Health Psychology Copyright 2003 by the American Psychological Association, Inc.

2003, Vol. 22, No. 6, 559 –569 0278-6133/03/$12.00 DOI: 10.1037/0278-6133.22.6.559

Relation Between Cardiovascular and Metabolic Disease and Cognition

in Very Old Age: Cross-Sectional and Longitudinal Findings From the
Berlin Aging Study

Paul Verhaeghen Markus Borchelt

Syracuse University Charité—Universitätsmedizin Berlin

Jacqui Smith
Max Planck Institute for Human Development

This study documented findings on the relation between cognitive functioning (perceptual speed,
memory, fluency, and knowledge) and cardiovascular and metabolic disease in a sample of very old
adults (ages 70 and older), both cross-sectionally (n ⫽ 516) and longitudinally (n ⫽ 206) in a 4-year
follow-up. After age, SES, sex, and dementia status were controlled for, 4 diagnoses were negatively
associated with cognition: congestive heart failure, stroke, coronary heart disease, and diabetes mellitus,
with a joint effect of 0.47 standard deviations. The impact of disease status was largest on perceptual
speed and fluency, memory was impacted only by diabetes, and knowledge was not related to any
somatic diagnosis. There was no differential decline in participants diagnosed with 1 of these 4 diseases
and those who were not. The only cardiovascular risk factor associated with cognitive performance was
alcohol consumption.

Key words: cognition, diabetes, cardiovascular health, aging

As people grow older, their performance on cognitive and
intellectual tests typically declines (for meta-analyses, see Verhae-
ghen, Marcoen, & Goossens, 1993; Verhaeghen & Salthouse,
1997). Simultaneously, the prevalence of somatic disorders with a
potential impact on cognition (e.g., cardiovascular disease [CVD]
and diabetes) increases after age 70 (e.g., Kannel & Belanger,
1991). These two combined findings raise questions as to whether
phenomena of so-called normal cognitive aging may be contami-
nated by the ill effects of deteriorating health. Although much is
known about the etiology of dementias in old age, information on
the causal pathways and interactions among cardiovascular and
metabolic disease, cardiovascular health risk factors, and non-
pathological cognitive aging is relatively scarce (for an overview,
see Bäckman, Small, Wahlin, & Larsson, 2000).
Previous studies have shown decreased cognitive performance
in individuals suffering from cardiovascular disorders (e.g., for
heart disease and hypertension, M. F. Elias & Elias, 1993; Hertzog,
Schaie, & Gribbin, 1978; Schaie, 1996; Starr, Whalley, Inch, &
Shering, 1993; Waldstein & Elias, 2001; for stroke, Greveson,
Gray, French, & James, 1991). The common cause underlying
these disorders may be atherosclerosis (Vingerhoets, 2001); ath-
erosclerosis may be linked to cognitive functioning through he-
modynamic (e.g., cerebrovascular symptoms due to calcification,
rupture, ulceration, hemorrhage), genetic (e.g., through the ApoE

Paul Verhaeghen, Center for Health and Behavior and Department of
Psychology, Syracuse University; Markus Borchelt, Forschungsgruppe
Geriatrie und Ernährungsmedizin am Evangelischen Geriatriezentrum Ber-
lin, GmbH, Charité—Universitätsmedizin Berlin, Berlin, Germany; Jacqui
Smith, Center for Lifespan Psychology, Max Planck Institute for Human
Development, Berlin.
This research was conducted within the context of the Berlin Aging Study
(BASE), directed by P. B. Baltes, psychology; H. Helmchen, psychiatry; K. U.
Mayr, sociology; and E. Steinhagen-Thiessen, internal medicine and geriatrics.
BASE is a project of the Committee on Aging and Societal Development in
collaboration with the Psychiatric Clinic of the Free University of Berlin, the
Virchow Clinic of the Humboldt University of Berlin, and the Max Planck
Institute for Human Development, where the study is coordinated. The Com-
mittee on Aging and Societal Development, initiated in 1987 by the former
Academy of Sciences and Technology in Berlin, has been sponsored by the
Berlin–Brandenburg Academy of Sciences since 1994. From 1989 to 1991,
BASE was financially supported by the Federal Ministry of Research and
Technology. Since 1992, financial support has been awarded through Federal
Ministry of Family Affairs, Senior Citizens, Women, and Youth Grants
314-1722-102/9 and 314-1722-102/9a. Paul Verhaeghen is supported by Na-
tional Institute on Aging Grant AG-16201.
We acknowledge helpful discussions with John Nesselroade, Michael
Rapp, and Martin Sliwinski.
Correspondence concerning this article should be addressed to Paul Ver-
haeghen, Department of Psychology, Syracuse University, 430 Huntington
Hall, Syracuse, New York 13244-2340; to Markus Borchelt, Evangelisches
Geriatriezentrum Berlin, Reinickendorfer Strasse 61, D-13347 Berlin, Ger-
many; or to Jacqui Smith, Max Planck Institute for Human Development,
Lentzeallee 94, D-14195 Berlin, Germany. E-mail: pverhaeg@psych.syr.edu,
markus.borchelt@charite.de, or smith@mpib-berlin.mpg.de

