American Journal of Epidemiology Vol. 165, No.

12
Copyright ª 2007 by the Johns Hopkins Bloomberg School of Public Health DOI: 10.1093/aje/kwm031
All rights reserved; printed in U.S.A. Advance Access publication April 3, 2007

Original Contribution

Cardiorespiratory Fitness as a Predictor of Nonfatal Cardiovascular Events in

Asymptomatic Women and Men

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

Xuemei Sui1, Michael J. LaMonte2, and Steven N. Blair3
1
The Cooper Institute, Dallas, TX.
2
Department of Social and Preventive Medicine, University of Buffalo, Buffalo, NY.
3
Department of Exercise Science, Arnold School of Public Health, University of South Carolina, Columbia, SC.

Received for publication August 29, 2006; accepted for publication November 20, 2006.

Prospective data relating cardiorespiratory fitness (CRF) with nonfatal cardiovascular disease (CVD) events are
limited to studies in men or studies of combined fatal and nonfatal CVD endpoints. The authors examined the
association between CRF and nonfatal CVD events in 20,728 men and 5,909 women without CVD at baseline. All
participants performed a maximal treadmill exercise test and completed a follow-up health survey in the Aerobics
Center Longitudinal Study (Dallas, Texas) between 1971 and 2004. There were 1,512 events in men and 159
events in women during an average follow-up of 10 years. Across incremental CRF groups, age- and examination
year-adjusted event rates per 10,000 person-years were 107.9, 75.2, and 50.3 in men (ptrend < 0.001) and 41.9,
27.7, and 20.8 in women (ptrend ¼ 0.002). After further adjustment for smoking, alcohol intake, family history of
CVD, and abnormal exercise electrocardiogram responses, hazard ratios were 1.00 (referent), 0.82 (95% confi-
dence interval (CI): 0.72, 0.94), and 0.61 (95% CI: 0.53, 0.71) in men, ptrend < 0.001, and were 1.00 (referent), 0.74
(95% CI: 0.49, 1.13), and 0.63 (95% CI: 0.40, 0.98) in women, ptrend ¼ 0.05. After adjustment for other CVD
predictors, the association remained significant in men but not in women.

cardiovascular diseases; cerebrovascular accident; exercise; primary prevention; women

Abbreviations: ACLS, Aerobics Center Longitudinal Study; CI, confidence interval; CRF, cardiorespiratory fitness; CVD,
cardiovascular disease; SD, standard deviation.

Cardiovascular disease (CVD) continues to exact a large CRF is inversely associated with CVD mortality in adults
economic and public health toll in the United States, ac- (12–16). Few prospective studies have reported on CRF and
counting for nearly 1 million deaths and 6 million hospital- nonfatal CVD risk, and those that have are limited to stud-
izations in 2003 (1). Physical inactivity is a major modifiable ies in men or to combined nonfatal/fatal endpoints (15, 17–
CVD risk factor (2) that is associated with increased risk of 21). Although it may be intuitive to expect that CRF would
fatal and nonfatal CVD events in women and men (3–10). confer protection against nonfatal CVD events in women
Cardiorespiratory fitness (CRF) is an objective, reproduc- and men as is seen for fatal CVD, this conclusion can not
ible, physiologic measure that reflects the functional influ- accurately be drawn from studies of combined nonfatal/
ences of physical activity habits, genetics, and disease status. fatal events or studies only in men. We examined the pro-
Because CRF is less prone to misclassification, it may better spective association between CRF and nonfatal CVD in
reflect the adverse health consequences of a sedentary life- women and men in the Aerobics Center Longitudinal Study
style than do self-reported physical activity exposures (11). (ACLS).

Correspondence to Dr. Xuemei Sui, Department of Exercise Science, Arnold School of Public Health, 921 Assembly Street, Columbia, SC 29208
(e-mail: msui@gwm.sc.edu).

