
functioning. Thus, we conclude that terminal decline is influenced by disease-related changes that affect cognitive
functioning. The longitudinal analyses showed interaction
effects for survival group and decline in only two cognitive
abilities. Thus, we find only weak support for a differential
decline in relation to impending death. Also, the diseaserelated effects on the cognition–survival association make it
unlikely that all individuals will experience a marked decline before death. Closer inspection of cause of death and
illnesses in late life is required to establish a more solid
basis for this cognitive phenomenon.
There are several limitations in our study that need to be
addressed. In terms of statistical power, there appears to be
no problem in the cross-sectional analyses in which we
found differences at baseline given the moderate effect sizes
and adequate sample size. Also, in the longitudinal analyses,
we have sufficient power to detect differences across time.
However, the power to detect interaction effects is typically
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
memory. However, when the effects of stroke and markers
of CVD were jointly considered, the association between
cognition and time to death was weakened and only observed in three out of the six abilities. Longitudinal analyses
provided limited support for the hypothesis of an accelerated decline in cognitive functioning prior to death. This
was observed only across a 2-year interval in crystallized
knowledge and in episodic memory.
Previous findings from studies on the relationship between cognitive performance and survival show that level of
cognitive performance at one point in time is related to distance to death. Our findings confirm this and indicate that all
cognitive abilities are affected, with the exception of shortterm memory. Thus, we did not find support for the hypothesis of White and Cunningham (1988), that mortality-related
effects are only present in abilities that are typically unaffected by aging (crystallized knowledge). Moreover, we found
that the cognitive abilities that were the strongest predictors
of mortality (in terms of regression coefficients) were the
fluid abilities (i.e., visuospatial ability, episodic memory, and
speed). The same pattern was found in the study of Maier
and Smith (1999), who concluded that the mortality-related
effects associated with intellectual functioning were pervasive rather than ability specific.
The next question of interest concerned the biobehavioral
mechanisms in the relationship between cognitive functioning and survival. As has already been noted, a reasonable
explanation is systematic decline due to changes associated
with compromised physical health. Cerebrovascular disease
is a prime candidate because it is highly prevalent in old age
and also related to cognitive decline (e.g., Carmelli, Swan,
LaRue, & Eslinger, 1997; Elias et al., 1990) and mortality.
Furthermore, as CVD is the leading cause of death (American Heart Association, 1998), and is known to elevate the
risk for stroke (a marker of cerebrovascular disease), it is reasonable to consider the impact of both of these conditions. In
the present study, stroke and two markers of CVD (myocardial infarction and congestive heart failure) were the conditions that most clearly differed between the survivors and
the deceased. Thus, it was of special interest to examine the
effect of these diseases. After adjusting for history of stroke
and markers of CVD, we found that crystallized knowledge
and inductive reasoning were no longer associated with
mortality. These findings suggest that CVD and cerebrovascular disease significantly affect the cognition–survival relationship. To our knowledge, there are no studies that have
systematically examined the effect of CVD or cerebrovascular disease on the cognition–survival association. Thus, at
this stage it can only be speculated how these factors are related. The neurobiological mechanisms between CVD and
cognitive functioning are most likely related to increased
risk for stroke, as well as diminished cardiac output and decreased cerebral perfusion, with the cerebral consequences
being compromised oxygen support.
The finding that stroke and markers of CVD are significant factors in the cognition–survival relationship for selected cognitive tasks remains to be further analyzed. Interestingly, a cognitive domain that largely remains unaffected
in normal aging (e.g., crystallized knowledge) was also affected by vascular diseases. Performance decline in tasks re-
COGNITIVE PERFORMANCE AND SURVIVAL
Acknowledgments
This study is supported by National Institute on Aging Grant NIA: AG
08861 of the National Institutes of Health.
The OCTO Twin Study is an ongoing longitudinal study conducted at the
Institute of Gerontology, University College of Health Sciences in
Jönköping, Sweden, in collaboration with the Center for Developmental
and Health Genetics at the Pennsylvania State University, Philadelphia,
Pennsylvania, and the Department of Medical Epidemiology at the Karolinska Institute in Stockholm, Sweden.
We thank registered nurses Lene Ahlbäck (1991–present), Agneta Carlholt (1993–present), Gunilla Hjalmarsson (1991–1993), Eva Georgsson
(1993–1995), and Anna-Lena Wetterholm (1993–1994), who traveled
throughout the country and examined the participants.