559
560 VERHAEGHEN, BORCHELT, AND SMITH
⑀4 allele), or metabolic (e.g., impaired insulin metabolism) mech-
anisms (for an overview of these mechanisms, see Everson, Hel-
kala, Kaplan, & Salonen, 2001). A metabolic disease often asso-
ciated with cognitive functioning is diabetes mellitus (e.g., Croxon
& Jagger, 1995; Haan, Shemanski, Jagust, Manolio, & Kuller,
1999; Perlmuter et al., 1984; Schaie, 1996). Diabetes most prob-
ably affects cognitive function through hyper- or hypoglycemia,
which interferes with acetylcholine synthesis (Kumari, Brunner, &
Fuhrer, 2000; Ryan, 2001).
In the present study, we examined the cross-sectional and lon-
gitudinal effects of prevalence and incidence of cardiovascular and
metabolic disorders on cognitive performance. Additionally, we
were interested in two questions of more theoretical importance.
First, it has often been claimed (e.g., Beers & Berkow, 2000) that
in very old age vulnerability to the negative effects of physiolog-
ical pathology increases to the point that cognitive performance in
this group may be determined more by the effects of organic
pathology than by the so-called normal aging process. In this
article, we investigated whether, consistent with the increased-
vulnerability hypothesis, the impact of somatic health increases
with age. In a next step, we examined whether the effects of
chronological age on cognition are indeed mediated by failing
health. A related question is whether the effects of health on
cognition are general or whether they remain restricted to specific
health problems. For instance, for each of the disorders noted
above, clear brain-related mechanisms exist that can explain the
relation with cognition. However, if one considers that in very old
age cognitive functioning becomes increasingly dependent on in-
tactness of the somatic substrate (see the common-cause hypoth-
esis of cognitive aging; Lindenberger & Baltes, 1994, 1997; Salt-
house & Czaja, 2000) and that the cognitive system shows
decreasing plasticity with advancing age (P. B. Baltes & Kliegl,
1992; Singer, Lindenberger, & Baltes, 2000; Verhaeghen & Mar-
coen, 1996), one might expect a rather general negative effect of
somatic health on cognitive functioning, apart from the specific
brain-mediated effects shown in younger cohorts. Given the high
prevalence of cardiovascular problems in old age and their poten-
tial impact on cognition (e.g., P. K. Elias, Elias, D’Agostino,
Silbershatz, & Wolf, 1999; Hill, 1989; Launer, Feskens, Kalmijn,
& Kromhout, 1996), the second aspect of somatic health that we
investigated here is the impact of cardiovascular risk factors on
cognitive functioning.
We should stress that the question about the relation between
health and cognition in very old age is important not only for
advancing understanding of cognitive functioning in this age group
but also because it has practical implications, especially in light of
the possibility of prevention and intervention. Recent evidence
indicates that real-life consequences of individual differences in
intelligence are further augmented in very old age (e.g., Allaire &
Marsiske, 1999; M. M. Baltes & Lang, 1997; Diehl, Willis, &
Schaie, 1995). For instance, using the same data set as analyzed
here, Lindenberger and Baltes (1995) demonstrated that cognition
explained 37% of individual differences in measures of activities
of daily living and up to 45% in expanded levels of everyday
competence (reflecting optional activities necessary for a satisfac-
tory life, e.g., leisure, social, and instrumental activities of daily
living; see also M. M. Baltes, Maas, Wilms, Borchelt, & Little,
1999). In advanced age, the level of cognitive functioning may be
one aspect that contributes considerably to the quality of one’s life
and to the types (and extent) of interaction with the physical and
social environments that one can still enjoy.
In sum, in the present study, we examined the relationships
between CVD, diabetes, CVD risk factors, and intellectual func-
tioning in very old age. Data were derived from the Berlin Aging
Study (BASE; P. B. Baltes & Mayer, 1999), a locally representa-
tive sample of men and women ages 70 to 103 (M ⫽ 85 years).
One of the strengths of this study is that it included a computerized
assessment of five intellectual abilities (Lindenberger & Baltes,
1997) and objective clinical assessments of somatic and mental
health status. Participants in the initial BASE cross-sectional sam-
ple (n ⫽ 516; stratified by age and sex) completed 14 sessions of
individual face-to-face assessment over a 3–5-month period. Sur-
vivors have been followed longitudinally. Here, we report cross-
sectional results at baseline (n ⫽ 516) and results from a 4-year
longitudinal follow-up (n ⫽ 206).
Method
More details about the variables assessed, the sample tested, and the
procedures used can be found in P. B. Baltes and Mayer (1999); P. B.
Baltes and Smith (1997); Lindenberger and Baltes (1997); and Linden-
berger, Mayr, and Kliegl (1993). Here, we offer a brief overview.
Sample
The original BASE consisted of an initial 14-session assessment (Time
1 [T1]) of 516 individuals (age range at first assessment ⫽ 70 –103, mean
age ⫽ 84.92 years, SD ⫽ 8.66), stratified by age and sex, resulting in 43
men and 43 women in each of six different age brackets (70 –74, 75–79,
80 – 84, 85– 89, 90 –94, and 95⫹ years). A total of 1,908 individuals, whose
addresses were obtained by a random draw from the city registry, had to be
contacted to obtain this sample. The initial assessment (T1) was conducted
between mid-1990 and June 1993. The longitudinal extension of BASE
involved four occasions (T1, Time 2 [T2], Time 3 [T3], and Time 4 [T4]),
each scheduled about 2 years apart. At the second wave of measurement,
only parts of the cognitive test battery were collected. Therefore, this
article reports on participants tested on two occasions, that is, the first
(1990 –1993; n ⫽ 516) and the third (1995–1996; n ⫽ 206) assessments.
On average, T3 data were collected 3.99 years (SD ⫽ 0.69) after T1 data.
More details on the longitudinal samples can be found in Singer, Verhae-
ghen, Ghisletta, Lindenberger, and Baltes (2003); details on selectivity
(i.e., drop-out effects) can be found in Lindenberger, Singer, and Baltes
(2002). In brief, as usual in longitudinal aging studies, there was evidence
for relatively strong selectivity effects. The longitudinal sample scored
0.74 standard deviations above the mean of the full cross-sectional sample
on the intelligence composite and 0.27 standard deviations on the SES
composite; 59% of the size of the attrition effect on intelligence and 30%
of the effect of attrition on SES was due to mortality. At T1, the total
sample was, on average, 84.92 years old (SD ⫽ 8.66); the participants had
received an average of 10.75 years of education (SD ⫽ 2.34); and 50%
were female, and 50% were male. At T3, mean age of the participating
sample was 83.62 years (SD ⫽ 7.00); average number of years of education
was 11.23 (SD ⫽ 2.45); and 51% were female, and 49% were male.
Measures
Cognitive assessment. The cognitive test battery consisted of four
intellectual abilities, each assessed by unit-weighted composites of two
tests: (a) perceptual speed (measured by Digit Letter and Identical Pictures;
Cronbach’s ␣ ⫽ .96 and .90, respectively), (b) episodic memory (measured
by Paired Associates and Memory for Text; Cronbach’s ␣ ⫽ .87 and .57,
respectively), (c) fluency (measured by Categories and Word Beginnings;
because of the test format, no alphas could be calculated), and (d) knowl-
 SPECIAL SECTION: HEALTH–COGNITION IN VERY OLD AGE
edge (measured by Vocabulary and Spot-a-Word; Cronbach’s ␣ ⫽ .82 and
.92, respectively). Note that in the original cross-sectional analysis at T1,
reported in Lindenberger and Baltes (1997), some of these abilities were
measured by three tests. Because of time constraints on the follow-up
assessments, only two tests per construct were administered at T3. A
unit-weighted intelligence composite was computed from the four cogni-
tive composites; correlations among the four abilities ranged from .63 to
.73. Autocorrelations over the 4 years between T1 and T3 were high: .69
for perceptual speed; .65 for memory; .76 for fluency, .79 for knowledge,
and .83 for the intelligence composite. A detailed description of the tests
can be found elsewhere (Lindenberger & Baltes, 1994, 1997; Lindenberger
et al., 1993). The five composites were scaled by linear transformation such
that scores at the first measurement occasion for the full sample conformed
to a T metric (i.e., M ⫽ 50, SD ⫽ 10).
SES. Four variables were used to represent SES: (a) income (on a
5-point scale); (b) occupational prestige (based on a standard rating scale
in Germany); (c) social class (on a 5-point scale); and (d) number of years
of education. Out of these four variables, we constructed a unit-weighted
SES composite (after z transformation), to be used as a covariate in our
analyses.
Somatic diagnoses. Of the possible diagnostic categories, 26 were used
in BASE. We concentrated on CVD (myocardial infarction, congestive
heart failure, coronary heart disease, stroke, and hypertension) and diabetes
mellitus. To sort individual participants into diagnostic categories, psychi-
atrists and internists held a consensus conference (one at T1, one at T3) in
which all available information (medical examination, interview, resting
electrocardiography [ECG], blood pressure, Doppler ultrasound, quantita-
tive computer tomography, flow cytometry, lymphocyte stimulation, major
histocompatibility compatibility types, serum, plasma, and urinalyses) was
brought together. For each individual, each illness’s diagnostic certainty
and objective and subjective severity were rated (for more details, see
Steinhagen-Thiessen & Borchelt, 1999). All discernible diseases and di-
agnosable pathological states were classified according to the four-digit
code of the 9th edition of the International Classification of Diseases
(World Health Organization, 1978), on a rather low aggregation level. All
diagnoses were coded independent of each other. Note, however, that some
diagnoses tend to overlap. For instance, myocardial infarction tends to be
superimposed on a diagnosis of coronary heart disease, but the association
is not absolute (in the present sample, 89 of 110 participants who were
diagnosed with myocardial infarction were also diagnosed with coronary
heart disease; 89 of 176 participants diagnosed with coronary heart disease
also had myocardial infarction). For the myocardial infarction and coronary
heart disease diagnoses, case history, medication history, interview with
family doctor, physical examination, and resting ECG findings were taken
into account. At least two of the following needed to be present for
coronary heart disease: typical angina, stenocardia, nitrate therapy, family
doctor’s diagnosis, and typical ECG abnormalities; myocardial infarction
was diagnosed on the basis of case history, including the interview with the
general physician, plus the ECG findings typical of myocardial infarction.
Interrater reliability for the diagnostic categories was .90. For the present
analyses, ratings were recoded into three categories: diagnosis absent,
diagnosis present, and diagnosis missing; the latter category typically
implied diagnostic uncertainty.
Cardiovascular health risk factors. Smoking behavior (current and
former smoking status, amount of units smoked per day currently or
formerly, starting age of current or former smoking, and quitting age of
former smoking) and alcohol consumption (current alcohol status, scored
as yes or no; number of units consumed per day) were assessed through a
questionnaire. Levels of high-density lipoprotein (HDL cholesterol, which
is correlated negatively with the risk for CVD) and low-density lipoprotein
(LDL cholesterol, which is correlated positively with the risk for CVD)
were determined from blood samples (no fasting period was observed prior
to blood sampling). The body mass index was calculated from height and
weight.
561
Statistical Analyses
To determine the influence of somatic diagnosis on level of performance
at the first measurement point and on 4-year change scores, we conducted
a series of hierarchical regression analyses. Age, sex, SES, and dementia
diagnosis were entered first as control variables, and each of the diagnoses
was entered in a second step. The effects of multimorbidity were tested in
a repeated measures analysis of variance (ANOVA). The effects of risk
factors were assessed through hierarchical regression, after we controlled
for age, sex, SES, dementia diagnosis, and the presence or absence of each
of the somatic diagnoses found in the previous analyses to be correlated
with cognition. Alpha level for statistical testing was set at .05.
Results
Somatic Diagnoses and Cognition: Cross-Sectional
Analyses
In Table 1, we report the significant results of the regression
analysis relating cardiovascular diagnoses and diabetes (scored as
present or absent) to cognition (scored as a continuous variable),
after we partialed out age, sex, the SES composite, and dementia
diagnosis. Because of missing values due to diagnostic uncer-
tainty, valid sample size (N) was lower than the number of par-
ticipants and varied from analysis to analysis. Two CVD diagnoses
were not related to cognition: hypertension and myocardial infarc-
tion. Four diagnoses proved to be significantly and negatively
correlated with intelligence. These were, in order of decreasing
importance, stroke, congestive heart failure, diabetes, and coronary
heart disease. The amount of variance accounted for after we
controlled for age, sex, SES, and dementia status was slight (be-
tween 0.5% and 2.0% of the variance), but the effect sizes were
moderate: B values (which indicate the average difference in
cognitive test score between participants with and without the
diagnosis) ranged from ⫺1.8 to ⫺3.2, or (given that SD ⫽ 10)
from ⫺0.18 to ⫺0.32 standard deviations. Taken together, the four
cognition-related somatic diagnoses explained 1.6% of the vari-
ance in intelligence (after we controlled for age, sex, SES, and
dementia). The combined effect of the four cognition-related di-
agnoses (i.e., the sum of the B values) was ⫺4.7, or ⫺0.47
standard deviations. Fluency and perceptual speed appeared to be
the variables most impacted by somatic health, yielding the larger
B values; memory was associated only with diabetes; knowledge
was not reliably associated with any of the four diagnoses.
We hypothesized that stroke would carry much of the variance
associated with other cardiovascular or metabolic disease diag-
noses and that the effects of these diagnoses would emerge only
when stroke was controlled for. To test this assertion, we first redid
the hierarchical regression analysis by using only the sample of
362 participants who did not receive a diagnosis of dementia,
stroke, or both at T1 (see Table 2). All of the significant effects
from the previous hierarchical regression analysis were confirmed,
with the exception of the relation between intelligence and coro-
nary heart disease and of the combined effects of the diagnoses on
perceptual speed. Additionally, we examined the partial correla-
tion matrix between all cardiovascular diagnoses and cognition in
this selected sample, controlling for age, SES, and gender. No new
significant correlations were found involving the intelligence com-
posite; knowledge correlated significantly with peripheral nervous
system disorders (r ⫽ ⫺.09, n ⫽ 354) and myocardial infarction
(r ⫽ ⫺.10, n ⫽ 331), and fluency correlated significantly with
562 VERHAEGHEN, BORCHELT, AND SMITH