1413 Am J Epidemiol 2007;165:1413–1423

1414 Sui et al.

MATERIALS AND METHODS electrocardiogram responses were broadly defined as

rhythm and conduction disturbances and ischemic ST-T
Study population wave abnormalities as described in detail elsewhere (28).
Participants were 20,728 men and 5,909 women aged We have found 90 percent agreement between the electro-
18
83 years who completed a baseline examination at the cardiogram interpretation recorded in our database and that
Cooper Clinic (Dallas, Texas) during 1971–2001. At base- of a group of three physicians who read a random sample of
line, all participants were free of known CVD, had normal 357 records of patients (28).
resting electrocardiograms, and were able to complete an Assessment of outcomes
exercise stress test to at least 85 percent of their age-
predicted maximal heart rate. All participants responded CVD events were ascertained from responses to mail-
to at least one mail-back health survey during follow-up. back health surveys in 1982, 1999, and 2004. The aggregate
Most participants were Caucasian and from middle and survey response rate across all survey periods in the ACLS

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

upper socioeconomic strata. Participants provided written is approximately 65 percent. Nonresponse bias is a concern
consent to participate in the follow-up study. in epidemiologic surveillance, and this issue has been in-
vestigated in the ACLS (29). Baseline health histories and
clinical measures were similar between responders and non-
Baseline examination responders and between early and late responders (29). Total
mortality rates also have been similar between responders
The physician’s examination and clinical measurements
and nonresponders (unpublished data). CVD endpoints were
were completed after an overnight fast of at least 12 hours
defined as diagnosis by a physician of myocardial infarction,
(12, 13). Body mass index (weight (kg)/height (m)2) was
stroke, or a coronary revascularization procedure (coronary
computed from measured height and weight. After a brief
artery bypass graft or percutaneous coronary intervention).
period of quiet sitting, blood pressure was recorded as the
In participants reporting multiple events, the first event was
first and fifth Korotkof sounds by use of auscultation meth-
used for analysis. The primary outcome was all CVD events.
ods (22). Serum samples were analyzed for lipids and glu-
Secondary outcomes were coronary heart disease events
cose with standardized automated bioassays. The presence
(myocardial infarction, coronary revascularization) and
of hypertension, diabetes, and dyslipidemia was based on
myocardial infarction and stroke as separate endpoints. In
a history of physician diagnosis or measured phenotypes
a random sample of these endpoints (n ¼ 50 each), we
that met clinical thresholds for each condition. Information
applied a standard definition for defining and adjudicating
on smoking habits (current smoker or not), alcohol intake
myocardial infarction, revascularization, and stroke (30,
(drinks per week), and physical activity habits (sedentary or
31). The percentage of agreement between reported events
active) was obtained from a questionnaire. Sedentary was
and participants’ medical records was 88 percent, 100 per-
defined as reporting no leisure-time physical activity in the 3
cent, and 89 percent for myocardial infarction, revasculari-
months before the examination.
zation, and stroke, respectively.
CRF was quantified as the duration of a symptom-limited
maximal treadmill exercise test using a modified Balke pro- Statistical analysis
tocol (12, 23). Exercise duration on this protocol is highly
correlated with measured maximal oxygen uptake (r > Follow-up time among noncases was computed as the
0.90) (24, 25). The test endpoint was volitional exhaustion difference between the date of the baseline examination
or termination by the supervising physician. The mean per- and the date of the last returned survey where the partici-
centage of age-predicted maximal heart rate achieved dur- pant reported being free of CVD. Follow-up time among
ing exercise was 100.3 (standard deviation (SD): 7.0) in cases was computed as the difference between the baseline
women and was 101.2 (SD: 7.0) in men. Maximal metabolic examination date and the reported date of the CVD event.
equivalents (METs) (1 MET ¼ 3.5 ml of oxygen uptake per If a diagnosis date was not provided, we used the midpoint
kilogram/minute) were estimated from the final treadmill between the date of the case-finding survey and either the
speed and grade (26). In previous ACLS reports that have baseline examination date or the date of the last returned
shown low CRF to be an independent predictor of mortality survey where the participant reported being free of CVD.
and nonfatal disease (12, 13, 27), we have defined low, The mean follow-up interval in years was 10.4 (SD: 8.1)
moderate, and high CRF exposures according to the lowest for men and 10.2 (SD: 7.8) for women. Cox proportional
20 percent, the middle 40 percent, and the upper 40 percent, hazards regression analysis was used to estimate hazard
respectively, of the age- and sex-specific distribution of tread- ratios and 95 percent confidence intervals of CVD events
mill duration in the overall ACLS population (table 1). To according to exposure categories. Multivariable analyses
maintain consistency in our study methods and because a included six covariables: age (years), examination year,
widely accepted clinical categorization of CRF does not current smoker (yes/no), alcohol intake (5 drinks/week
exist, we used the above approach. CRF by this definition or not), abnormal exercise electrocardiogram responses
was positively associated with reported physical activity (present or not), and family history of CVD (present or
status. The percentages of participants classified as being not). We conducted additional analyses that further ad-
physically active in the low, moderate, and high CRF groups justed for baseline differences in the following four factors
were 28.8, 54.9, and 86.8 in men and were 33.5, 59.5, and that may be intermediate in the causal pathway between
86.7 in women (ptrend < 0.001, each). Abnormal exercise CRF and CVD: body mass index (<25 vs. 25 kg/m2),

