1194
Expected Gains in Life Expectancy From
Various Coronary Heart Disease
Risk Factor Modifications
Joel Tsevat, MD, MPH; Milton C. Weinstein, PhD; Lawrence W. Williams, MS;
Anna N.A. Tosteson, ScD; and Lee Goldman, MD, MPH
Background. Despite much evidence that modifying risk factors for coronary heart disease
decrease morbidity and mortality, little is known about the impact of risk-factor
modification on life expectancy.
Methods and Results. We used the Coronary Heart Disease Policy Model, a state-transition
computer simulation of the US population, to forecast potential gains in life expectancy firom
risk-factor modification for the cohort of Americans turning age 35 in 1990. Among 35-year-old
men, we projected that the population-wide increase in life expectancy would be about 1.1 years
from strict blood pressure control, 0.8 years from smoking cessation, 0.7 years from reduction of
serum cholesterol to 200 mg/dl, and about 0.6 years from weight loss to ideal body weight. For
women, reducing cholesterol to 200 mg/dl would have the greatest estimated impact-a gain of 0.8
years -whereas smoking cessation, blood pressure control, or weight loss would yield populationwide gains of 0.7, 0.4, and 0.4 years, respectively. Gains for 35-year-old individuals having a given
risk factor are greater. We estimate that, on average, male smokers would gain 2.3 years from
quitting smoking; males with hypertension would gain 1.1-5.3 years from reducing their diastolic
blood pressure to 88 mm Hg; men with serum cholesterol levels exceeding 200 mg/dl would gain
0.5-4.2 years from lowering their serum cholesterol level to 200 mg/dl; and overweight men would
gain an average of 0.7-1.7 years from achieving ideal body weight. Corresponding projected gains
for at-risk women are 2.8 years from quitting smoking, 0.9-5.7 years from lowering blood pressure,
0.4-6.3 years from decreasing serum cholesterol, and 0.5-1.1 years from losing weight. Eliminating
coronary heart disease mortality is estimated to extend the average life expectancy of a 35-year-old
man by 3.1 years and a 35-year-old woman by 3.3 years.
Concluions. Population-wide gains in life expectancy from single risk-factor modifications are
modest, but gains to individuals at risk can be more substantial. (Circulaton 1991;83:
1194-1201)
can
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
heart disease (CHD) in the United States. An
estimated 5.4 million Americans have symptomatic CHD, and countless others have asymptomatic
CHD. There are 680,000 hospital admissions for
myocardial infarction and half a million deaths from
CHD each year.1 The annual cost of CHD in 1980
was approximately $80 billion, with direct health
costs totaling $30 billion.2
Much morbidity, mortality, and expense could be
prevented through risk factor modification. Cigarette
smoking, diastolic blood pressure, serum cholesterol
level, and body weight are modifiable risk factors3-6
See p 1452
for CHD, and cigarette smoking is the foremost
preventable cause of death in the United States.7 The
medical profession, among others, has launched pub-
From the Division of Clinical Epidemiology (J.T., L.W.W.,
A.N.A.T., L.G.), the Cardiovascular Division (L.G.), and the
Division of General Medicine, Departments of Medicine, Beth
Israel Hospital and Brigham and Women's Hospital, Harvard
Medical School; and the Department of Health Policy and Management, Harvard School of Public Health (M.C.W., A.N.A.T),
Boston.
Presented in abstract form at the 11th Annual Meeting of the
Society for Medical Decision Making in Minneapolis, Minn., in
October 1989.
Supported in part by grant 86-3192 from the Henry J. Kaiser
Family Foundation, Menlo Park, Calif., and grant 1R01-HS06258-01 from the Agency for Health Care Policy and Research.
Address for reprints: Lee Goldman, MD, MPH, Division of
Clinical Epidemiology, Brigham and Women's Hospital, 75
Frances Street, Boston, MA 02115.
Received October 17, 1989; revision accepted November 6,
1990.
It is difficult to overstate the impact of coronary
Tsevat et al Gains in Life Expectancy
lic health campaigns, as exemplified by the fact that
the American Medical Association has "declared war
on cholesterol."8
To date, however, there is little information on the
impact of risk factor modification on life expectancy.
How much would life expectancy be prolonged if
Americans stop smoking, lower their serum cholesterol levels, control their blood pressures, or lose
weight? Using the Coronary Heart Disease Policy
Model,9-12 a computer simulation of CHD in the US
population, we have forecasted potential gains in
both population-wide and individual life expectancy
from various risk factor modifications for the cohort
of Americans turning age 35 years in 1990.
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Methods
The Model
The Coronary Heart Disease Policy Model9-12 is a
state-transition computer simulation model that consists of three submodels: the Demographic-Epidemiologic Submodel, the Bridge Submodel, and the
Disease History Submodel (Figure 1).
The Demographic-Epidemiologic Submodel assesses each individual's risk of developing CHD
based on age, sex, smoking status (no, yes; if yes,
average number of cigarettes per day), diastolic
blood pressure (.94 mm Hg, 95-104 mm Hg, >105
mm Hg), relative weight (.109% of ideal, 110-129%
of ideal, >130% of ideal), and serum cholesterol
level (.249 mg/dl, 250-299 mg/dl, .300 mg/dl).