Table 1
Results of Hierarchical Multiple Regression Analysis Predicting Cognitive Performance by Selected Somatic Diagnoses

Intelligence Knowledge Fluency Memory Perceptual speed

Step and variable ⌬R 2

B ⌬R 2
B ⌬R 2
B ⌬R 2
B ⌬R2 B

Diabetes (n ⫽ 398; prevalence ⫽ 19%)

Step 1 .444* .224* .343* .328* .451*
Age ⫺0.45* ⫺0.25* ⫺0.38* ⫺0.36* ⫺0.53*
Sex ⫺0.15 ⫺1.50 ⫺0.92 2.53* ⫺0.63
SES ⫺0.05 ⫺0.03 ⫺0.03 ⫺0.11* ⫺0.03
Dementia diagnosis ⫺10.35* ⫺8.28* ⫺9.25* ⫺9.20* ⫺8.57*
Step 2 .005* .001 .005* .010* .002
Diabetes ⫺1.81* ⫺1.57 ⫺1.82* ⫺2.52* ⫺1.15

Coronary heart disease (n ⫽ 455; prevalence ⫽ 37%)

Step 1 .447* .228* .342* .337* .453*
Age ⫺0.44* ⫺0.24* ⫺0.37* 0.04* ⫺0.53*
Sex ⫺0.08 ⫺1.54 ⫺0.75 2.73* ⫺0.70
SES ⫺0.05 ⫺0.01 ⫺0.02 ⫺0.12* ⫺0.03
Dementia diagnosis ⫺10.61* ⫺8.61* ⫺9.46* ⫺9.33* ⫺8.76*
Step 2 .007* .006 .011* .003 .002
Coronary heart disease ⫺1.74* ⫺1.61 ⫺2.14* ⫺1.19 ⫺1.00

Congestive heart failure (n ⫽ 493; prevalence ⫽ 65%)

Step 1 .450* .224* .347* .334* .457*
Age ⫺0.45* ⫺0.25* ⫺0.37* ⫺0.37* ⫺0.53*
Sex ⫺0.06 ⫺1.45 ⫺0.86 2.73* ⫺0.64
SES ⫺0.05 ⫺0.02 ⫺0.02 ⫺0.10* ⫺0.03
Dementia diagnosis ⫺10.37* ⫺8.14* ⫺9.46* ⫺9.19* ⫺8.56*
Step 2 .011* .004 .023* .001 .011*
Congestive heart failure ⫺2.37* ⫺1.37 ⫺3.49* ⫺0.81 ⫺2.41*

Stroke (n ⫽ 504; prevalence ⫽ 9%)

Step 1 .443* .220* .343* .332* .448*
Age ⫺0.44* ⫺0.23* ⫺0.38* ⫺0.37* ⫺0.52*
Sex 0.03 ⫺1.33 ⫺0.66 2.65* ⫺0.49
SES ⫺0.06 ⫺0.03 ⫺0.04 ⫺0.11* ⫺0.03
Dementia diagnosis ⫺10.37* ⫺8.45* ⫺9.25* ⫺9.03* ⫺8.61*
Step 2 .020* .004 .012* .001 .014*
Stroke ⫺3.24* ⫺2.20 ⫺3.72* ⫺0.95 ⫺4.16*

