Framingham Risk Score and Prediction of
Lifetime Risk for Coronary Heart Disease
Donald M. Lloyd-Jones, MD, ScM, Peter W.F. Wilson, MD, Martin G. Larson, ScD,
Alexa Beiser, PhD, Eric P. Leip, MS, Ralph B. D’Agostino, PhD, and Daniel Levy, MD
We investigated whether the Framingham risk score,
which was designed to estimate the 10-year risk of
coronary heart disease (CHD), differentiates lifetime risk
for CHD. All subjects in the Framingham Heart Study
examined from 1971 to 1996 who were free of CHD
were included. Subjects were stratified into age- and
gender-specific tertiles of Framingham risk score, and
lifetime risk for CHD was estimated. We followed 2,716
men and 3,500 women; 939 developed CHD and 1,363
died free of CHD. At age 40 years, in risk score tertiles
1, 2, and 3, respectively, the lifetime risks for CHD were
38.4%, 41.7%, and 50.7% for men and 12.2%, 25.4%,
and 33.2% for women. At age 80 years, risks were
16.4%, 17.4%, and 38.8% for men and 12.8%, 22.4%,
and 27.4% for women. The Framingham risk score strat-
ified lifetime risk well for women at all ages. It per-
T he Framingham Heart Study has published multi-
variate models for estimation of the 10-year abso-
lute risk of developing coronary heart disease
(CHD).1,2 Framingham multivariate equations and
risk scores are applicable to a wide range of popula-
tions, with only minor adjustments for baseline inci-
dence,3 and they have formed the basis for a number
of recent clinical guidelines that address appropriate
selection of patients for drug therapy in primary pre-
vention of CHD.2,4,5 However, some experts have
expressed reluctance to place exclusive emphasis on
short-term risk estimates to determine all treatment
decisions.2,6,7 Subjects with low or intermediate 10-
year risk for CHD may still be at high risk in the
longer term because any single risk factor can produce
cumulative damage and high risk if left untreated for
many years.6 – 8 Our group9 and others10,11 have pro-
moted the concept of lifetime risk estimation for
From the Department of Preventive Medicine and Division of Cardiol-
ogy, Feinberg School of Medicine, Northwestern University, Chicago,
Illinois; the National Heart, Lung, and Blood Institute’s Framingham
Heart Study, Framingham, Massachusetts; the Department of Epidemi-
ology and Preventive Medicine, Boston University School of Medicine,
the Statistics and Consulting Unit, Boston University, and the Depart-
ment of Epidemiology and Biostatistics, Boston University School of
Public Health, Boston, Massachusetts. Dr. Lloyd-Jones is supported by
grant 1 K23 HL04253 from the National Institutes of Health, Be-
thesda, Maryland. The Framingham Heart Study is supported by
contract N01-HC-25195 from the National Institutes of Health/Na-
tional Heart, Lung, and Blood Institute, Bethesda, Maryland. Manu-
script received October 16, 2003; revised manuscript received and
accepted March 19, 2004.
Address for reprints: Donald M. Lloyd-Jones, MD, ScM, Depart-
ment of Preventive Medicine, Feinberg School of Medicine, North-
western University, 680 N. Lake Shore Drive, Suite 1120, Chicago,
Illinois. E-mail: dlj@northwestern.edu.
20 ©2004 by Excerpta Medica, Inc. All rights reserved.
The American Journal of Cardiology Vol. 94 July 1, 2004
formed less well in younger men but improved at older
ages as remaining life expectancy approached 10
years. Lifetime risks contrasted sharply with shorter term
risks: at age 40 years, the 10-year risks of CHD in
tertiles 1, 2, and 3, respectively, were 0%, 2.2%, and
11.6% for men and 0%, 0.7%, and 2.3% for women. The
Framingham 10-year CHD risk prediction model dis-
criminated short-term risk well for men and women.
However, it may not identify subjects with low short-
term but high lifetime risk for CHD, likely due to changes
in risk factor status over time. Further work is needed to
generate multivariate risk models that can reliably pre-
dict lifetime risk for CHD. 䊚2004 by Excerpta Medica,
Inc.
(Am J Cardiol 2004;94:20 –24)
CHD. Lifetime risk estimates, which account for the
risk of the disease of interest and the competing risk of
death from other causes, provide a simple conceptual
basis for estimating the absolute risk of developing
disease over the remaining lifespan. As such, they
may be more easily understood than relative risks by
the lay public and can be used by clinicians and policy
makers to predict the population burden of disease and
to compare lifetime risks between diseases. In the
present analysis, we investigated whether Framing-
ham 10-year risk equations could reliably stratify life-
time risk for CHD in men and women free of CHD at
selected ages.
METHODS
Subjects: The Framingham Heart Study was estab-
lished in 1948, when 5,209 residents, 28 to 62 years
old, of Framingham, Massachusetts, were enrolled in
a prospective epidemiologic cohort study. In 1971, an
additional 5,124 subjects (offspring of original cohort
subjects and spouses of offspring) were enrolled in the
Framingham Offspring Study. Study design and entry
criteria for the 2 cohorts have been detailed else-
where.12,13 For the present analysis, to reflect more
contemporary experience, all subjects who were ex-
amined between 1971 and 1996 were eligible pro-
vided that they were examined at least once between
the ages of 40 and 94 years and they had follow-up
after the earliest eligible examination. Subjects who
were free of CHD at their initial examination during
the study period were included.
Case ascertainment: At each examination, interim
cardiovascular events were ascertained. CHD events
included angina pectoris, coronary insufficiency, myo-
0002-9149/04/$–see front matter
doi:10.1016/j.amjcard.2004.03.023
 TABLE 1 Subjects and Framingham Heart Study Risk Score by Age- and Gender-Specific Tertiles of Risk Score
Men* Women*