Based on the categories for each of these factors, the
entire US population age 35-84 is divided into 5,400
cells, each with a specific value for each factor. For
simplicity, the model's outputs are collapsed into
10-year age ranges, thus yielding 540 strata.
For the current analyses, subjects turning age 35
years in 1990 who are free of clinical coronary heart
disease enter the Demographic-Epidemiologic Submodel. That cohort is then followed for 50 years
during which, in any given year, persons may 1)
develop clinical CHD (in which case they move to the
Bridge Submodel, 2) age by 1 year without developing CHD and proceed to the subsequent year's cell in
the Demographic-Epidemiologic Submodel, 3) die of
other causes, or 4) turn age 85 years and exit the
model.
The Bridge Submodel characterizes subjects for
the first 30 days after they develop CHD, and the
Disease History Submodel considers all events that
occur to such persons after that 30-day period:
interventions such as coronary artery bypass surgery,
recurrent CHD events, CHD deaths, and deaths
from other causes.
Assumptions of the Model
The Demographic-Epidemiologic Submodel was initially constructed using the 1980 US census13 and the
estimated proportion of persons without CHD.14 The
distribution of smoking status, diastolic blood pressure,
relative weight, and serum cholesterol level in the US
Non-CHD
Deaths
1195
CHD
Deaths
FIGURE 1. Diagram of the Coronary Heart Disease Policy
Model. Persons tuming age 35 years enter the DemographicEpidemiologic Submodel. Possible state transitions during the
next 50 years are indicated by the arrows. All 85-year-old
survivors exit the model (see text). CHD, coronary heart
disease.
population was taken from the Second Health and
Nutrition Examination Survey (HANES),15 and the
four risk factor distributions were, for simplicity, assumed to be independent.
The number of US residents turning age 35 years
in 1990 was estimated from projections of the US
Bureau of the Census.16,17 Risk factor changes with
age were estimated from cross-sectional population
data.'5
For each of the 540 cells, relative risk (,l) coefficients for CHD incidence and for total mortality were
based on data from the Framingham Heart Study.3
We obtained the age- and gender-specific coefficients
for all risk factors in the Demographic-Epidemiologic
Submodel from the Framingham Heart Study's 30year follow-up. We then smoothed these raw coefficients by age for each gender with a weighted leastsquares regression of the Framingham coefficients on
the midpoint of each 10-year age interval, where the
weights represented the inverse of the coefficients'
standard errors. CHD incidence rates for ages 35-74
years were based on the Framingham Heart Study18
with a secular adjustment for the decline in CHD
incidence19 since the beginning of the Framingham
Heart Study and with extrapolation to ages 75-84
years and linear interpolation to smooth the agespecific annual rates.9 Mortality rates for CHD and
other causes were based on US vital statistics and
were assumed to apply to a population with risk
factors distributed according to the HANES data.
The Disease History Submodel tracks the cohort of
patients who survive the first month after their CHD
event (i.e., graduates of the Bridge Submodel)
through 12 states. These patients with CHD are
categorized according to their previous history:
whether they are in their first or subsequent year
after the initial event and whether their history
includes one or more cardiac arrests. one or more
myocardial infarctions, one or more coronary artery
bypass graft operations, any combination of these, or
none of these (i.e., angina only). Each of the 12
resulting states is further subclassified by age and
1196
Circulation Vol 83, No 4 April 1991
gender, making a total of 1,200 Disease History cells
(50 ages x 2 genders x 12 CHD states).
During each model year, patients in the Disease
History Submodel face the chance of having a cardiac
arrest, a myocardial infarction, or a coronary artery
bypass operation (or any combination of these). Each
event has a specific case-fatality rate tailored to the
cell in which the person started that year. Methods
used to calculate these event rates are described in
detail elsewhere.9 Survivors of a given year begin the
next year in the appropriate Disease History cell.
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Interventions
Risk factor interventions, such as reductions in
diastolic blood pressure, are simulated interactively
with the Coronary Heart Disease Policy Model.
Primary interventions, that is, interventions aimed at
preventing CHD in persons with no history of CHD,
are simulated by adjusting the Demographic-Epidemiologic Submodel. The user can adjust risk factors
by a relative amount (e.g., a 10% reduction in
cholesterol), adjust risk factors by an absolute
amount (e.g., a 5-mm Hg decrease in diastolic blood
pressure), or redefine the mean value (e.g., change
the mean number of cigarettes smoked per day from
12 to zero) for any or all cells. In the DemographicEpidemiologic Submodel, the probability of dying
from causes other than CHD is a function of not only
age and gender but also diastolic blood pressure and
mean number of cigarettes smoked; thus, by adjusting diastolic blood pressure or smoking, one simultaneously adjusts the risk of developing CHD and of
dying from non-CHD causes.
We simulated secondary prevention (of recurrence
of CHD events) by reducing the following Disease
History probabilities: the probability of an arrest,
myocardial infarction, or coronary artery bypass operation given a history of angina, arrest, myocardial
infarction, or coronary artery bypass operation; and
the mortality rate from chronic coronary artery disease. We determined the percent reductions for
these variables as follows: First, at the midpoint of
each 10-year age interval, that is, at ages 40, 50, 60,
70, and 80 years, we used the frequency distribution
and the mean risk factor values in the DemographicEpidemiologic Submodel for persons having their
first CHD "event" (angina, myocardial infarction, or
cardiac arrest with or without myocardial infarction).