Diabetes, coronary heart disease, congestive heart failure, and stroke (n ⫽ 423; prevalence ⫽ 62%)
Step 1 .451* .223* .341* .342* .454*
Age ⫺0.42* ⫺0.23* ⫺0.39* ⫺0.38* ⫺0.54*
Sex 0.34 ⫺1.26 ⫺0.35 3.12* ⫺0.37
SES ⫺0.05 ⫺0.01 ⫺0.02 ⫺0.12* ⫺0.03
Dementia diagnosis ⫺10.37* ⫺8.64* ⫺9.05* ⫺9.06* ⫺8.62*
Step 2 .016* .006 .003* .011 .015*
Diabetes ⫺1.36 ⫺0.42 ⫺1.15 ⫺2.27* ⫺0.78
Coronary heart disease ⫺0.78 ⫺1.08 ⫺0.79 ⫺0.90 0.12
Congestive heart failure ⫺1.78* ⫺0.72 ⫺2.87* ⫺0.32 ⫺2.15*
Stroke ⫺0.79 ⫺0.03 ⫺1.72 1.05 ⫺2.00

Note. When values for diagnosis are missing, that is, when diagnosis is uncertain, n ⬍ 516.
* p ⬍ .05.

peripheral vascular disease (r ⫽ ⫺.10, n ⫽ 354). Given that the
change in result was minor and did not affect more than one
ability, we concluded that the presence of stroke patients did not
have a masking effect on the data. Therefore, to benefit from the
statistical power associated with the full sample size, all subse-
quent analyses were conducted using the original sample.
The sample was split into a group of older adults without any of
the four cognition-related diagnoses (n ⫽ 105) and a group of
individuals with one or more of these four diagnoses (n ⫽ 318).
Table 3 reports differences between these two groups for the
cognitive variables, age, sex, dementia status, and SES. For all
variables, significant effects were found; on all variables with
an evaluative direction, the healthy sample fared better. The
effect on cognition seemed to be moderately large: For the
intelligence composite, the two health groups were 6.3 T score
units, that is, 0.63 standard deviations apart. After age, sex,
 SPECIAL SECTION: HEALTH–COGNITION IN VERY OLD AGE 563

Table 2
Results of Hierarchical Multiple Regression Analysis Predicting Cognitive Performance by Selected Somatic Diagnoses of
Participants Who Did Not Receive a Diagnosis of Dementia, Stroke, or Both at Time 1

Intelligence Knowledge Fluency Memory Perceptual speed

Step and variable ⌬R 2

B ⌬R 2
B ⌬R 2
B ⌬R 2
B ⌬R2 B

Diabetes (n ⫽ 358)
Step 1 .235* .063* .161* .178* .307*
Age ⫺0.49* ⫺0.26* ⫺0.44* ⫺0.40* ⫺0.59*
Sex 1.27 ⫺0.37 0.39 4.12* 0.20
SES ⫺0.10* ⫺0.08 ⫺0.06 ⫺0.14* ⫺0.07
Step 2 .014* .004 .014* .017* .005
Diabetes ⫺2.81* ⫺3.23* ⫺1.58

Coronary heart disease (n ⫽ 320)

Step 1 .231* .059* .156* .182* .304*
Age ⫺0.50* ⫺0.26* ⫺0.44* ⫺0.39* ⫺0.60*
Sex 1.55 ⫺0.22 0.81 4.44* 0.24
SES ⫺0.11* ⫺0.09 ⫺0.06 ⫺0.15* ⫺0.07
Step 2 .006 .002 .012* .004 .001
Coronary heart disease ⫺1.39 ⫺0.86 ⫺2.12* ⫺1.28 ⫺0.47

Congestive heart failure (n ⫽ 346)

Step 1 .237* .064* .158* .184* .307*
Age ⫺0.50* ⫺0.27* ⫺0.44* ⫺0.40* ⫺0.59*
Sex 1.38 ⫺0.34 0.46 4.42* 0.17
SES ⫺0.10* ⫺0.08 ⫺0.05 ⫺0.14* ⫺0.06
Step 2 .014* .003 .033* .002 .012
Congestive heart failure ⫺2.30* ⫺1.12 ⫺3.69* ⫺0.88 ⫺2.16*

Diabetes, coronary heart disease, and congestive heart failure (n ⫽ 306)

Step 1 .239* .064* .162* .185* .311*
Age ⫺0.51* ⫺0.27* ⫺0.45* ⫺0.40* ⫺0.62*
Sex 1.59 ⫺0.20 0.86 4.58* 0.19
SES ⫺0.10* ⫺0.08 ⫺0.05 ⫺0.15* ⫺0.06
Step 2 .023* .006 .039* .017 .014
Diabetes ⫺2.14 ⫺1.46 ⫺1.45 ⫺2.80* ⫺1.57
Coronary heart disease ⫺0.67 ⫺0.66 ⫺1.06 ⫺0.94 0.38
Congestive heart failure ⫺1.59 ⫺0.30 ⫺3.04* ⫺0.22 ⫺1.85

Note. When values for diagnosis are missing, that is, when diagnosis is uncertain, n ⬍ 362.
* p ⬍ .05.

SES, and dementia diagnosis were controlled for, the two
groups were 0.34 standard deviations apart, F(1, 417) ⫽ 15.43,
MSE ⫽ 53.73, p ⬍ .05.
Figure 1 depicts means and standard errors for the intelligence
composite as a function of multimorbidity of the four cognition-
related diagnoses (diagnosis absent: n ⫽ 105; one diagnosis
present, n ⫽ 136; two diagnoses present, n ⫽ 122; three diagnoses
present, n ⫽ 55; and all four diagnoses present, n ⫽ 5). A
univariate ANOVA with intelligence as the dependent variable;
multimorbidity as the predictor; and age, sex, SES, and dementia
status as covariates showed a significant effect of multimorbidity,
F(4, 414) ⫽ 4.90, MSE ⫽ 53.58, p ⬍ .05. Repeated contrast
analysis, however, showed that there was a significant difference
in intelligence score only between the group with no diagnosis and
the group with one diagnosis; all further repeated contrasts proved
nonsignificant, indicating that the groups of participants with one
or more of the diagnoses were not reliably different in cognitive
performance, as can clearly be seen in Figure 1.
Figure 2 depicts a scatter plot of intelligence scores (at T1) as a
function of age, separated by individuals who were diagnosed with
one or more of the four cognition-related somatic disorders and
those who were not, along with the freely estimated regression
lines for each group. Regression analysis including a dummy
intercept term for health group and a Health Group ⫻ Age inter-
action term (Berry & Feldman, 1985) showed that the two regres-
sion lines were reliably different from each other, R2 ⫽ .33; R2 for
inclusion of the dummy intercept and the interaction term ⫽ .029,
F(2, 419) ⫽ 8.99, p ⬍ .05. The line for the participants with at
least one cognition-related somatic diagnosis had a reliably smaller
intercept (57.44 vs. 63.73, evaluated at age 70) and a reliably
shallower slope (⫺0.54 vs. ⫺0.76) than the line for individuals
without one of these disorders.
Somatic Diagnoses and Cognition: Four-Year
Longitudinal Analyses
For the 4-year longitudinal analyses (T1 to T3), the sample was
split into three groups, one for each of the four cognition-related
diagnoses: (a) the group of participants who did not have that
particular cognition-related diagnosis at T1 and T3 (diagnosis
564 VERHAEGHEN, BORCHELT, AND SMITH

Table 3
Descriptive Statistics for Participants, Along With Results of the
Independent Samples t Test

Participants Participants
withouta withb

Variable M SD M SD t(421)

