Cognition and Neuropsychology 491(P. Elias, D•Agostino, Elias, & Wolf, 1995). The classic age
by hypertension model predicts that the combination of
age and hypertension will produced accelerated decline in
cognitive function over time relative to the decline observed
in the absence of hypertension.
A study comparing 10-year change in cognitive function-
ing on the Wechsler Adult Intelligence Scale (WAIS) for
60- to 79-year-old Duke Longitudinal Study (DLS) partici-
pants produced the “rst data consistent with the classic aging
by hypertension model. The DLS started in 1955 with re-
spondents ages 60 to 103 and followed them for 11 repeated
measures until 1976 (see Busse et al., 1985; Siegler, 1983).
Wilkie and Eisdorfer (1971) reported that study participants,
de“ned as clearly hypertensive (diastolic BP 106 mmHg)
and 60 to 79 years of age at entry into the study, exhibited
over a decade signi“cant decline in cognitive functioning
relative to a normotensive cohort (diastolic BP range 65
to 95 mmHg) and a borderline hypertensive cohort (96 to
105 mmHg) of comparable age.
It is sobering to note that no severely hypertensive indi-
viduals survived long enough to participate in the same study
between 70 and 79 years of age. However, both the nor-
motensive and borderline hypertensive individuals exhibited
statistically signi“cant decline in WAIS performance scores
over a 10-year period while the •moderately hypertensiveŽ
participants exhibited signi“cantly more decline over 10
years than the normotensive participants.
This “nding was consistent with the classic aging times
disease interaction model and served as a major stimulus to
other studies, although it involved a very small sample of
subjects and did not involve controls for antihypertension
drugs and hypertension-related disease, which could have ac-
counted for the higher rate of cognitive decline for the hyper-
tensive subjects.
Cross-Sectional Data
There have been several reports of interactions of age and
hypertension for samples of adults less than 40 to 50 years
of age, but “ndings were opposite those predicted by the clas-
sic age by hypertension model. In two studies, differences in
test performance between middle-aged hypertensive and nor-
motensive individuals have been smallerthan the differences
between young adult hypertensive and normotensive individ-
uals. This was true for a wide range of measures of attention,
memory, executive functions, and psychomotor abilities
(Waldstein, 1995). However, the range of ages employed in
these studies makes a difference with respect to interactions.
Wilkie and Eisdorfer (1971) found signi“cant negative corre-
lations between diastolic blood pressure and every subtest in
the Wechsler Adult Intelligence test in a 70- to 79-year-old
cohort, but no signi“cant correlations for 60- to 79-year -old
cohort. However, no evidence of age times blood pressure
interactions was obtained in a large-sample cross-sectional
study involving three age cohorts of 1,695 men and women
(55 to 64, 65 to 74, and 75 to 88 years) participating in the
Framingham Heart Study (P. Elias et al., 1995).
Models advanced by Waldstein (1995) and Waldstein and
Katzel (2001) show that there are a number of physiological
and morphological changes in the brain in the presence of
young adult hypertension that could explain lowered cogni-
tive functioning. However, in terms of the cumulative effects
of blood pressure on the brain, it is dif“cult to explain
why hypertension in old age should not be associated with
disproportionately accelerated change in cognitive function.
Structural and functional changes in the brain seen with hy-
pertension are progressive and cumulative and generally irre-
versible once they occur.
Waldstein (1995) advanced a U-shaped age by hyperten-
sion interaction model to explain the observation, based on
aggregating data from all cross-sectional studies, that young
and elder individuals are more adversely affected by hyper-
tension than middle-aged subjects. This model “ts the cross-
sectional data in a general way, but the data are inconsistent
with contemporary longitudinal studies.
M. Elias et al. (1990) have provided a •signal-to-noise-
ratio explanationŽ of poorer test performance in association
with youth and old age. The argument is that apparently dis-
proportionate effects of hypertension on cognitive test per-
formance in youth affects the cohort against which they are
compared. In youth, hypertension occurs more against a
background of relatively good health than it does in middle or
advanced age. The prevalence hypertension-related patho-
physiology and comorbidity increase with age. Thus, as an
individual ages, hypertension becomes a risk factor seen
against a background of multiple disease and other risk fac-
tors (e.g., diabetes, high cholesterol, high homocysteine
levels, B12 de“ciency; M. Elias, Elias, Robbins, Wolf, &
D•Agostino, 2001). While these confounds can be adjusted
out statistically, this can be the case only if subclinical dis-
eases could be recognized and diagnosed. This objective is
impossible to reach without great cost. One possibility may
be to follow the same individuals over time.
In fact, dif“culties in explaining cross-sectional results
may be due to methodological rather than conceptual (model
building) de“ciencies. Cross-sectional studies are associated
with a number of methodological challenges relating to the
fact that the same individuals are not followed over time.
This problem is particularly acute in case control studies.
Sample bias due to self-selection for studies (M. Elias,