To calculate a percent reduction in CHD risk from a
given intervention, we applied those values to the
formula
% decrease in CHD risk = (1-e- PA)x 100
where ,B represents the relative risk coefficient for CHD
incidence and A is the difference between the preintervention and postintervention risk factor values.
For example, among 35- to 44-year-old women
developing CHD, 53.2% had a "low" serum cholesterol level of less than 250 mg/dl (mean, 195.2
mg/dl), 27.5% had an "intermediate" level of 250-
299 mg/dl (mean, 267.4), and 19.3% had a "high"
level exceeding 300 mg/dl (mean, 326.8); the ,B for
this gender and age group was 0.01309. In the
intervention in which we lowered the serum cholesterol level to 240 mg/dl for everyone whose serum
cholesterol level was greater than 240 mg/dl, there
was a 0.4-mg/dl reduction in the mean serum cholesterol level of the "low" group (generated by those
whose levels were 241-249 mg/dl), a 27.4-mg/dl
reduction in the "intermediate" group (from 267.4 to
240 mg/dl), and an 86.8-mg/dl decrease in the "high"
group (from 326.8 to 240 mg/dl). Using the formula,
we calculated the percent reduction in CHD risk for
the "low" (0.5% reduction), "intermediate" (30.1%
reduction), and "high" (67.9% reduction) groups and
took a weighted average of the three (weighted
average, 21.6%). Last, we reduced the probabilities
of each of the four Disease History Submodel variables mentioned above by that weighted average.
The last step in modeling a secondary intervention
was to adjust the probability of dying from non-CHD
causes (which in the Disease History Submodel,
unlike the Demographic-Epidemiologic Submodel, is
not done automatically). Here, for each smoking and
blood pressure intervention, we took the percent
reduction in the non-CHD death rate calculated in
the Demographic-Epidemiologic Submodel and applied it to the Disease History Submodel as well.
Calculating Life Expectancy
Persons surviving to age 85 years exit the Coronary
Heart Disease Policy Model. To achieve an accurate
estimate of life expectancy, it was necessary to add
the life expectancy at age 85 years to the surviving
85-year-old persons. Thus, in the baseline run, men
reaching age 85 years were credited with an additional 5.0 years of life expectancy, and women were
credited with 6.4 years.20
We used two different methods to estimate life
expectancy for interventions. First, to simulate eliminating CHD, we ran the model using a zero probability of developing CHD. We then adjusted the
incremental life expectancy for 85-year-old persons
by reducing the expected number of deaths each year
after age 85 years by the fraction of deaths caused by
ischemic heart disease in persons over age 85 years
(35% for men, 36% for women)21 and subjecting the
cohort of 85-year-old persons to the mortality rate
for causes of death other than ischemic heart disease.
The derived life expectancy for 85-year-old persons if
CHD is eliminated was 6.7 years and 8.6 years for
men and women, respectively. Here, we assumed no
change in mortality from noncardiac causes. To the
extent that persons not dying of CHD might instead
die at a more rapid rate from a competing disease,
this method may overestimate life expectancy; on
the other hand, if CHD were eliminated, there
might be a reduction in mortality from related
diseases such as cerebrovascular disease, rendering
our estimates too low.
Tsevat et al Gains in Life Expectancy
In a second method, we calculated life expectancy
for a 35-year-old person by adjusting simulated life
expectancy by a correction factor to allow a small
gain after age 85 years. The correction factor was the
ratio of the total life-years accumulated by persons
dying between ages 35 and 84 years in the baseline
run to the life-years accumulated between ages 35
and 84 years in a given intervention run. Here, a
life-year was defined as a year of life after age 35
years; that is, a member of the cohort who dies at age
55 years would get 20 life-years. The baseline life
expectancy at age 85 years was multiplied by this
ratio, and this adjusted life expectancy for 85-yearold persons was then incorporated into the model to
calculate a life expectancy for 35-year-old persons.
When we compared the two methods for estimating
life expectancy if CHD is eliminated, the outcomes
were nearly identical. Our results for all simulations are
presented based on the second approach.
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Population-Wide Simulations
Population-wide simulations distribute gains in
life expectancy across all 35-year-old personsthose who have the given risk factor and those who
do not. We considered the following populationwide interventions:
Serum cholesterol. 1) Reducing serum cholesterol
level to 240 mg/dl if greater than 240 mg/dl; 2)
reducing serum cholesterol level to 200 mg/dl if
greater than 200 mg/dl; 3) reducing serum cholesterol level by 10 mg/dl; 4) reducing serum cholesterol
level by 20 mg/dl.
Smoking. 1) Reducing the mean number of cigarettes smoked by 50%; 2) eliminating smoking.
Diastolic blood pressure. 1) Reducing diastolic
blood pressure to 88 mm Hg if greater than 88
mm Hg.
Relative weight. 1) Reducing weight to ideal body
weight.