Intelligence 55.35 9.90 49.03 9.56 5.83*

Speed 55.15 9.66 48.90 9.82 5.67*
Memory 53.47 10.70 49.32 9.66 3.71*
Fluency 55.78 11.31 48.97 8.93 6.32*
Knowledge 53.84 8.79 49.48 10.25 3.91*
Age 81.00 8.44 85.72 8.50 4.95*
% women 41 53 2.06*
% demented 11 23 2.63*
SES composite 51.80 10.25 49.52 9.83 2.04*
a
Participants without one or more cognition-related diagnoses (n ⫽
105). b Participants with one or more cognition-related diagnoses (n ⫽ Figure 2. Composite intelligence score as a function of age in the group
318). of participants with at least one of the four cognition-related diagnoses (i.e.,
* p ⬍ .05. diabetes, congestive heart failure, coronary heart disease, or stroke) and the
group of participants without any cognition-related diagnosis. CVS ⫽
cardiovascular syndrome; Diab ⫽ diabetes.
absent), (b) the group of participants who received that particular
diagnosis prior to T1 and held it at T3 (diagnosis present), and (c)
the group of participants who received the diagnosis at T3 but not rates in our sample were high (e.g., the American Heart Associa-
T1 (incidence). Analogous groups were constructed for the pres- tion, 2002, estimates prevalence of congestive heart failure in the
ence, absence, and incidence of any of the four cognition-related 75-year-and-older age bracket at 9.8%). The discrepancy may have
diagnoses (called hereafter the summary data). The breakdown in been due (a) to oversampling of the very oldest cohorts in our
number of participants for each of the three groups was as follows: study (i.e., compared with the population, we had a disproportion-
for coronary heart disease, 96, 70, and 24 for diagnosis absent, ate number of the oldest old—age 85 and older), (b) to the careful
diagnosis present, and incidence, respectively; for congestive heart objective individual assessment of disorders in our study, and (c)
failure, 46, 101, and 49, respectively; for diabetes, 147, 40, and 18, to the inclusion of mild cases in the diagnosis-present category.
respectively; for stroke, 183, 14, and 7, respectively; and for the Figure 3 displays the means and standard errors for the intelli-
summary data, 18, 108, and 41, respectively. Note that prevalence gence composite of the three groups for each of the four diagnoses
and the summary data as a function of time in study, controlled for
age, sex, SES, and dementia status. The graph shows that for all
diagnoses, with the possible exception of stroke, the slopes for the
three health groups were remarkably similar. To test for group
differences in the longitudinal slopes, we conducted a repeated
measures ANOVA for each of the five comparisons, with time of
measurement and type of cognition (i.e., speed, memory, fluency,
and knowledge) as within-subject factors; health group (i.e., diag-
nosis absent, diagnosis present, and incidence) as a between-
subjects factor, and age, sex, SES, and dementia status as covari-
ates. None of the interactions involving time by health group
reached significance in any of the four cognition-related diagnoses,
nor did they in the summary data. There was, however, a marginal
trend toward a three-way Time ⫻ Type of Cognition ⫻ Health
Group interaction for the stroke diagnosis, F(6, 591) ⫽ 2.03,
MSE ⫽ 19.43, p ⫽ .06. Decline was more precipitous in the stroke
incidence group (⫺5.53 T score units) than in the other two stroke
groups (diagnosis absent ⫽ ⫺2.19; diagnosis present ⫽ ⫺1.93);
the decline in the stroke incidence group was more apparent for
speed and fluency (⫺8.20 and ⫺6.31, respectively) than for mem-
ory and knowledge (⫺0.01 and ⫺4.35, respectively). Note that the
incidence group for stroke was very small (n ⫽ 7), and, therefore,
power for the detection of the interaction was very low. This
finding may warrant treating the trend as at least clinically
Figure 1. Composite intelligence score as a function of the number of significant.
cognition-related diagnoses. Error bars represent plus or minus one stan- With the exception of stroke, F(2, 197) ⫽ 2.63, MSE ⫽ 253.41,
dard error. CVS ⫽ cardiovascular syndrome; Diab ⫽ diabetes. p ⫽ .75, all health group main effects were significant. Figure 3
 SPECIAL SECTION: HEALTH–COGNITION IN VERY OLD AGE 565

Figure 3. Four-year longitudinal changes (corrected for age, sex, and SES) for diagnosis-absent, diagnosis-
present, and incident cases for the four cognition-related diagnoses as well as for any combination of the four
cognition-related diagnoses. Error bars represent plus or minus one standard error. T1 ⫽ Time 1; T3 ⫽ Time 3.