Address correspondence to Linda B. Hassing, Department of Psychology, Göteborg University, Box 500, SE-405 30 Göteborg, Sweden. E-mail:
[email protected]
References
American Heart Association. (1998). 1999 heart and stroke statistical update. Dallas, TX: Author.
American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: Author.
Berg, S. (1996). Aging, behavior, and terminal decline. In J. E. Birren &
K. W. Schaie (Eds.), Handbook of the psychology of aging (4th ed., pp.
323–337). New York: Academic Press.
Bosworth, H. B., & Schaie, K. W. (1999). Survival effects in cognitive
function, cognitive style, and sociodemographic variables in the
Seattle Longitudinal Study. Experimental Aging Research, 25, 121–
139.
Bosworth, H. B., Schaie, K. W., & Willis, S. L. (1999). Cognitive and sociodemographic risk factors for mortality in the Seattle Longitudinal
Study. Journal of Gerontology: Psychological Sciences, 54B, P273–
P282.
Bosworth, H. B., Schaie, K. W., Willis, S. L., & Siegler, I. C. (1999). Age
and distance to death in the Seattle Longitudinal Study. Research on
Aging, 21, 723–738.
Botwinick, J., West, R., Storandt, M. (1978). Predicting death from behavioral test performance. Journal of Gerontology, 33, 755–762.
Carmelli, D., Swan, G. E., LaRue, A., & Eslinger, P. J. (1997). Correlates
of change in cognitive functions in survivors from the Western Collaborative Group Study. Neuroepidemiology, 16, 285–295.
Cederlöf, R., & Lorich, U. (1978). The Swedish Twin Registry. In W. E.
Nance, G. Allen, & P. Parisi (Eds.), Twin research: Biology and epidemiology (pp. 189–195). New York: Alan R. Riss.
Cox, D. R. (1972). Regression models and life regression tables. Journal of
the Royal Statistical Society (B), 34, 187–220.
Deeg, D. J. H., Hofman, A., & van Zonneveld, R. J. (1990). The association
between change in cognitive function and longevity in Dutch elderly.
American Journal of Epidemiology, 132, 973–982.
Dureman, I., & Sälde, H. (1959). Psykometriska och experimental-psykologiska metoder för klinisk tillämpning [Psychometric and experimentalpsychological methods for clinical use]. Uppsala, Sweden: Almqvist &
Wiksell.
Elias, M. F., Elias, J. W., & Elias, P. K. (1990). Biological and health influences on behavior. In J. E. Birren & K. W. Schaie (Eds.), Handbook of
the psychology of aging (3rd ed., pp. 79–102). San Diego, CA: Academic Press.
Hassing, L. B., Small, B. J., von Strauss, E., Fratiglioni, L., & Bäckman, L.
(2002). Mortality-related differences and changes in episodic memory
among the oldest old: Evidence from a population-based sample of
nonagenarians. Aging, Neuropsychology, and Cognition, 9(1), 11–20.
Howieson, D. B., Dame, A., Camicioli, R., Sexton, G., Payami, H., &
Kaye, J. A. (1997). Cognitive markers preceding Alzheimer’s disease in
the healthy oldest old. Journal of the American Geriatrics Society, 45,
584–589.
Jacobs, D. M., Sano, M., Dooneief, G., Marder, K., Bell, K. L., & Stern, Y.
(1995). Neuropsychological detection and characterization of preclinical Alzheimer’s disease. Neurology, 45, 957–962.
Johansson, B. (1988/1989). The MIR—Memory in Reality Test. Stockholm,
Sweden: Psykologiförlaget AB.
Johansson, B., & Berg, S. (1989). The robustness of the terminal decline
phenomenon: Longitudinal data from the Digit-Span Memory Test.
Journal of Gerontology: Psychological Sciences, 44B, P184–P186.
Johansson, B., & Zarit, S. H. (1997). Early cognitive markers of the incidence of dementia and mortality: A longitudinal population-based
study of the oldest old. International Journal of Geriatric Psychiatry,
12, 53–59.
Jonsson, C.-O., & Molander, L. (1964). Manual till CVB-skalan [Manual
of the CVB-Scales]. Stockholm, Sweden: Psykologi Förlaget.
Linn, R. T., Wolf, P. A., Bachman, D. L., Knoefel, J. E., Cobb, J. L., Belanger, A. J., et al. (1995). The “preclinical phase” of probable Alzheimer’s disease. Archives of Neurology, 52, 485–490.