Index Age (yrs) Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3

40 n ⫽ 590 n ⫽ 621 n ⫽ 548 n ⫽ 794 n ⫽ 691 n ⫽ 677

Framingham risk score† 1.6 ⫾ 1.7 4.0 ⫾ 0.9 6.2 ⫾ 1.0 ⫺5.1 ⫾ 1.1 ⫺2.3 ⫾ 0.8 1.6 ⫾ 2.3
50 n ⫽ 884 n ⫽ 780 n ⫽ 653 n ⫽ 1,032 n ⫽ 959 n ⫽ 962
Framingham risk score 4.3 ⫾ 1.5 6.7 ⫾ 0.5 8.7 ⫾ 0.9 2.8 ⫾ 1.6 5.7 ⫾ 0.9 9.2 ⫾ 2.0
60 n ⫽ 660 n ⫽ 664 n ⫽ 636 n ⫽ 925 n ⫽ 879 n ⫽ 828
Framingham risk score 6.3 ⫾ 1.2 8.5 ⫾ 0.5 10.7 ⫾ 1.1 6.7 ⫾ 1.9 9.9 ⫾ 1.0 13.3 ⫾ 2.0
70 n ⫽ 408 n ⫽ 415 n ⫽ 367 n ⫽ 647 n ⫽ 628 n ⫽ 547
Framingham risk score 8.4 ⫾ 1.1 10.7 ⫾ 0.5 12.9 ⫾ 1.0 7.3 ⫾ 1.8 11.1 ⫾ 0.9 14.5 ⫾ 1.9
80 n ⫽ 129 n ⫽ 142 n ⫽ 119 n ⫽ 305 n ⫽ 284 n ⫽ 241
Framingham risk score 10.3 ⫾ 1.2 12.5 ⫾ 0.4 14.7 ⫾ 1.1 9.3 ⫾ 1.6 12.6 ⫾ 0.7 15.8 ⫾ 1.8

*Total number of subjects ⫽ 2,716 men and 3,500 women. Subjects contribute to all index ages they achieve free of CHD.
†
The possible range of risk scores is from ⫺6 in younger subjects to ⫹19 in the oldest subjects.
Mean ⫾ SD.