In a given intervention simulation, the risk factor
was modified to a certain level and maintained at that
level. For example, in one intervention, serum cholesterol level was reduced to 240 mg/dl and not
allowed to rise over time. To calculate gains in life
expectancy, population-wide interventions were compared with a no-intervention baseline run in which
risk factor distributions were allowed to change naturally. That is, in the baseline run, 35-year-old persons assumed the risk factor profile of 45-year-old
persons when they turned age 45 years, of 55-yearold persons when they turned age 55 years, and so on.
Thus, because smoking is less prevalent in the elderly,15 over time a certain proportion of smokers quit
smoking even in the baseline run. Conversely, diastolic blood pressures and serum cholesterol levels
are higher in the elderly,15 so that the levels rose
accordingly in the baseline run.
At-Risk Individuals Simulations
We considered the following interventions for atrisk 35-year-old persons.
1197
Serum cholesterol. 1) Reducing serum cholesterol
level to 200 mg/dl if 200-239 mg/dl; 2) reducing
serum cholesterol level to 200 mg/dl if 240-299
mg/dl; 3) reducing serum cholesterol level to 200
mg/dl if 300 mg/dl or greater.
Smoking. 1) Reducing the mean number of cigarettes smoked by 50%; 2) quitting smoking.
Diastolic blood pressure. 1) Reducing diastolic
blood pressure to 88 mm Hg if 90-94 mm Hg; 2)
reducing diastolic blood pressure to 88 mm Hg if
95-104 mm Hg; 3) reducing diastolic blood pressure
to 88 mm Hg if 105 mm Hg or greater.
Relative weight. 1) Reducing weight to ideal body
weight if weight exceeds ideal body weight by less
than 30%; 2) reducing weight to ideal body weight if
weight exceeds ideal body weight by 30% or more.
Whereas the "population-wide" analysis sought to
assess the gains from an intervention when compared
with the "natural course of events," the "at-risk
individuals" analysis was structured to answer a
different question: How much life expectancy can a
35-year-old individual having a cardiac risk factor
expect to gain if he or she modified that risk factor
compared with maintaining that risk factor at its level
at age 35 years?
To model the gain for an at-risk individual, we
again compared the life expectancy in a baseline run
to the life expectancy after a given intervention.
Here, however, for each risk factor simulation, we
developed a separate baseline run in which the risk
factor of interest was "frozen" at its level at age 35
years. For example, to simulate the benefits to smokers of quitting smoking, smokers quit smoking in the
intervention, but in the baseline run, they continued
to smoke the same number of cigarettes per day until
they exited the Demographic-Epidemiologic Submodel. (Diastolic blood pressures and serum cholesterol levels were allowed to rise with age in the
baseline smoking and smoking intervention simulations.) Similarly, when determining the benefits of
controlling hypertension, we froze (only) their diastolic blood pressures in the baseline run and modified it in the intervention run.
Because assumptions differ between the "population-wide" and "at-risk" individual simulations, population-wide gains cannot be directly mapped onto
gains for at-risk individuals. By freezing risk factor
levels at their level at age 35 years in baseline "at-risk
individuals" simulations, gains from cholesterol and
blood pressure interventions are slightly diminished,
and gains from smoking cessation are slightly magnified compared with the analogous population-wide
interventions. Gains from weight loss are not affected
by the differing assumptions.
Sensitivity Analyses
We performed sensitivity analyses on each intervention to determine how varying the P3 coefficients
would affect the outcome. For each simulation, the
given /3 coefficient was varied from one standard
error less than its mean to one standard error greater
1198
Circulation Vol 83, No 4 April 1991
TABLE 1. Population-Wide Gains in Life Expectancy for 35-Year-Old Persons
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
LE (yr)
Intervention
Male
Female
None
38.2
44.6
Reduce cholesterol level
To 240 mg/dl if >240 mg/dl
38.5
45.0
To 200 mg/dl if >200 mg/dl
38.9
45.4
By 10 mg/dl
44.7
38.5
By 20 mg/dl
38.7
44.9
Reduce number of cigarettes smoked
38.7
By 50%
44.9
39.1
Eliminate smoking
45.2
Reduce diastolic blood pressure
39.3
45.0
To 88 mm Hg if >88 mm Hg
Reduce weight
To ideal if >ideal
44.9
38.8
41.4
Eliminate CHD
47.8
Life expectancies and gains in life expectancy rounded to nearest 0.1 year.
LE, life expectancy; CHD, coronary heart disease.
than its mean. Using a smaller value for the ,B
coefficient implies that the risk factor has a smallerthan-expected impact on the incidence of CHD; thus,
intervening on that risk factor would yield less impressive gains in life expectancy than those under
baseline assumptions. Conversely, a larger ,B coefficient indicates that the risk factor has a more substantial impact on the overall incidence of CHD,
meaning that intervening would be more fruitful than
those under baseline assumptions.
