Predictor of Long-term Cardiovascular Mortality in a French Male
Athanase Benetos; Michel Safar; Annie Rudnichi; Harold Smulyan;
Jacques-Lucien Richard; Pierre Ducimetière; ; Louis Guize
the Investigations Préventives et Cliniques (A.B., A.R., L.G.),
Paris, France; Institut National de la Santé et de la Recherche
Médicale (INSERM) U337 (A.B., M.S.) and U258 (J.-L.R., P.D., L.G.),
Paris, France; and Department of Medicine (H.S.), State University
of New York, Syracuse, NY.
Correspondence to Athanase Benetos, MD, PhD, Investigations
Préventives et Cliniques (IPC), 23 rue de Lubeck, 75116 Paris,
Abstract Studies on the usefulness of blood pressure as a
prognostic factor in cardiovascular disease have more
often involved investigations of the levels of diastolic
or systolic blood pressure. However, blood pressure may
be divided into two other components: steady (mean
pressure) and pulsatile (pulse pressure). In this study,
the relationship of pulse pressure to cardiovascular mortality
was investigated in 19 083 men 40 to 69 years old who were
undergoing a routine systematic health examination and
were being followed up after a mean period of 19.5 years.
Subjects were divided into four groups according to age
(40 to 54 and 55 to 69 years) and mean arterial pressure
(<107 and =" src="/math/ge.gif"107 mm Hg). Each group was
further divided into four subgroups according to the pulse
pressure level. A wide pulse pressure (evaluated according to
the quartile group or as a continuous quantitative variable)
was an independent and significant predictor of all-cause,
total cardiovascular, and, especially, coronary mortality
in all age and mean pressure groups. No significant
association between pulse pressure and cerebrvascular
mortality was observed. In conclusion, in a large
population of men with a relatively low cardiovascular
risk, a wide pulse pressure is a significant independent
predictor of all-cause, cardiovascular, and, especially,
Key Words: mortality • blood pressure • pulse pressure •
standard definition of hypertension includes an elevation
of SBP and/or DBP. In adhering to this definition, many
epidemiological investigations of the long-term risks of
the disease fail to include patients with high SBP but
normal or low DBP.1–3 In so doing, patients
with the widest PPs were excluded. Despite this systematic entrance
bias, these studies have shown increasing cardiovascular risk
of higher SBP and PP at the same DBP.4 In 1971,
investigators with the Framingham Heart Study5
emphasized the greater risks of an elevated SBP compared
with the DBP in patients over the age of 55. Although
probably involved, PP was rarely mentioned.6
Although a large PP measured at the brachial artery with use
of the cuff method is not an accurate representation of the
proximal aortic PP, it does suggest a stiffened aorta. Such
stiffening, through a variety of mechanisms,7,8
tends to raise the SBP and lower the DBP. The former,
which increases left ventricular pulsatile work, is
associated with left ventricular hypertrophy and requires
a greater coronary blood flow. The latter reduces the
pressure on which coronary flow is dependent, and together
they increase the vulnerability of the heart to ischemia. All
this suggests that PP itself could be a major predictor of
Evidence is beginning to accumulate in support of this view.
In 1994, Madhavan et al9 reported a series of 2207
untreated hypertensive subjects followed for an average
of 4.8 years. This study showed that subjects in the
upper tertile of pretreatment PP (=" src="/math/ge.gif"63 mm
Hg) had a greater mortality than those in the lower tertiles
and that PP, but not SBP or DBP, was an independent predictor
of myocardial infarction. In a later study4 from
the same group, an expanded number of 5730 treated and
untreated hypertensive patients were reported. After
adjustment for other risk factors, PP was the only
measure of blood pressure significantly and independently
related to the in-treatment incidence of myocardial
1989, a study from France10 described findings for 18 336
men and 9351 women who had been followed for an average of 9.5
years. These were unselected subjects who volunteered for
free medical examinations. Blood pressure data were
divided statistically into steady and pulsatile
components. There was an association between the pulsatile
component and left ventricular hypertrophy in both sexes, as
well as a positive correlation with death from coronary artery
disease in women. The results in both sexes were weakened by
the small number of deaths in each group, which was a
consequence of the relatively short duration of
purpose of the present report was to assess the effect of
the initial PP on the long-term risks of cardiovascular mortality
in the male subjects of this self-selected cohort who had been
followed for 10 additional years, to a total follow-up period
of 19.5 years.