shows that, in general, the diagnosis-absent group performed at a
higher level than the diagnosis-present group. At the first time of
measurement, the incidence group was not reliably distinguishable
from the diagnosis-absent group for most diagnoses, with the
notable exception of diabetes.
Cardiovascular Risk Factors and Cognition
The only cardiovascular risk factor that was reliably associated
with cognitive performance at T1 after age, sex, SES, dementia
status, and the presence or absence of the four cognition-related
diagnoses were controlled for was alcohol consumption, that is, an
affirmative answer by the participant to the question regarding
whether he or she presently ever drank alcohol (380 participants
indicated consuming alcohol). For the intelligence composite, the
increment in R2 was .02; the effect size (i.e., B for the regression
line) was ⫺3.66. The effect of alcohol consumption was signifi-
cant for all four cognitive variables (knowledge: B ⫽ ⫺3.44;
fluency: B ⫽ ⫺2.63; memory: B ⫽ ⫺3.80; and speed: B ⫽
⫺2.61). None of the cardiovascular risk factors were reliably
associated with 4-year declines in cognition. When only the sam-
ple of participants without a dementia and/or stroke diagnosis
was considered, the pattern of results did not change, with the
exception of a nonsignificant effect for alcohol consumption on
speed. For the intelligence composite, the increment in R2 was
.03; the effect size was ⫺3.52. For knowledge, B ⫽ ⫺2.88; for
fluency: B ⫽ ⫺2.92; for memory: B ⫽ ⫺4.24; and for speed: B ⫽
⫺1.94, ns.
566 VERHAEGHEN, BORCHELT, AND SMITH
Discussion
In the present study, we investigated selected aspects of somatic
health and their relation to cognition in a relatively large sample of
very old (i.e., ages 70 and older) adults, followed over a period of
4 years.
Somatic Diagnoses and Health: Regression Analyses
Three cardiovascular disorders—namely, congestive heart fail-
ure, myocardial infarction, and stroke—were found to be reliably
associated with cognitive functioning, even after the effects of age,
sex, SES, and dementia status were controlled for. Diabetes also
had a negative impact on cognition.
There are essentially three ways to assess the impact of these
disorders on the aging cognitive system. The first is to calculate the
percentage of variance explained by diagnostic status over and
above the effects of age, sex, SES, and dementia status. This
indicates the impact of disease over and above the effects associ-
ated with aging and biographical variables. The increments in R2
ranged between .005 and .020, and, therefore, somatic disease
status explained 2% or less of the variance over and above the
effects of age, sex, SES, and dementia. All disorders influenced
perceptual speed and fluency, but none impacted knowledge; the
only reliable effect on episodic memory performance was noted
for diabetes. Therefore, the effects of disease status are not general
but remain restricted to speeded tests.
The second way of evaluating the impact of somatic disorders
on cognition is to examine the regression coefficients, which
indicate the difference in score between participants with and
without the disorders. This difference was relatively large: It
ranged from ⫺0.18 standard deviations to ⫺0.32 standard devia-
tions. An alternative way of presenting these coefficients is to scale
them compared with the effects of aging as indicated by the
regression weight for age. Then, diabetes had an effect equal to
that of 4.0 years of aging (in other words, performance of an
80-year-old with diabetes was comparable to that of an 84-year-old
without diabetes); coronary heart disease of 3.9 years of aging,
congestive heart failure of 5.3 years of aging, and stroke of 7.4
years of aging. These effects were small compared with the effect
of dementia, which decreased performance by 1 standard deviation
(an effect about equal to 24 years of aging).
The third way of assessing the impact of somatic diagnosis is to
look at its mediating effects on cognition, that is, by calculating
how much of the variance in cognition associated with age (as
opposed to total variance) is due to individual differences in
disease status (see Salthouse, 1991, pp. 322–323, for the calcula-
tion). Taken together, the four cognition-related diagnoses ex-
plained 17% of the age-related variance in the intelligence com-
posite (knowledge: 22%; fluency: 19%; memory: 24%, and speed:
19%).
Summarized, we can state that somatic status has an impact on
cognitive performance. This impact is relatively modest, however,
compared with the effects of dementia.
Somatic Diagnoses: Are Health Effects General?
Our results go against the hypothesis that in very old age the
cognitive system becomes vulnerable to any negative change in
health status. Only a limited number of diagnoses were reliably
associated with decreased cognitive performance. Moreover, for
each of the four cognition-related disorders, physiological CNS-
related mechanisms exist that can explain the relation with cogni-
tion. CVDs interfere with normal blood flow and thus restrict
oxygenation of the CNS. Stroke has a direct impact on neuronal
tissue. Diabetes mellitus may affect cognitive function through
hyper- or hypoglycemia, which interferes with acetylcholine syn-
thesis (Kumari et al., 2000). Thus, it appears that the effects of
somatic health on cognition are mediated by mechanisms that are
directly brain-related. Additional evidence that this may be the
case comes from the finding that these disorders do not impact all
cognitive variables. Vocabulary, presumably an estimate of the
extent of an individual’s knowledge base acquired over a lifetime,
is not sensitive to the effects of health; episodic memory is affected
only by diabetes. Fluency and speed, the two variables affected by
all four diagnoses, are tests that require fast online responses from
the individual, and it makes sense that these aspects of cognition
are negatively affected by oxygenation deficiencies or neurotrans-
mitter depletion.
Note, however, that our data may contain a number of false
negatives. That is, we did not find reliable correlations between
intelligence and somatic diagnosis for some disorders that have
been known to correlate with cognition in other studies involving
older adults. The most noticeable false negatives are hypertension
(e.g., Waldstein, 1995) and myocardrial infarction. One possible
reason might be that our cognitive battery (which was, after all, not
designed to capture the effects of specific somatic syndromes but
to assess a wide range of cognitive functions) was too broad to
precisely measure the effects of these diseases on particular sub-
sets of cognition. For instance, the effects of hypertension seem to
be most noticeable in episodic memory (Denicoff, Joffe, Laksh-
manan, Robbins, & Rubinow, 1990; Wahlin, Robins-Wahlin,
Small, & Bäckman, 1998; Waldstein, 1995). Maybe our episodic
memory measure (which was associated only with diabetes) was
simply not sensitive enough.
Somatic Diagnoses: Decreased Vulnerability in Very
Old Age?
Old age is generally assumed to be associated with increased
cognitive vulnerability. For instance, the disablement process
model by Verbrugge and Jette (1994) claims that increased mor-
bidity due to the cumulation of risk factors in old age leads to
functional limitations, such as a decrease in cognitive functioning,
which in turn may lead to increased disability in a number of areas
of daily living. Some results of our study, however, counter this
hypothesis and, in fact, point at decreasing vulnerability of the
cognitive system vis-à-vis somatic morbidity.
A first result pointing in the direction of the decreasing impact
of somatic disorders on the cognitive system is the lack of effect of
multimorbidity on cognitive performance. Rather than showing a
cumulative effect of added diagnoses, our data imply a knock-out
effect; that is, the presence of one of the four cognition-related
diseases led to a reliable reduction in cognitive performance, but
the further presence of other cognition-related diseases did not
decrease performance beyond the initial reduction (see Figure 1).
A second result that runs counter to the hypothesis of increased
vulnerability is that the negative effect of disease status was larger
in the younger strata of our sample than in the oldest groups (see
Figure 2). Cognitive performance of the group of individuals
 SPECIAL SECTION: HEALTH–COGNITION IN VERY OLD AGE
without any of the four cognition-related diagnoses and the group
of individuals with one or more cognition-related diagnoses, in
fact, converged at around age 95 (such convergence has been
found previously in the relation between cognition and hyperten-
sion; Schultz, Dineen, Elias, Pentz, & Wood, 1979; Waldstein,
1995).
Two longitudinal findings augment this surprising cross-
sectional result. First, the bottom right panel of Figure 3 shows that