Results
Calibration of the Model
For the baseline (no intervention) run, the model
calculated life expectancy for a 35-year-old man to be
38.2 years, nearly identical to the projection by 1980
national vital statistics of 38.1 years for men 35 years
of age.20 For 35-year-old women, the model calculated a life expectancy of 44.6 years, only 0.2 years
greater than published life-table life expectancy.20
Population-Wide Gains in Life Expectancy
Baseline analysis. For population-wide simulations,
the estimated population-wide gains in life expectancy
for persons turning age 35 years in 1990 ranged from
0.2 to 1.1 years (Table 1). The largest estimated gains in
male population-wide life expectancy were realized
through reducing diastolic blood pressure to 88 mm Hg
or eliminating smoking. For women, reducing serum
cholesterol level to 200 mg/dl and smoking cessation
were projected to have the greatest impacts on population-wide life expectancy. If by reducing risk factors
or by other interventions it were possible to eliminate
CHD, the model projected that the life expectancies of
a 35-year-old man and woman would increase by 3.1
and 3.3 years, respectively.
Sensitivity analyses. Reducing the coefficients by
one standard error reduced the projected gain in
Male
Gain in LE (yr)
Female
...
0.3
0.7
0.2
0.4
0.5
0.8
0.2
0.3
0.4
0.8
0.4
0.7
1.1
0.4
0.6
3.1
0.4
3.3
population-wide life expectancy achievable through
risk factor intervention. For example, the estimated
gain in population-wide life expectancy from eliminating smoking was only 0.5 years for men and 0.4
years for women if the /3 coefficients were one
standard error less than the mean Framingham estimates (Table 2). Conversely, if the actual P3 coefficients for smoking were one standard error greater
than the mean, then eliminating smoking would be
projected to extend male population-wide life expectancy by 1.2 years and female population-wide life
expectancy by 0.8 years.
Gains for Individuals at Risk
Baseline analysis. Gains to at-risk 35-year-old individuals from single risk factor modifications ranged
TABLE 2. Sensitivity Analyses of Population-Wide Gains for
35-Year-Old Persons
Range of
Intervention
Reduce cholesterol level
To 240 mg/dl if >240 mg/dl
To 200 mg/dl if >200 mg/dl
By 10 mg/dl
By 20 mg/dl
Reduce number of cigarettes smoked
By 50%
Eliminate smoking
Reduce diastolic blood pressure
To 88 mm Hg if >88 mm Hg
Reduce weight
To ideal if >ideal
LE, life expectancy.
population-wide
gain in LE (yr)
Male
Female
0.2-0.3
0.5-0.8
0.2-0.3
0.3-0.5
0.2-0.8
0.4-1.4
0.1-0.3
0.1-0.6
0.3-0.6
0.5-1.2
0.2-0.5
0.4-0.8
1.0-1.2
0.3-0.6
0.4-0.8
0.3-0.4
Tsevat et al Gains in Life Expectancy
TABLE 3. Gains in Life Expectancy for 35-Year-Old Individuals
at Risk
Intervention
Reduce cholesterol level
To 200 mg/dl if 200-239 mg/dl
To 200 mg/dl if 240-299 mg/dl
To 200 mg/dl if >300 mg/dl
Reduce number of cigarettes smoked
By 50%
Eliminate smoking
Reduce diastolic blood pressure
To 88 mm Hg if 90-94 mm Hg
To 88 mm Hg if 95-104 mm Hg
To 88 mm Hg if >105 mm Hg
Reduce weight
To ideal if <30% over ideal
To ideal if .30% over ideal
Gain in LE (yr)
Female
Male
0.5
1.7
4.2
0.4
1.5
6.3
1.2
2.3
1.5
2.8
1.1
2.3
5.3
0.9
1.7
5.7
0.7
1.7
0.5
1.1
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
LE, life expectancy.
from 0.5 to 5.3 years for men and from 0.4 to 6.3 years
for women (Table 3). Reducing a serum cholesterol
level exceeding 300 to 200 mg/dl or reducing a
diastolic blood pressure exceeding 105 to 88 mm Hg
yielded the greatest benefits. Gains from weight loss
were relatively larger for overweight men than for
overweight women, whereas quitting smoking was
slightly more beneficial for women.
Sensitivity analyses. Several baseline projections,
such as gains from treating borderline hypercholesterolemia, changed little in the sensitivity analyses. Conversely, estimated gains from smoking interventions,
cholesterol interventions in those with the highest
serum cholesterol levels, and blood pressure interventions in those with the highest diastolic blood presTABLE 4. Sensitivity Analyses of Gains in Life Expectancy for
35-Year-Old Individuals at Risk
Range of gain in LE for
individuals at risk (yr)
Female
Male
Intervention
Reduce cholesterol level
0.4-0.6
0.2-0.8
To 200 mg/dl if 200-239 mg/dl
1.3-2.0
0.7-2.6
To 200 mg/dl if 240-299 mg/dl
3.4-5.1
2.7-10.5
To 200 mg/dl if .300 mg/dl
Reduce number of cigarettes smoked
0.9-1.8
0.7-1.7
By 50%
1.8-3.3
1.3-3.4
Eliminate smoking
Reduce diastolic blood pressure
1.0-1.2
0.6-1.2
To 88 mm Hg if 90-94 mm Hg
1.1-2.3
2.0-2.5
To 88 mm Hg if 95-104 mm Hg
3.6-7.7
4.6-5.7
To 88 mm Hg if .105 mm Hg
Reduce weight
0.4-0.6
0.4-0.9
To ideal if <30% over ideal
1.1-2.3
0.9-1.4
To ideal if .30% over ideal
LE, life expectancy.