The French Public Health System (Securité Sociale–CNAM) offers
to all working and retired persons and their families free
health examinations every 5 years. The IPC is one of the
largest medical centers in France; since 1970, 15 000
examinations of persons living in the Paris area have
been performed annually. In the present study, we present
data that describe a population composed of all 19 083
men aged 40 to 69 years who underwent a systematic health
check-up in the center IPC during the period of May 1972
through May 1977. These subjects were selected from the 39 495 men
examined during this period at the Center IPC. Clerks and
working executives represented 88% of the studied
blood pressure was measured by a nurse in the right arm
with the use of a manual sphygmomanometer. After a 10-minute
rest period, pressure was measured three times, and the mean
of the last two measurements was calculated. The first and the
fifth Korotkoff's phases were used to define SBP and DBP.
Smoking status was assessed with a self-administered
questionnaire with dichotomic (yes or no) questions
regarding tobacco use. Plasma cholesterol was measured
with a Technicon SMA 12.
follow-up study period ended on December 1994 (mean follow-up,
19.5 years). Deceased subjects were identified from the
mortality records of the Institut National de
Statistiques et d'Etudes Economiques. A patient of our
cohort was considered to be deceased when he had the
first name, last name, sex, and date of birth as a person
recorded in the Institut National de Statistiques et d'Etudes
Economiques mortality records during the period of the
follow-up. Through the use of this matching procedure,
the identification error was <1%. Only subjects
fulfilling all four of these criteria were considered to
be deceased. Individuals matching for sex, last name, and
only one of the other two criteria were excluded from the
study. All other subjects were considered to be alive at
the end of the follow-up period. On the basis of this
procedure, 3653 subjects of our cohort were considered to
have died during the follow-up period. Causes of mortality
were taken from the death certificates. These data were
provided by the Department of Mortality of the INSERM
(Unit SC 8). Causes of death were codified according to
the International Classification of Disease
(eighth revision until 1978 and ninth revision after
Subjects were divided into four groups according to age (young, 40
to 54 years; older, 55 to 69 years) and MBP
(MBP=2/3DBP+1/3SBP; low MBP, <107 mm Hg; high MBP, ="
src="/math/ge.gif"107 mm Hg). In each of the four groups
identified according to age and MBP, the role of PP was
studied either as a qualitative parameter (separation
according to the four quartiles of PP defined in the whole
population) or as a continuous quantitative parameter.
The qualitative separation was accomplished by dividing
each group into PP quartiles defined as PP145,
45<PP250, 50<PP3<65, and PP4="
src="/math/ge.gif"65 mm Hg. This classification is the
closest to the quartiles distribution in the whole
population by steps of 5 mm Hg.
comparisons among the PP groups within each of the four
original groups (defined according to age and MBP), mean values
of morphometric parameters, blood pressure, and total
cholesterol were compared with the use of a one-way
analysis of variance and a 2 test for tobacco
status; deaths for the different causes of mortality were
compared with the use of a trend 2 test; and
the difference in survival probability for the different causes
of mortality were tested by using a Cox analysis with
adjustments for age, total cholesterol, and tobacco
also assessed (in a multiple logistic regression) the respective
roles of MBP and PP, both of which are considered to be
continuous quantitative variables. To test whether the
effects of one variable was affected by the other, we
introduced the interactive term. In the case of
significant interaction MBPxPP, we reconstructed the effect
of each variable in the hypothetical populations corresponding
to selected levels (mean, first, and third quartiles) of the
statistical analyses were performed with SAS software.
1 provides a summary of the mean values of morphometric
parameters, blood pressure, total cholesterol, and tobacco
consumption among the different groups according to the
PP level. In both groups of young patients (Table 1A),
age, weight, height, BMI, and tobacco consumption did not
show any clinically significant differences among the
four PP subgroups. SBP levels progressively increased
from the first PP group to the higher PP groups, whereas
mean values of DBP were significantly lower in the subgroups
with the higher PP. Total cholesterol was increased from
groups PP1 to PP4. In the two
groups of older patients (Table 1B), age was higher in
the subgroups with the higher PP values. For the other
parameters, the same trends were observed as for younger patients.
2 shows death rates for the different causes of mortality.