TABLE 2 Lifetime Risk for Coronary Heart Disease by Tertile of Framingham Risk Score at Specific Index Ages
Men Women
Index Age
(yrs) Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3

40* 38.4% (34.7–42.2) 41.7% (38.0–45.3) 50.7% (46.8–54.5) 12.2% (9.2–15.2) 25.4% (21.3–29.5) 33.2% (29.8–36.6)
50† 44.2% (40.1–48.3) 49.2% (45.6–52.7) 54.4% (51.0–57.8) 23.1% (20.0–26.2) 30.5% (27.0–34.0) 34.8% (31.9–37.8)
60† 37.1% (33.7–40.5) 42.3% (39.2–45.5) 47.9% (44.7–51.1) 19.7% (16.9–22.6) 32.2% (29.2–35.2) 36.6% (33.6–39.6)
70† 28.5% (25.1–31.9) 32.0% (28.6–35.4) 44.9% (41.2–48.5) 16.3% (13.3–19.1) 23.6% (20.7–26.5) 36.1% (32.9–39.4)
80† 16.4% (12.7–20.1) 17.4% (13.6–21.2) 38.8% (34.2–43.4) 12.8% (9.8–15.8) 22.4% (19.0–25.8) 27.4% (23.8–31.0)

Values are lifetime risk for CHD (95% confidence interval).

*Lifetime risk through age 84 years of age (limited length of follow-up).
†
Lifetime risk through 94 years of age.

cardial infarction, and death due to CHD, as described
in detail previously.9 All suspected cardiovascular
events were reviewed by a panel of 3 physicians, who
applied established criteria14 for such events. Deaths
were assigned to 1 of 6 mutually exclusive causes:
CHD, stroke, other cardiovascular disease, cancer,
other, or unknown.
Calculation of Framingham risk score: The risk score
was calculated for each subject using the risk score of
Wilson et al.1 In this algorithm, subjects receive a
point score based on categorical values of age, total
cholesterol, high-density lipoprotein cholesterol,
blood pressure, smoking, and diabetes. The scoring
sheet is available in the study by Wilson et al1
and online at www.nhlbi.nih.gov/about/framingham/
riskabs.htm. We calculated the risk score at each ex-
amination a subject attended and assigned a risk score
based on the mean of the subject’s calculated risk
scores in the 5 years before each index age of 40, 50,
60, 70, and 80 years. Men and women were stratified
separately into tertiles of risk score for each index age.
Statistical analysis: All statistical analyses were per-
formed with SAS statistical software.15 For each index
age, we stratified subjects according to tertile of risk
score. Lifetime risk for CHD was then calculated
separately for men and women in each tertile at each
index age. We also calculated lifetime risk for hard
CHD events, ignoring angina pectoris as a first event.
Lifetime risk was calculated by using a modified
technique of survival analysis, as described previous-
ly.9,16,17 Because few subjects survived past age 94
years, lifetime risk estimates were calculated only
through age 94 years. Each subject in the study sample
was followed from entry through 1996 until the year
of a first CHD event, the year of death, attainment of
age 95 years, or the last year in which the subject
underwent a follow-up examination. Hazards, age-
specific incidences, cumulative incidence, and sur-
vival probabilities were calculated first by Kaplan-
Meier analysis.18 Because the Kaplan-Meier cumulative
incidence does not reflect the competing risk of death
from other causes before the development of CHD, ad-
justment was made for this competing risk to yield a true
remaining lifetime risk19 of CHD.
RESULTS
Study sample: We followed 2,716 men and 3,500
women from 1971 to 1996. During follow-up, 939
subjects developed CHD and 1,363 died free of CHD.
Table 1 lists the number of subjects contributing data
and the mean Framingham risk scores for each age-
and gender-specific risk score tertile. There was a
stepwise increase in mean risk score with advancing
age, because advancing age confers increased risk for
CHD and because of a greater burden of CHD risk
factors with advancing age.
Lifetime risk for CHD by Framingham risk score: Ta-
ble 2 presents the lifetime risks for CHD by tertile of
risk score at each index age for men and women.
Figure 1 displays the cumulative risk curves and life-