1199
sures changed markedly. For example, varying the P
coefficients from one standard error less than the
mean to one greater than the mean created a range of
projected gains for men who quit smoking of 1.3-3.4
years (baseline estimate, 2.3 years) (Table 4).
Discussion
CHD is the leading cause of death in the United
States. It may be surprising, then, to find that even
the strictest risk factor modifications-eliminating
smoking or reducing body weight, diastolic blood
pressure, or serum cholesterol to ideal levels -would
yield apparently modest gains in population-wide life
expectancy. It is important to realize, however, that
in this model, the estimated gains in population-wide
life expectancy apply across the entire population,
not just to affected individuals. Gains to at-risk
individuals can be much more substantial and are
directly proportional to the degree that their level
exceeds the target.
Although the gains in life expectancy from many of
the simulations are modest, risk factor modification
can also reduce morbidity. A number of studies56,22
have shown striking reductions in ischemic heart disease events from risk factor modification. Although
not the focus of this analysis, a delay or avoidance of
coronary events by risk factor reduction10'2 would
improve quality of life and reduce medical care expenses independent of any effects on longevity. Intervention may indeed be cost-effective: For example,
although national expenditures for treating patients
with hypertension would be high, the aggregate net
cost per additional quality-adjusted year of life can be
as low as several thousand dollars, depending on age
and pretreatment diastolic blood pressure.1"'23
Comparison With Prior Studies
At least four previous analyses have estimated the
impact of risk factors or CHD per se on life expectancy. The Centers for Disease Control recently
calculated that smoking exacts a toll of 0.015 years
from the average population-wide life expectancy.24
Taylor and colleagues,25 who performed an analogous analysis of gains in life expectancy for affected
individuals from risk factor modification, found that
cholesterol reduction could generally achieve a gain
in life expectancy of weeks to months for people with
hypercholesterolemia. Persons at highest risk, defined as having a systolic blood pressure, cigarette
smoking habit, and total serum cholesterol level each
at the 90th percentile and a high density lipoprotein
cholesterol level at the 10th percentile for their age
and gender, were estimated to live 2-11 months
longer (depending on age and gender) by reducing
their serum cholesterol levels by 6.7% and to live
5-29 months longer by reducing it by 20%. By
comparison, they projected gains of 23-70 months
from quitting smoking and of 19-34 months from
reducing systolic blood pressure by 14.3%. Taylor
and coworkers25 did not consider gains from weight
reduction.
1200
Circulation Vol 83, No 4 April 1991
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Cohen and Lee26 estimated that heart disease
shortens male life expectancy by 6.3 years and female
life expectancy by 5.4 years. They also estimated that
smoking costs the male smoker 5.9-6.2 years and the
female smoker 1.2-2.2 years of life expectancy. Being
30% overweight was projected to shorten one's life by
3.6 years, and being 20% overweight was projected to
shorten one's life by 2.5 years. Finally, Tsai and
colleagues27 projected that eliminating all cardiovascular disease would prolong the life expectancy of
35-year-old individuals by 12.9 years.
Certain projections from our model are quite close
to previous projections. For example, our projected
gains from control of hypercholesterolemia and hypertension agree with those of Taylor and coworkers.25 Our projected gains for female smokers are in
the same range as Cohen and Lee's26 estimates.
Meanwhile, there are discrepancies with other projections. Cohen and Lee26 and Tsai and colleagues27
project larger gains from the elimination of CHD
than do we. The differences may be explained in part
by the fact that their projections were based on data
from an era when CHD was more prevalent and
treatments were fewer; also, Tsai and coauthors27
considered gains from eliminating all cardiovascular
disease, not just CHD. In contrast, the Centers for
Disease Control24 projected smaller losses in population-wide life expectancy from smoking than does
the Coronary Heart Disease Policy Model. The difference between their estimate and ours may emanate from their consideration of the entire population, children as well as adults.
Limitations
Like any forecasting tool, our model has certain
limitations. First, for simplicity, the Coronary Heart
Disease Policy Model assumes that risk factors are
distributed independently of each other; that is, they
are conditional only on age and gender. Furthermore, the model assumes that intervening on one risk
factor does not affect other risk factors. In actuality,
reduction in weight could also improve serum cholesterol level and diastolic blood pressure, whereas
reductions in smoking may lead to weight gain. Thus,
the assumption of independence of risk factors could
underestimate or overestimate the impact of any
single risk factor intervention.
A major problem of such an analysis is that the
coefficients of risk are derived from population-based
studies. Because of misclassification of both exposure
status and to a lesser extent outcome status, the
coefficients likely underestimate the true relation
between risk factors and disease.28 Underestimating
relative risk results in a corresponding underestimation of the benefits of intervention. The sensitivity
analyses demonstrate the effect of varying the relative risk on the projected gain; some projections are
affected to a greater degree than others (Tables 2
and 4)
A third limitation is that we applied cross-sectional
data longitudinally. That is, in the model, the cohort
of 35-year-old persons has CHD incidence rates
unique to its 10-year age range. When the cohort
turns 45 years, it adopts new CHD incidence rates,
but these are based on patients who would have been
45 years old in 1990. Of course, one cannot know the
true incidence rates until our starting cohort itself
turns 45 years in the year 2000.