there was no significant trend toward divergence over time for the
group of individuals not diagnosed with any of the cognition-
related disorders at any of the two measurement points and the
group of participants diagnosed with one or more of the cognition-
related disorders at both measurement points. Performance in the
diagnosis-absent group declined 2.7 points over the course of 4
years; decline in the diagnosis-present group equaled 2.0 points.
Clearly, participants diagnosed with one or more of the cognition-
related disorders did not decline reliably faster than participants
without these disorders.
Second, Figure 3 also demonstrates that the decline experienced
by the group of participants receiving their first cognition-related
diagnosis between the first and second measurement points was
not larger than the decline in the diagnosis-free group (the inci-
dence group declined 2.6 points.) This finding implies that the
impact of a particular cognition-related diagnosis received prior to
entering the study was quite large (i.e., the diagnosis-present group
performed lower than the diagnosis-absent group), but the impact
of receiving a new diagnosis in between the first and second
measurement points was quite limited and, in fact, so limited it
could not be traced in our data (i.e., the incidence group did not
decline more than the diagnosis-absent group). This finding sug-
gests that the impact of disorders on cognitive performance is
larger earlier in life (i.e., in the period before T1) than later in life
(i.e., in the period between T1 and T3). Stroke, however, is a
possible exception to this pattern, with a marginally significant
interaction indicating that incident stroke cases decline faster.
How can this negative relation between age and the impact of
disease on cognitive performance be explained? At least two
mechanisms seem feasible. First, the convergence may be due to
survival effects and selective attrition (see Waldstein, 1995, 2000,
for a similar argument). That is, with advancing age, vulnerability
for disease increases, resulting in higher mortality and, in survi-
vors, in disabilities that might lead a prospective participant to
decide not to get involved in a study as exacting as BASE. The net
result is that the group of younger-older adults in the study is less
selected and may even be quite representative of the population;
the group of very old participants, however, is likely to be highly
positively selected and represent only the highest levels of func-
tioning present in the population (Lindenberger, Singer, & Baltes,
2002; Singer et al., 2003). This introduces a bias in the data: The
oldest participants in the study may well be functioning at such
high level in comparison to their age peers that the presence of
somatic disorder may not matter much, whereas the presence of
such a disorder may well throw the cognitive system of younger,
less selected older adults off-balance. Also, note that all prevalent
cases are survivors, and very old adults who survive a potentially
life-threatening condition may be of exceptional constitution.
A second explanation has opposite valence. It can be argued that
with advancing age, biological changes increasingly resemble dis-
ease states. There is marked age-related decline, for instance, in
brain volume, notably in the prefrontal cortex and hippocampal
567
areas, which subserve higher order cognition (for a meta-analytic
review, see Raz, 2000). Blood flow in the CNS (the likely locus for
the cognitive effects of cardiovascular disorders) is decreased by
the effects of atherosclerosis, accumulated over a lifetime, as well
as from age-related changes in elastin and collagen and from
age-related calcium deposits (Beers & Berkow, 2000). Negative
age changes occur in the cholinergic system, the probable site of
impact for diabetes (for a review, see Muir, 1997). It is then
possible that in very late life, subclinical changes in the substrate
of the cognitive system are already overlapping to a large degree
with the effects typically shown in CVDs, stroke, and diabetes.
Therefore, the impact of these diseases on cognitive functioning in
very old age may be smaller than their impact in younger adults,
not because of decreased vulnerability but because the disease
process adds little to the cumulative changes in brain physiology
that have occurred over the course of a very long life. This
process-overlap interpretation, then, assumes that the observed
phenomena are real and that the observed decrease in vulnerability
in very old age is caused, paradoxically, by an underlying decreas-
ing intactness of the substrate.
Although it is impossible to distinguish empirically between the
two interpretations on the basis of our results, previously published
analyses using BASE data are compatible with this view of rela-
tively global negative changes in the mind– body continuum.
These include the observation of increased dedifferentiation of
cognitive abilities, that is, very high intercorrelations among abil-
ities that are typically less highly correlated in early adulthood
(Lindenberger & Baltes, 1997; Singer et al., 2003), increased
interrelatedness of the cognitive system and the sensorimotor
system (P. B. Baltes & Lindenberger, 1997; Lindenberger &
Baltes, 1997; Singer et al., 2003), and decreased cognitive plas-
ticity in this very old sample (Singer et al., 2000).
Somatic Correlates of Cognition: Cardiovascular
Risk Factors
After the four cognition-related diagnoses and dementia were
taken into account, only alcohol consumption was found to be
related to cognition. This finding implies that the possible effects
of smoking, cholesterol, and obesity on cognition are mediated
through their effects on CVD. There were no reliable differences
in decline associated with cardiovascular risk factors. These results
echo reports from the literature on the diminishing effects of
cardiovascular risk factors on mortality in the very old (e.g., Corti
et al., 1997; Krumholz, Seeman, Merrill, Mendes de Leon, &
Vaccarino, 1994; Stevens et al., 1998; Woo, Ho, Yuen, Yu, & Lau,
1998). Note, however, that the assessment of risk factors was
limited: The lack of fasting preceding the cholesterol measures and
the noninclusion of other risk factors shown in previous research to
be correlated with cognition (e.g., glucose, insulin, homocysteine)
are clear limitations of the present study.
Conclusions
We found that four cardiovascular and metabolic diagnoses
were negatively associated with cognition after age, SES, sex, and
dementia status were controlled for: congestive heart failure,
stroke, coronary heart disease, and diabetes mellitus, with a joint
effect of 0.47 standard deviations. The impact of disease status was
largest on perceptual speed and fluency, memory was impacted
568 VERHAEGHEN, BORCHELT, AND SMITH
only by diabetes, and knowledge was not related to any somatic
diagnosis. There was no differential decline in participants diag-
nosed with one of these four diseases and those who were not. The
only cardiovascular risk factor associated with cognitive perfor-
mance was alcohol consumption.
References
Allaire, J. C., & Marsiske, M. (1999). Everyday cognition: Age and
intellectual ability correlates. Psychology and Aging, 14, 627– 644.
American Heart Association. (2002). 2002 heart and stroke update. Dallas,
TX: Author.
Bäckman, L., Small, B. J., Wahlin, A., & Larsson, M. (2000). Cognitive
functioning in very old age. In F. I. M. Craik & T. A. Salthouse (Eds.),
The handbook of aging and cognition (2nd ed., pp. 499 –558). Mahwah,
NJ: Erlbaum.
Baltes, M. M., & Lang, F. R. (1997). Everyday functioning and successful
aging: The impact of resources. Psychology and Aging, 12, 433– 443.
Baltes, M. M., Maas, I., Wilms, H.-U., Borchelt, M., & Little, T. D. (1999).
Everyday competence in old and very old age: Theoretical consider-
ations and empirical findings. In P. B. Baltes & K. U. Mayer (Eds.), The
Berlin Aging Study: Aging from 70 to 100 (pp. 384 – 402). New York:
Cambridge University Press.
Baltes, P. B., & Kliegl, R. (1992). Further testing of limits of cognitive
plasticity: Negative age differences in a mnemonic skill are robust.
Developmental Psychology, 28, 121–125.
Baltes, P. B., & Lindenberger, U. (1997). Emergence of a powerful
connection between sensory and cognitive functions across the adult life
span: A new window to the study of cognitive aging? Psychology and
Aging, 12, 12–21.
Baltes, P. B., & Mayer, K. U. (Eds.). (1999). The Berlin Aging Study:
Aging from 70 to 100. New York: Cambridge University Press.
Baltes, P. B., & Smith, J. (1997). A systemic–wholistic view of psycho-
logical functioning in very old age: Introduction to a collection of
articles from the Berlin Aging Study. Psychology and Aging, 12, 395–
409.
Beers, M. H., & Berkow, R. (Eds.). (2000). The Merck manual of geriat-
rics. Rahway, NJ: Merck.
Berry, W. D., & Feldman, S. (1985). Sage University Paper Series on