Am J Epidemiol 2007;165:1413–1423
 Cardiorespiratory Fitness and Cardiovascular Events 1415

TABLE 1. Age- and sex-specific maximal treadmill exercise duration and estimated
metabolic equivalent levels of cardiorespiratory fitness, Aerobics Center Longitudinal
Study, Dallas, Texas, 1971–2004*

Age and Men Women

quintile Duration (minutes) METsy Duration (minutes) METs

20–39 years
1 <15.0 <10.4 <10.3 <8.2
2 15.0–<18.0 10.4–<11.7 10.3–<13.0 8.2–<9.4
3 18.0–<20.3 11.7–<13.1 13.0–<15.0 9.4–<10.4
4 20.3–23.6 13.1–14.4 15.0–18.0 10.4–11.7

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

5 >23.6 >14.4 >18.0 >11.7
40–49 years
1 <13.5 <9.9 <8.9 <7.6
2 13.5–<16.1 9.9–<10.8 8.9–<11.0 7.6–<8.5
3 16.1–<19.0 10.8–<12.2 11.0–<13.0 8.5–<9.4
4 19.0–22.0 12.2–13.5 13.0–16.0 9.4–10.8
5 >22.0 >13.5 >16.0 >10.8
50–59 years
1 <11.0 <8.5 <7.0 <6.7
2 11.0–<13.3 8.5–<9.9 7.0–<9.0 6.7–<7.6
3 13.3–<16.0 9.9–<10.8 9.0–<10.7 7.6–<8.5
4 16.0–19.2 10.8–12.3 10.7–13.2 8.5–9.6
5 >19.2 >12.3 >13.2 >9.6
60 years
1 <7.8 <7.2 <5.5 <5.8
2 7.8–<10.5 7.2–<8.5 5.5–<7.0 5.8–<6.7
3 10.5–<13.1 8.5–<9.5 7.0–<9.0 6.7–<7.6
4 13.1–16.4 9.5–10.8 9.0–11.3 7.6–8.6
5 >16.4 >10.8 >11.3 >8.6

* Treadmill exercise testing was performed by use of a modified Balke-Ware protocol as

described in Materials and Methods. Low fitness: quintile 1; moderate fitness: quintiles 2 and 3;
high fitness: quintiles 4 and 5. Among participants in the current analysis, the distribution of low,
moderate, and high fitness by the above definition was 19%, 40%, and 41% in men and 15%,
35%, and 50% in women.
y METs, metabolic equivalents; 1 MET ¼ 3.5 ml of oxygen uptake per kilogram/minute.