Our model also assumes an immediate benefit
from risk factor intervention, whereas in reality,
there may be a short delay in assuming normal risk.
For example, the risk of CHD attributable to smoking declines by 50% 1 year after quitting,29 and it
takes perhaps 3 years for a reduction in cholesterol
level to yield its full benefit.5,25,30-32 Yet in risk factor
intervention modeling, it is advantageous to emphasize the current risk factor profile over the historical
profile.22 Moreover, we tested the effect of a lag
period in a complementary analysis of hypercholesterolemia: After 25 years, the results were nearly
identical to the analysis having no lag period.10
The accuracy of the Coronary Heart Disease Policy
Model's projections is partly dependent on the accuracy of US vital statistics data. Those data indicate
that mortality from ischemic heart disease (ICD
codes 410-414)33 has declined since 1980, yet much
of this decline is offset by rises in deaths from cardiac
arrest and congestive heart failure.34 Thus, changes
in coding may account for some of the apparent
decline in ischemic heart disease mortality.
In addition, recent data from the Health Interview
Study indicate that the prevalence of ischemic heart
disease has increased by more than 10% in the last
decade.35 This trend, in an era of improving coronary
risk factors, is probably explained by more aggressive
and earlier diagnosis of CHD and a prolonged expectancy of persons with CHD as opposed to an
increase in atherosclerosis. In future updates of the
model, we hope to capture such changing risk factor
distributions and prevalences, along with the projected impact of new therapies such as thrombolysis
and percutaneous transluminal coronary angioplasty.
Conclusion
Planning agencies and payors may find projected
population-wide gains useful as a bench mark for
what may be worth targeting on a national level.
Previous analyses found that eliminating cancer -a
veritable pie in the sky- could extend average
population-wide life expectancy at birth by 2.3-2.7
years.26,36 Our analysis shows that eliminating
CHD, an equally impossible task, would prolong the
life of the average 35-year-old by a little over 3
years. The single risk factor interventions considered would prolong population-wide life expectancy
by 0.2-1.1 years.
Health care providers and patients may be more
interested in potential gains for the individual at risk.
Here, gains can be more substantial and are a function
of the preintervention and postintervention risk factor
level. It is up to the individual and to society to weigh
the potential gains against the sacrifices.
Tsevat et al Gains in Life Expectancy
References
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
1. Use of drugs to cut cholesterol debated. Intem Med News
1989;22:43
2. Rice DP, Hodgson TA, Kopstein AN: The economic cost of
illness: A replication and update. Health Care Financ Rev
1985;7:61-80
3. US Department of Health and Human Services: The Framingham Study: An Epidemiological Investigation of Cardiovascular
Disease. Some Risk Factors Related to the Annual Incidence of
Cardiovascular Disease and Death Using Pooled Repeated Biennial Measurements: Framingham Heart Study, 30-Year Followup. Section 34, NIH publication No. 87-2703. Bethesda, Md,
National Heart, Lung, and Blood Institute, 1987
4. Multiple Risk Factor Intervention Trial Research Group:
Coronary heart disease death, non-fatal acute myocardial
infarction and other clinical outcomes in the Multiple Risk
Factor Intervention Trial. Am J Cardiol 1986;58:1-13
5. The Lipid Research Clinics Program: The Lipid Research
Clinics Coronary Primary Prevention Trial results: I. Reduction in incidence of coronary heart disease. JAMA4 1984;251:
351-364
6. Frick MH, Elo 0, Haapa K, Heinonen OP, Heinsalmi P, Helo
P, Huttunen JK, Kaitaniemi P, Koskinen P, Manninen V, et al:
Helsinki Heart Study: Primary-prevention trial with gemfibrozil in middle-aged men with dyslipidemia. N Engl J Med
1987;317:1237-1245
7. Office on Smoking and Health: The Health Consequences of
Smoking: Cancer-A Report of the Surgeon GeneraL DHHS
publication No. (PHS) 82-50179. Rockville, Md, US Department of Health and Human Services, Public Health Service,
1982
8. Perrone J: AMA declares war on cholesterol. Am Med News
December 16, 1988; 9
9. Weinstein MC, Coxson PG, Williams LW, Pass TM, Stason
WB, Goldman L: Forecasting coronary heart disease incidence, mortality, and cost: The Coronary Heart Disease Policy
Model. Am J Public Health 1987;77:1417-1426
10. Goldman L, Weinstein MC, Williams LW: Relative impact of
targeted versus population-wide cholesterol interventions on
the incidence of coronary heart disease: Projections of the
Coronary Heart Disease Policy Model. Circulation 1989;80:
254-260
11. Edelson JT, Weinstein MC, Tosteson ANA, Williams LW,
Lee TH, Goldman L: Long-term cost-effectiveness of various
initial monotherapies for mild to moderate hypertension.