CORONARY ARTERY DISEASE/FHS RISK SCORE AND LIFETIME RISK FOR CHD 21
FIGURE 1. Cumulative risk for CHD for men and women at index ages of 40, 50, 60, 70, and 80 years, by tertile of Framingham risk
score, adjusted for the competing risk of death. Lifetime risk estimates through the last age are shown at the right margin of each
graph.
time risk for CHD for each index age. For women at
all ages, the risk score appeared to discriminate risk
well, with 1.5- to 3-fold gradients in remaining life-
time risk between the highest and lowest tertiles. In
men, the risk score discriminated risk less well at
younger ages, but it performed better at older ages as
remaining life expectancy approached 10 years. In
men and women, overall lifetime risk for CHD de-
creased with advancing index age because of increas-
ing competing risk of death and decrease of suscepti-
ble patients at younger ages. Lifetime risks for hard
CHD events, excluding angina pectoris as an initial
CHD event, are presented in Table 3. Similar patterns
of risk discrimination were observed as for all CHD
events, but absolute lifetime risk for hard CHD was
slightly lower.
Ten-year versus lifetime risk for CHD: The Framing-
ham risk score stratified 10-year cumulative risk well,
even in the context of the competing risk of death free
of CHD, for men and women at all ages (Table 4). At
older ages, as the remaining life expectancy de-
creased, the 10-year and lifetime risks more closely
approximated each other. When comparing Tables 2
22 THE AMERICAN JOURNAL OF CARDIOLOGY姞 VOL. 94
and 4, it should be noted that younger subjects in the
lowest risk tertiles, who had very low 10-year risks of
CHD, still had a substantial lifetime risk for CHD. For
example, at 50 years of age, men and women in the
lowest tertile of risk score had low 10-year cumulative
risks of CHD but had remaining lifetime risks that
were 10 times higher.
In addition, at 40 and 50 years of age, no group had
a 10-year cumulative risk of ⬎20%, the threshold
value recommended for aggressive treatment in cur-
rent clinical practice guidelines.2,4 In men, only the
highest tertile at 60, 70, and 80 years of age was
⬎20% risk threshold in 10 years; in women, only the
highest tertile at 80 years of age was ⬎20% risk
threshold in 10 years (Table 5).
DISCUSSION
The Framingham risk score, which was designed to
predict 10-year risk for CHD, was very effective in
predicting the short-term cumulative risk for CHD,
even in the context of competing risk of death from
noncoronary causes. In women, the risk score was also
very effective at stratifying the remaining lifetime risk
JULY 1, 2004
 performed less well in predicting
TABLE 3 Lifetime Risk for Hard Coronary Heart Disease, Excluding Angina
Pectoris, by Tertile of Framingham Risk Score at Specific Index Ages lifetime risk for CHD in younger
subjects because of changes in risk
Men Women factor status that occurred over time.
Index Age
(yrs) Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3 Rates of diabetes and hypertension
40* 31.2% 33.6% 46.8% 9.7% 15.1% 25.9%
increase sharply with age, which
50† 40.9% 42.4% 48.4% 16.9% 22.9% 27.4% may alter the long-term risk of
60† 31.9% 37.4% 43.5% 14.8% 26.3% 30.9% younger patients in an unpredictable
70† 25.2% 27.8% 42.4% 12.3% 19.9% 31.7% fashion.
80† 16.1% 14.4% 34.2% 11.2% 19.6% 24.9% In addition, the competing risk of
*Lifetime risk through 84 years of age (limited length of follow-up). death from noncoronary causes,
†
Lifetime risk through 94 years of age. which is factored into lifetime risk
estimates, may have unpredictable
effects on lifetime risk. For example,
TABLE 4 Ten-year Cumulative Risk of Coronary Heart Disease (adjusted for short-term risk for CHD is much
competing risk of death) by Tertile of Framingham Risk Score at Index Age higher for smokers than for non-