Quantitative Applications in the Social Sciences: Multiple regression in
practice (50, Serial No. 07-050).
Corti, M.-C., Gurlanik, J. M., Salive, M. E., Harris, T., Ferrucci, L., Glynn,
R. J., & Havlik, R. J. (1997). Clarifying the direct relation between total
cholesterol levels and death from coronary heart disease in older per-
sons. Annals of Internal Medicine, 126, 753–761.
Croxon, S. C. M., & Jagger, C. (1995). Diabetes and cognitive impairment:
A community-based study of elderly subjects. Age and Ageing, 24,
421– 424.
Denicoff, K. D., Joffe, R. T., Lakshmanan, M. C., Robbins, J., & Rubinow,
D. R. (1990). Neuropsychiatric manifestations of altered thyroid state.
American Journal of Psychiatry, 147, 94 –99.
Diehl, M., Willis, S. L., & Schaie, K. W. (1995). Everyday problem solving
in older adults: Observational assessment and cognitive correlates. Psy-
chology and Aging, 10, 478 – 491.
Elias, M. F., & Elias, P. K. (1993). Hypertension affects neurobehavioral
functioning: So what’s new? Psychosomatic Medicine, 55, 44 –50.
Elias, P. K., Elias, M. F., D’Agostino, R. B., Silbershatz, H., & Wolf, P. A.
(1999). Alcohol consumption and cognitive performance in the Framing-
ham Heart Study. American Journal of Epidemiology, 150, 580 –589.
Everson, S. A., Helkala, E.-L., Kaplan, G. A., & Salonen, J. T. (2001).
Atherosclerosis and cognitive functioning. In S. R. Waldstein & M. F.
Elias (Eds.), Neuropsychology of cardiovascular disease (pp. 105–120).
Mahwah, NJ: Erlbaum.
Greveson, G. C., Gray, C. S., French, J. M., & James, O. F. W. (1991).
Long-term outcome for patients and carers following hospital admission
for stroke. Age and Ageing, 20, 337–344.
Haan, M. N., Shemanski, L., Jagust, W. J., Manolio, T. A., & Kuller, L.
(1999). The role of APOE ⑀4 in modulating effects of other risk factors
for cognitive decline in elderly persons. Journal of the American Med-
ical Association, 282, 40 – 46.
Hertzog, C., Schaie, K. W., & Gribbin, K. (1978). Cardiovascular disease
and changes in intellectual functioning from middle to old age. Journal
of Gerontology, 33, 872– 883.
Hill, R. D. (1989). Residual effects of cigarette smoking on cognitive
performance in normal aging. Psychology and Aging, 4, 251–254.
Kannel, W. B., & Belanger, A. J. (1991). Epidemiology of heart failure.
American Heart Journal, 121, 951–957.
Krumholz, H. M., Seeman, T. E., Merrill, S. S., Mendes de Leon, C. F., &
Vaccarino, V. (1994). Lack of association between cholesterol and
coronary heart disease mortality and morbidity and all-cause mortality in
persons older than 70 years. Journal of the American Medical Associ-
ation, 272, 1335–1340.
Kumari, M., Brunner, E., & Fuhrer, R. (2000). Minireview: Mechanisms
by which the metabolic syndrome and diabetes impair memory. Journals
of Gerontology: Medical Sciences, 55, 228 –232.
Launer, L. J., Feskens, J. M., Kalmijn, S., & Kromhout, D. (1996).
Smoking, drinking, and thinking: The Zutphen Elderly Study. American
Journal of Epidemiology, 143, 219 –227.
Lindenberger, U., & Baltes, P. B. (1994). Sensory functioning and intel-
ligence in old age: A strong connection. Psychology and Aging, 9,
339 –355.
Lindenberger, U., & Baltes, P. B. (1995). Kognitive Leistungsfähigkeit im
hohen Alter: Ergebnisse aus der Berliner Altersstudie [Cognitive abili-
ties in late life: Findings from the Berlin Aging Study]. Zeitschrift für
Psychologie, 203, 283–318.
Lindenberger, U., & Baltes, P. B. (1997). Intellectual functioning in old
and very old age: Cross-sectional results from the Berlin Aging Study.
Psychology and Aging, 12, 410 – 432.
Lindenberger, U., Mayr, U., & Kliegl, R. (1993). Speed and intelligence in
old age. Psychology and Aging, 8, 207–220.
Lindenberger, U., Singer, T., & Baltes, P. B. (2002). Longitudinal selec-
tivity in aging populations: Separating mortality-associated versus ex-
perimental components in the Berlin Aging Study (BASE). Journals of
Gerontology: Psychological Sciences, 57B, 474 – 482.
Muir, J. L. (1997). Acetylcholine, aging, and Alzheimer’s disease. Phar-
macology, Biochemistry & Behavior, 56, 687– 696.
Perlmuter, L. C., Hakami, M. K., Hodgson-Harrington, C., Ginsberg, J.,
Katz, J., Singer, D. E., & Nathan, D. M. (1984). Decreased cognitive
function in aging non-insulin-dependent diabetic patients. American
Journal of Medicine, 77, 1043–1048.
Raz, N. (2000). Aging of the brain and its impact on cognitive perfor-
mance: Integration of structural and functional findings. In F. I. M. Craik
& T. A. Salthouse (Eds.), The handbook of aging and cognition (2nd ed.,
pp. 1–90). Mahwah, NJ: Erlbaum.
Ryan, C. M. (2001). Diabetes-associated cognitive dysfunction. In S. R.
Waldstein & M. F. Elias (Eds.), Neuropsychology of cardiovascular
disease (pp. 61– 82). Mahwah, NJ: Erlbaum.
Salthouse, T. A. (1991). Theoretical perspectives on cognitive aging.
Hillsdale, NJ: Erlbaum.
Salthouse, T. A. (1993). Speed and knowledge as determinants of adult age
differences in verbal tasks. Journals of Gerontology: Psychological
Sciences, 48, 29 –36.
Salthouse, T. A., & Czaja, S. (2000). Structural constraints on process
explanations in cognitive aging. Psychology and Aging, 15, 44 –55.
Schaie, K. W. (1996). Intellectual development in adulthood: The Seattle
Longitudinal Study. New York: Cambridge University Press.
Schultz, N. R., Dineen, J. T., Elias, M. F., Pentz, C. A., & Wood, W. G.
(1979). WAIS performance for different age groups of hypertensive and
control subjects during the administration of a diuretic. Journal of
Gerontology, 34, 246 –253.
Singer, T., Lindenberger, U., & Baltes, P. B. (2000, April). Memory
 SPECIAL SECTION: HEALTH–COGNITION IN VERY OLD AGE
plasticity in very old age: Testing-the-limits of a mnemonic skill. Poster
session presented at the Cognitive Aging Conference, Atlanta, GA.
Singer, T., Lindenberger, U., & Baltes, P. B. (2003). Plasticity of memory
for new learning in very old age: A story of major loss? Psychology and
Aging, 18, 306 –317.
Singer, T., Verhaeghen, P., Ghisletta, P., Lindenberger, U., & Baltes, P. B.
(2003). The fate of cognition in very old age: Six-year longitudinal
findings in the Berlin Aging Study (BASE). Psychology and Aging, 18,
318 –331.
Starr, J. M., Whalley, L. J., Inch, S., & Shering, P. A. (1993). Blood
pressure and cognitive function in healthy older people. Journal of the
American Geriatrics Society, 41, 753–756.
Steinhagen-Thiessen, E., & Borchelt, M. (1999). Morbidity, medication,
and functional limitations in very old age. In P. B. Baltes & K. U. Mayer
(Eds.), The Berlin Aging Study: Aging from 70 to 100 (pp. 131–166).
New York: Cambridge University Press.
Stevens, J., Cai, J., Pamuk, E. R., Williamson, D. F., Thun, M. J., & Wood,
J. L. (1998). The effect of age on the association between body-mass
index and mortality. The New England Journal of Medicine, 338, 1–7.
Verbrugge, L. M., & Jette, A. M. (1994). The disablement process. Social
Science and Medicine, 38, 1–14.
Verhaeghen, P., & Marcoen, A. (1996). On the mechanisms of plasticity in
young and older adults after instruction in the method of loci: Evidence
for an amplification model. Psychology and Aging, 11, 164 –178.
Verhaeghen, P., Marcoen, A., & Goossens, L. (1993). Facts and fiction
about memory aging: A quantitative integration of research findings.
Journals of Gerontology: Psychological Sciences, 48, P157–P171.
569
Verhaeghen, P., & Salthouse, T. A. (1997). Meta-analyses of age–
cognition relations in adulthood: Estimates of linear and non-linear age
effects and structural models. Psychological Bulletin, 122, 231–249.
Vingerhoets, G. (2001). Cognitive consequences of myocardial infarction,
cardiac arrhythmias, and cardiac arrest. In S. R. Waldstein & M. F. Elias
(Eds.), Neuropsychology of cardiovascular disease (pp. 143–163). Mah-
wah, NJ: Erlbaum.
Wahlin, A., Robins-Wahlin, T.-B., Small, B. J., & Bäckman, L. (1998).
Influences of thyroid stimulating hormone on cognitive functioning in
very old age. Journals of Gerontology: Psychological Sciences, 53,
234 –239.
Waldstein, S. R. (1995). Hypertension and neuropsychological function: A
lifespan perspective. Experimental Aging Research, 21, 321–352.
Waldstein, S. R. (2000). Health effects on cognitive aging. In P. C. Stern
& L. L. Carstensen (Eds.), The aging mind: Opportunities in cognitive
research (pp. 189 –217). Washington, DC: National Academy Press.
Waldstein, S. R., & Elias, M. F. (2001). Neuropsychology of cardiovas-
cular disease. Mahwah, NJ: Erlbaum.
Woo, J., Ho, S. C., Yuen, Y. K., Yu, L. M., & Lau, J. (1998). Cardiovas-
cular risk factors and 18-month mortality and morbidity in an elderly
Chinese population aged 70 years and over. Gerontology, 44, 51–55.
World Health Organization. (1978). Manual of the international statis-
tical classification of diseases, injuries, and causes of death. Based
on the recommendations of the Ninth Revision Conference, 1975, and
adopted by the 29th World Health Assembly. Geneva, Switzerland:
Author.