All-cause, cardiovascular, and coronary heart disease
mortality rates were constantly higher as PP increased.
observed in all groups (young versus old and low MBP versus
high MBP). However, no significant association was observed
between PP and cerebrovascular mortality.
shows survival probabilities for cardiovascular mortality,
with adjustments for age, total cholesterol, and tobacco
consumption. In the four groups of subjects, lower
survival probabilities were observed in subjects with
higher PP, especially in the subgroup with PP values of
=" src="/math/ge.gif"65 mm Hg. The differences among PP
subgroups progressively increased throughout the follow-up period.
These differences become significant after 10 years of
follow-up. Similar patterns were observed in survival
probabilities for all-cause and coronary heart mortality
but not for cerebrovascular mortality (data not shown).
Finally, the respective roles of MBP and PP, considered in this
analysis as continuous quantitative variables, were
evaluated after adjustment for age (Table 3 and Fig 2).
In younger patients, when MBP and PP were used together
in the model (model 3 in Table 3A), both are highly
significant predictors for all-cause, noncardiovascular,
total cardiovascular, and coronary heart disease mortality.
The effects of the MBP and PP were additive (no interaction
MBPxPP was observed). For cerebrovascular mortality, MBP
but not PP was a strong predictor. In the older subjects,
MBP and PP were both predictors for the different causes
of mortality (Table 3B). However, difference from what we
observed in younger subjects, a significant negative
interaction was observed in older individuals between PP and MBP
for total cardiovascular (P=.012) and coronary heart
disease (P=.013) mortality. After that, we
evaluated the effect of PP at three different levels of
MBP (mean, first, and third quartiles) (Fig 2, top) and
the effect of MBP at the same three levels of the PP (Fig
2, bottom). This analysis showed that both PP and MBP
were significant predictors for total cardiovascular and
coronary heart disease mortality, with the more pronounced effects
of each parameter when values of the other parameter were
Therapeutic decision making and management in patients with
mild-to-moderate hypertension are complicated by the wide
variation in their clinical characteristics. Clinicians
therefore have sought more precise means to describe the
outlook for individual patients. One approach has been to
use different measures of baseline blood pressure, such
as 24-hour recordings and variability of blood pressure to
prognostically stratify patients. Our study is the first
to clearly show that in a large male unselected
population with a relatively low risk (volunteers for
free medical examinations), PP measurement may help in
the evaluation of the individual risk and therefore in the
therapeutic decision making. Although the use of a single
blood pressure measurement reduced statistical power, our
results demonstrate that increased PP is a predictor of
global mortality and cardiovascular mortality,
independent of other known cardiovascular factors such as
age, mean blood pressure, total cholesterol, and smoking.
Interestingly, increased PP was a predictor of coronary heart
disease mortality, whereas its predictive value was not
significant for cerebrovascular mortality.
previous analysis of the same cohort was not able to establish
a significant relationship between PP and cardiovascular
mortality in male subjects. We believe that this was due
to the fact that in this low-risk population, the
duration of the follow-up of the previous analysis was
too short (9 years) to evaluate the role of PP. As shown
from the survival curves in the present study (Fig 1)
differences among the four subgroups according to the PP
levels, only become clear after
10th year of follow-up.
Physiologically, PP describes the oscillation around the mean
arterial pressure (calculated as DBP+1/3PP) and is influenced
by hemodynamic mechanisms that differ from those controlling
mean arterial pressure. MBP is the pressure that would be
present in the aorta and its major arteries during a
given cardiac cycle if the cardiac output was
nonpulsatile.7,8 Although mean arterial pressure
remains nearly constant along the arterial tree, PP increases
markedly from central to peripheral arteries as a consequence
of a substantial increase in SBP and a slight lowering of DBP.
At a given stroke volume and velocity of ventricular ejection,
the mechanisms influencing PP are related to the status of
conduit arteries, that is, the viscoelastic properties of
the arterial wall and timing of the reflected waves.
Increased stiffness and earlier wave reflections within
the thoracic aorta increase the PP due to an increase in
SBP and a decrease in DBP.7,11 Alternatively, increased
stroke volume or ventricular ejection rate may be
responsible for an increase in SBP with no change in DBP.