Men Women smokers, but lifetime risk for CHD is
Index Age remarkably similar because of the
(yrs) Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3
substantially higher risk of death
40 0% 2.2% 11.6% 0% 0.7% 2.3% from competing causes and the con-
50 4.4% 11.1% 16.1% 2.2% 1.1% 8.0% sequent shorter lifespan in smokers.
60 10.2% 17.4% 22.2% 3.8% 8.2% 12.7%
70 18.4% 19.9% 27.5% 5.9% 9.8% 15.5%
The effect of smoking on lifetime
80 15.1% 15.5% 35.0% 10.6% 16.7% 25.8% risk may account in part for the
poorer long-term predictive ability of
the Framingham risk score and for
the particularly poor discrimination
TABLE 5 Time in Years to Age at Which Cumulative Lifetime Risk for Coronary among younger men, who were more
Heart Disease Was ⬎20% Threshold, by Tertile of Framingham Risk Score at
Different Index Ages likely to be smokers.
One important area of interest in
Men Women current clinical guidelines is global
Index Age
(yrs) Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3 risk stratification to identify subjects
40 29 26 16 ⬎45 42 28
at high risk for events who may merit
50 22 17 11 36 33 22 early intervention with drug therapy
60 17 11 8 ⬎35 21 15 to decrease their risk. In the Preven-
70 12 10 6 ⬎25 18 12 tion Conference V sponsored by the
80 ⬎15 ⬎15 3 ⬎15 12 6 American Heart Association6,7 and
the Third Report of the National
Cholesterol Education Program Ex-
for CHD, with 1.5- to 3-fold higher absolute risk in the pert Panel on Detection, Evaluation, and Treatment of
highest versus lowest tertile of risk score at all ages. In High Blood Cholesterol in Adults (NCEP-ATP III),2
men, the risk score discriminated lifetime risk well the panels emphasized lifetime risk estimation, not
only at older ages, likely because in older men the just short-term risk estimation. Our data suggest that
remaining lifespan approached 10 years. For younger selecting a 10-year risk for CHD of 20% as a threshold
men, there was a stepwise increase in lifetime risk for for screening and treatment of risk factors4 may be
CHD with increasing risk score, but the risk score did inappropriately high for clinical practice, given the
not stratify remaining lifetime risk well. extremely high absolute lifetime risks seen in younger
There was a striking disparity between the magni- and middle-aged subjects. The NCEP-ATP III report
tudes of short- and long-term CHD risk, even among used our data on lifetime risk for CHD according to
subjects predicted to be at low risk. Men in the lowest cholesterol levels to justify treatment for subjects at
tertile of risk score at 50 years of age experienced a low 10-year but high lifetime risk for CHD.2
10-year cumulative risk of only 1 in 25 but had a Short-term CHD risk-prediction equations based
lifetime risk of nearly 1 in 2. Similarly, women in the on Framingham data form the basis of risk stratifica-
lowest tertile at 50 years of experienced a 10-year tion for several current clinical practice guidelines,
cumulative risk of 1 in 50 but had a remaining lifetime including the recommendations of the Second Joint
risk of 1 in 4. Task Force of European and Other Societies on Cor-
We calculated the risk score at selected index ages, onary Prevention,4 the Joint British Recommendations
which did not account for subsequent changes in life- on Prevention of Coronary Heart Disease in Clinical
style, blood pressure, lipids, and diabetes status, or Practice,5 and NCEP-ATP III.2 The Framingham 10-
medications that could alter CHD risk. We previously year risk equations performed well in predicting short-
found that, at all index ages, cholesterol levels alone term risk for CHD in white and black cohorts in the
separate remaining lifetime risk for CHD very well.8 It United States. Among Japanese American men, His-
is likely that the composite Framingham risk score panic men, and Native American women, the equa-