Ljungquist, B., Berg, S., & Steen, B. (1996). Determinants of survival: An
analysis of the effects of age at observation and length of the predictive
period. Aging and Clinical Experimental Research, 8, 22–31.
Maier, H., & Smith, J. (1999). Psychological predictors of mortality in old
age. Journal of Gerontology: Psychological Sciences, 54B, P44–P54.
Masur, D. M., Sliwinski, M., Lipton, R. B., Blau, A. D., & Crystal, H. D.
(1994). Neuropsychological prediction of dementia and the absence of
dementia in healthy elderly persons. Neurology, 44, 1427–1432.
McClearn, G. E., Johansson, B., Berg, S., Pedersen, N. L., Ahern, F., Petrill, S. A., et al. (1997, June 6). Substantial genetic influence on cognitive abilities in twins 80 or more years old. Science, 276, 1560–1563.
Palmore, E., & Cleveland, W. (1976). Aging, terminal decline, and terminal
drop. Journal of Gerontology, 31, 76–81.
Reimanis, G., & Green, R. F. (1971). Imminence of death and intellectual
decrement in the aging. Developmental Psychology, 5, 270–272.
Siegler, I. C. (1975). The terminal drop hypothesis: Fact or artifact? Experimental Aging Research, 1, 169–185.
Siegler, I. C., McCarty, S. M., & Logue, P. E. (1982). Wechsler memory
scale scores, selective attrition, and distance from death. Journal of
Gerontology, 37, 176–181.
Simmons, S. F., Johansson, B., Zarit, S. H., Ljungquist, B., Plomin, R., &
McClearn, G. (1997). Selection bias in samples of older twins? A comparison between octogenarian twins and singletons in Sweden. Journal
of Aging and Health, 9, 553–567.
Small, B. J., & Bäckman, L. (1997). Cognitive correlates of mortality: Evidence from a population-based sample of very old adults. Psychology
and Aging, 12, 309–313.
Small, B. J., & Bäckman, L. (1999). Time to death and cognitive performance. Current Directions in Psychological Science, 8, 168–172.
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016
much lower, such as those between survival group and time.
It is often stated that with greater sample size, effects could
have been detected. However, when effect size is very small,
as was the case in the interaction analyses, sample size is
not the only critical factor. A design with more measurement occasions and more individuals would of course be
beneficial for examining potential interactions between time
and survival groups. However, the fact that we did not find
more marked decline and differences in decline rates between
survivors and decedents (see Figure 2) may also be related
to the fact that our sample excluded individuals with dementia. This exclusion was, however, deliberate so we could examine other common health-related factors in the cognition–
survival relationship.
To summarize, the present study indicates that there are
differences in cognitive performance between those who die
earlier and those who survive longer. Moreover, at least
some of these differences can be ascribed to CVD and cerebrovascular disease. Our findings of few interaction effects
between performance change and subsequent survival provide weak support for an accelerated decline closer to death.
Future studies should address the underlying mechanisms in
the cognition–mortality relationship by using longitudinal
designs with closer measurement occasions and extensive
health-related measurements.
P275
P276
HASSING ET AL.
Small, B. J., Herlitz, A., Fratiglioni, L., Almkvist, O., & Bäckman, L.
(1997). Cognitive predictors of incident Alzheimer’s disease: A prospective longitudinal study. Neuropsychology, 11, 413–420.
Smits, C. H. M., Deeg, D. J. H., Kriegsman, D. M. W., & Schmand, B.
(1999). Cognitive functioning and health as determinants of mortality in
an older population. American Journal of Epidemiology, 150, 978–986.
SPSS. (2000). SPSS advanced models (Version 10.0) [Computer software].
Chicago: Author.
Thurstone, L. L., & Thurstone, T. G. (1949). Manual to SRA primary mental abilities. Chicago: Science Research Associates.
Tulving, E. (1983). Elements of episodic memory. Oxford, England: Oxford University Press.
Wechsler, D. (1991). Manual for the Wechsler Adult Intelligence Scale–
Revised. New York: Psychological Corp.
White, N., & Cunningham, W. R. (1988). Is terminal drop pervasive or specific? Journal of Gerontology: Psychological Sciences, 43, P141–
P144.
Received June 27, 2000
Accepted July 9, 2001
Decision Editor: Margie E. Lachman, PhD
Downloaded from http://psychsocgerontology.oxfordjournals.org/ at Pennsylvania State University on September 12, 2016