JAMA 1990;263:408-413
12. Tosteson ANA, Weinstein MC, Williams LW, Goldman L:
Long-term impact of smoking cessation on the incidence of
coronary heart disease: Projections of the Coronary Heart
Disease Policy Model. Am J Public Health 1990;80:1481-1486
13. U.S. Bureau of the Census: Estimates of the Civilian Population
of the United States by Age, Sex, and Race: 1980-83. Current
Population Reports, Population Estimates and Projections,
series P-25, number 949. Washington, DC, Government Printing Office, 1984
14. National Center for Health Statistics: Unpublished data from
the Health Interview Survey, 1979
15. National Center for Health Statistics: Unpublished data from
the Second Health and Nutrition Examination Survey,
1976-1980
16. US Bureau of the Census: Preliminary Estimates of the Population of the United States by Age, Sex, and Race, 1970-1981.
Current Population Reports, Population Estimates and Projections, series P-25, number 922. Washington, DC, Government Printing Office, 1982
17. U.S. Bureau of the Census: Projections of the Population of the
United States by Age, Sex, and Race: 1982-2050. Current
1201
Population Reports, Population Estimates and Projections,
series P-25, number 922. Washington, DC, Government Printing Office, 1982
18. U.S. Department of Health, Education, and Welfare: The
Framingham Study: An Epidemiological Investigation of Cardiovascular Disease: Some Characteristics Related to the Incidence
of Cardiovascular Disease and Death, 18-Year Follow-up. Section 30. Bethesda, Md, Public Health Service, 1973
19. Pell S, Fayerweather WE: Trends in the incidence of myocardial infarction and in associated mortality and morbidity in a
large employed population. NEngl JMed 1985;312:1005-1011
20. National Center for Health Statistics: Vital Statistics of the
United States, 1980, Volume II, Mortality, Part A. DHHS
publication No. (PHS) 85-1101. Public Health Service, Washington, DC, US Government Printing Office, 1985
21. National Center for Health Statistics: Vital Statistics of the
United States, 1980, Volume II, Mortality, Part B. DHHS
publication No. (PHS) 85-1102. Public Health Service, Washington, DC, US Government Printing Office, 1985
22. Fries JF, Green LW, Levine S: Health promotion and the
compression of morbidity. Lancet 1989;1:481-483
23. Weinstein MC, Stason WB: Hypertension: A Policy Perspective.
Cambridge, Mass, Harvard University Press, 1976
24. State-specific estimates of smoking-attributable mortality and
years of potential life lost-United States, 1985. JAMA 1989;
261:23-25
25. Taylor WC, Pass TM, Shepard DS, Komaroff AL: Cholesterol
reduction and life expectancy: A model incorporating multiple
risk factors. Ann Intern Med 1987;106:605-614
26. Cohen BL, Lee IS: A catalog of risks. Health Physics 1979;36:
707-722
27. Tsai SP, Lee ES, Hardy RJ: The effect of a reduction in
leading causes of death: Potential gains in life expectancy. Am
J Public Health 1978;68:966-971
28. Copeland KT, Checkoway H, McMichael AJ, Holbrook RH:
Bias due to misclassification in the estimation of relative risk.
Am J Epidemiol 1977;105:488-495
29. Fielding JE: Smoking: Health effects and control (first of two
parts). N Engl J Med 1985;313:491-498
30. Oliver MF, Heady JA, Morris JN, Cooper J: A co-operative
trial in the primary prevention of ischaemic heart disease using
clofibrate: Report from the Committee of Principal Investigators. Br Heart J 1978;40:1069-1118
31. Dayton S, Pearce ML, Hashimoto S, Dixon WJ, Tomiyasu U:
A controlled clinical trial of a diet high in unsaturated fat in
preventing complications of atherosclerosis. Circulation 1969;
40(suppl):1-63
32. Miettinen M, Turpeinen 0, Karvonen MJ, Elosuo R,
Paavilainen E: Effect of cholesterol-lowering diet on mortality
from coronary heart-disease and other causes: A twelve-year
clinical trial in men and women. Lancet 1972;2:835-838
33. World Health Organization: Manual of the International Statistical Classification of Diseases, Injuries and Causes of Death:
Based on Recommendations of the Ninth Revision Conference,
1975. Geneva, Switzerland, World Health Organization, 1977
34. National Center for Health Statistics: Vital Statistics of the
United States, 1979-1981, Volume II, Mortality, Part A. Washington, DC, Government Printing Office, 1981
35. National Center for Health Statistics: National Health Interview Survey. Vital and Health Statistics, Series 10, Number 164.
DHHS publication No. (PHS) 87-1592. Hyattsville, Md, October 1987
36. Keyfitz N: What difference would it make if cancer were
eradicated? An examination of the Taeuber paradox. Demography 1977;14:411-418
KEY WORDS * coronary heart disease * risk factor *
expectancy * intervention
life
Expected gains in life expectancy from various coronary heart disease risk factor
modifications.
J Tsevat, M C Weinstein, L W Williams, A N Tosteson and L Goldman
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Circulation. 1991;83:1194-1201
doi: 10.1161/01.CIR.83.4.1194
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 1991 American Heart Association, Inc. All rights reserved.
Print ISSN: 0009-7322. Online ISSN: 1524-4539
The online version of this article, along with updated information and services, is located on
the World Wide Web at:
http://circ.ahajournals.org/content/83/4/1194
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally
published in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the
Editorial Office. Once the online version of the published article for which permission is being requested is
located, click Request Permissions in the middle column of the Web page under Services. Further
information about this process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Circulation is online at:
http://circ.ahajournals.org//subscriptions/