CORONARY ARTERY DISEASE/FHS RISK SCORE AND LIFETIME RISK FOR CHD 23
tions tended to overestimate risk.3 However, after
recalibration for baseline differences in risk factor
prevalence and underlying rates of incident CHD, the
Framingham functions performed well in these ethnic
populations.3 Thus, these equations appear broadly
generalizable with minor modifications. Our data sug-
gest that new equations may be desirable to predict
lifetime risk for CHD, particularly for younger men.
These results must be considered in light of several
potential limitations. The Framingham Heart Study
cohort is composed almost exclusively of white pa-
tients, which may limit the generalizability of our
findings to other ethnic groups. Certain ethnic popu-
lations have different prevalences of CHD risk factors,
suggesting that their lifetime risks may differ from
those we observed. Hence, additional studies of the
lifetime risk for CHD in other large cohorts would be
useful. Because of their participation in periodic ex-
aminations, Framingham subjects may have been mo-
tivated to modify risk factors, which could have de-
creased lifetime risk for CHD. We may have
underestimated the intrinsic lifetime risk for CHD for
subjects with the highest levels of risk because they
may have been more likely to receive antihyperlipi-
demic, antihypertensive, or other therapies during fol-
low-up. Current methods in lifetime risk estimation
cannot account for changes in risk factor status over
time due to aging, changes in lifestyle, or medication.
1. Wilson PW, D’Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel
WB. Prediction of coronary heart disease using risk factor categories. Circulation
1998;97:1837–1847.
2. NCEP Expert Panel. Third Report of the National Cholesterol Education
Program (NCEP) expert panel on detection, evaluation, and treatment of high
blood cholesterol in adults (Adult Treatment Panel III). Final report. Circulation
2002;106:3143–3421.
3. D’Agostino RB, Grundy SM, Sullivan LM, Wilson P, for the CHD Risk
24 THE AMERICAN JOURNAL OF CARDIOLOGY姞 VOL. 94
Prediction Group. Validation of the Framingham coronary heart disease predic-
tion scores: results of a multiple ethnic groups investigation. JAMA 2001;286:
180 –187.
4. Wood D, De Backer G, Faergeman O, Graham I, Mancia G, Pyorala K.
Prevention of coronary heart disease in clinical practice: recommendations of the
Second Joint Task Force of European and Other Societies on Coronary Preven-
tion. Atherosclerosis 1998;140:199 –270.
5. Wood DA, Durrington P, McInnes G, Poulter N, Rees A, Wray R. Joint British
recommendations on prevention of coronary heart disease in clinical practice.
Heart 1998;80(suppl):S1–S29.
6. Grundy SM, Bazzarre T, Cleeman J, D’Agostino RB, Hill M, Houston-Miller
N, Kannel WB, Krauss R, Krumholz HM, Lauer RM. Prevention conference V.
Beyond secondary prevention: identifying the high-risk patient for primary pre-
vention. Medical office assessment. Circulation 2000;101:e3–e11.
7. Smith SC, Greenland P, Grundy SM. Prevention conference V. Beyond
secondary prevention: identifying the high-risk patient for primary prevention.
Executive summary. Circulation 2000;101:111–116.
8. Lloyd-Jones DM, Wilson PWF, Larson MG, Leip EP, Beiser A, D’Agostino
RB, Cleeman JI, Levy D. Lifetime risk for coronary heart disease by cholesterol
levels at selected ages. Arch Intern Med 2003;163:1966 –1972.
9. Lloyd-Jones DM, Larson MG, Beiser A, Levy D. Lifetime risk of developing
coronary heart disease. Lancet 1999;353:89 –92.
10. Cummings SR, Black DM, Rubin SM. Lifetime risks of hip, Colles’, or
vertebral fracture and coronary heart disease among white postmenopausal
women. Arch Intern Med 1989;149:2445–2448.
11. Fraser GE, Lindsted KD, Beeson WL. Effect of risk factor values on lifetime
risk of and age at first coronary event. The Adventist Health Study. Am J
Epidemiol 1995;142:746 –758.
12. Dawber TR, Kannel WB, Lyell LP. An approach to longitudinal studies in a
community: the Framingham study. Ann NY Acad Sci 1963;107:539 –556.
13. Feinleib M, Kannel WB, Garrison RJ, McNamara PM, Castelli WP. The
Framingham Offspring Study. Design and preliminary data. Prev Med 1975;4:
518 –525.
14. Abbott RD, McGee DL. The Framingham Study: an epidemiological inves-
tigation of cardiovascular disease, section 37: the probability of developing
certain cardiovascular diseases in eight years at specified values of some char-
acteristics. Bethesda, MD: National Heart, Lung and Blood Institute, 1987.
15. SAS/STAT User’s Guide, Version 8. Cary, NC: SAS Institute, Inc., 1999.
16. Seshadri S, Wolf PA, Beiser A, Au R, McNulty K, White R, D’Agostino RB.
Lifetime risk of dementia and Alzheimer’s disease: the impact of mortality on risk
estimates in the Framingham Study. Neurology 1997;49:1498 –1504.
17. Vasan RS, Beiser A, Seshadri S, Larson MG, Kannel WB, D’Agostino RB,
Levy D. Residual lifetime risk for developing hypertension in middle-aged
women and men. JAMA 2002;287:1003–1010.
18. Kaplan EL, Meier P. Nonparametric estimation from incomplete observa-
tions. J Am Stat Assoc 1958;53:457–481.
19. Gaynor JJ, Feuer EJ, Tan CC. On the use of cause-specific failure and
conditional failure probabilities. J Am Stat Assoc 1993;88:402–409.
JULY 1, 2004