In the present study, we showed that the widest PPs were
due to both an increase in SBP and a decrease in DBP. Thus,
the changes in PP may be considered as markers of increased
arterial stiffness, with consequences for the cardiovascular
present study, in the younger individuals (40 to 54 years), PP
and MBP have additive effects in the evaluation of the risk
for the different causes of mortality (except for
cerebrovascular, for which MBP but not PP is a strong
predictor). Interestingly, in older individuals (55 to 70
years), a negative interaction was observed between MBP
and PP, suggesting that the effect of PP in total
cardiovascular and, especially, coronary heart disease
mortality is enhanced in individuals with low MBP (Table 3b
and Fig 2). Taken together, these results show that increased
PP was a major predictor of coronary mortality even in the
presence of values of MBPs conventionally accepted as
being within the normal range (MBP <107 mm Hg). Indeed,
the coronary circulation is the only circulation with
volume flow that is governed by the DBP rather than the
SBP.12 Thus, any decrease in DBP as a
consequence of increased arterial stiffness may limit coronary
blood flow, particularly in the presence of associated
stenosis of the coronary arteries.11 In
addition to a decrease in DBP, increased arterial
stiffness is responsible for an increase in SBP, which,
through increased end-systolic stress, promotes cardiac
hypertrophy. In hypertensive subjects, a positive and
significant association has been previously observed between
increased PP and increased cardiac mass independent of mean
arterial pressure.13 Therefore, it is reasonable to
suggest that an increased PP, through both these
mechanisms, increases coronary risk.
present study, no comparable risk was observed for PP in
the cerebral circulation. This finding may be in part due to
a loss of statistical power as a consequence of the relatively
small number of cerebrovascular deaths compared with coronary
ischemic deaths. However, in our cohort, MBP was the most
significant predictor of cerebrovascular mortality, and
mean arterial pressure (but not PP) is the perfusion
pressure of the cerebral circulation.
conclusion, the present study has shown that in male subjects with
normal or elevated mean arterial pressure, increased PP is a
strong predictor of general and cardiovascular mortality,
affecting especially the coronary but not the
Return to Medicardium Main Information Page
Medicardium Protocol & Suggested Use
Medicardium Ingredients & Instruction
This study was performed with the help of INSERM (Paris). We
thank the Caisse Nationale
d'Assurance Maladie (CNAM) for support of this study.
The authors are grateful to Jean-François Morcet and Jean
Pierre Huby for their help and advice in analysis of the
data. We thank Dr Anne Safar for help in preparing the manuscript.
Received April 8, 1997; first decision April 24, 1997; accepted
June 25, 1997.
1. Collins R, Peto R, MacMahon S, Hebert P, Fiebach NH,
Everlein KA, Godwin J, Quizilbash N, O'Taylor J, Hennekens
C. Blood pressure, stroke, and coronary heart disease, Part
2: short-term reductions in blood pressure: overview of
randomised drug trials in their epidemiological context.
2. MacMahon S, and Rodgers A. Blood pressure,
antihypertensive treatment and stroke risk. J Hypertens.
3. Polk FB, Cutter G, Dugherty RM. Hypertension detection
and follow-up program: baseline physical examination
characteristic of the hypertensive participants.
4. Fang J, Madhavan S, Cohen H, Alderman MH. Measures of
blood pressure and myocardial infarction in treated
hypertensive patients. J Hypertens.
5. Kannel WB, Gordon T, Schwartz MJ. Systolic versus
diastolic blood pressure and risk of coronary heart disease:
the Framingham study. Am J Cardiol.
6. Kannel WB. Hypertension and the risk of cardiovascular
disease. In: Laragh JH, Brenner BM, eds. Hypertension,
Pathophysiology, Diagnosis, and Management. New York,
NY: Raven Press Publishers; 1990:101-117.
7. Nichols WV, O'Rourke MF. McDonald's Blood Flow in
Arteries: Theoretic, Experimental, and Clinical Principles,
ed 3. London/Melbourne: E Arnold; 1990:77–142, 216–269,
8. Safar ME. Pulse pressure in essential hypertension:
clinical and therapeutical implications. J Hypertens.
9. Madhavan S, Ooi WL, Cohen H, Alderman MH. Relation of
pulse pressure and blood pres9. sure reduction to the
incidence of myocardial infarction. Hypertension.
10. Darne B, Girerd X, Safar M, Cambien F, Guize L.
Pulsatile versus steady component of blood pressure: a
cross-sectional analysis and a prospective analysis on
cardiovascular mortality. Hypertension.
11. Kelly R, Tunin R, Kass D. Effect of reduced aortic
compliance on left ventricular contractile function and
energetics in vivo. Circ Res.
12. Hoffman JIE. A critical view of coronary reserve.
Circulation. 1987;75(suppl I):I-6-I-11.
13. Pannier B, Brunel P, El Aroussy W, Lacolley P, and
Safar ME. Pulse pressure and echocardiographic findings in
essential hypertension. J Hypertens.