hypertension, diabetes, and dyslipidemia (present or not for RESULTS

each), although authors debate whether or not an exposure-
outcome relation should be adjusted for biologic intermedi- There were 1,512 CVD events (489 myocardial infarc-
ates (32). To reduce the influence of ascertainment bias due tions, 290 strokes, 733 revascularizations) during 215,984
to variable survey response patterns, we stratified analyses man-years of exposure and 159 CVD events (53 myocardial
on survey year by use of the STRATA statement in Proc infarctions, 62 strokes, 44 revascularizations) during 60,158
PHREG (SAS, version 9.1, statistical software; SAS Insti- woman-years of exposure. Compared with noncases, indi-
tute, Inc., Cary, North Carolina). Tests of linear trends viduals who developed CVD were older, had lower CRF,
across exposure categories were computed with ordinal and had higher prevalence of sedentary habits and other
scoring. The proportional hazards assumption was exam- major CVD risk factors (table 2).
ined by comparing the cumulative hazard plots grouped on An inverse gradient (ptrend < 0.001) of total CVD event
exposure; no appreciable violations were noted. The poten- rates was observed across CRF groups in men (table 3).
tial influence of undetected subclinical disease at baseline After adjustment for covariables, men with moderate and
was evaluated by excluding events that occurred during the high CRF had an 18 percent and 39 percent lower CVD risk
first year of follow-up; little change was noted. All p values than did men with low CRF (ptrend < 0.001). The inverse
are two sided, and p < 0.05 was regarded as statistically association remained significant after additional adjust-
significant. ment for body mass index, hypertension, diabetes, and

Am J Epidemiol 2007;165:1413–1423
1416 Sui et al.

TABLE 2. Baseline characteristics of study participants by sex and cardiovascular

disease event status, Aerobics Center Longitudinal Study, Dallas, Texas, 1971–2004

Men Women
Noncases Cases Noncases Cases
(n ¼ 19,216) (n ¼ 1,512) (n ¼ 5,750) (n ¼ 159)
Characteristic
(mean (SD*) (mean (SD) (mean (SD) (mean (SD)
or %) or %) or %) or %)

Age (years) 43.9 (9.6) 50.3 (8.7) 44.4 (10.2) 52.3 (10.0)
Body mass index
(kg/m2) 26.1 (3.6) 26.4 (3.3) 23.0 (3.8) 23.6 (3.5)
Treadmill time
(minutes) 17.6 (5.0) 15.4 (4.8) 13.1 (4.6) 10.5 (4.1)

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

Maximal METs* 11.5 (2.5) 10.5 (2.3) 9.4 (2.1) 8.2 (1.9)
Lipids (mmol/liter)
Total cholesterol 5.4 (1.0) 5.9 (1.0) 5.2 (1.0) 5.7 (1.0)
HDL* cholesterol 1.2 (0.3) 1.1 (0.3) 1.6 (0.4) 1.6 (0.4)
Triglycerides 1.5 (1.2) 1.8 (1.3) 1.1 (0.8) 1.3 (0.8)
Fasting blood glucose
(mmol/liter) 5.6 (0.9) 5.8 (1.4) 5.2 (0.7) 5.5 (1.2)
Blood pressure
(mmHg)
Systolic 121.7 (13.7) 126.5 (14.9) 113.4 (14.5) 121.7 (15.6)
Diastolic 81.0 (9.7) 83.4 (9.7) 75.7 (9.5) 80.2 (9.5)
Sedentary (%) 36.7 42.1 30.9 34.1
Current smoker (%) 18.3 19.4 9.2 14.5
Alcohol intake (5
drinks/week)y (%) 40.6 39.3 18.9 24.5
Abnormal exercise
ECG* (%) 4.4 14.6 4.9 13.2
Hypertensionz (%) 30.7 45.0 17.4 40.9
Diabetes mellitus§ (%) 4.9 9.2 3.0 4.4
Hypercholesterolemia{ (%) 18.5 34.1 13.5 25.2
Hypertriglyceridemia# (%) 14.7 21.5 4.5 11.3
Low HDL cholesterol** (%) 54.5 68.1 25.9 40.3
Dyslipidemiayy (%) 82.7 91.9 52.6 70.4
Family history of CVD* (%) 15.8 18.5 18.5 17.0

* SD, standard deviation; METs, metabolic equivalents; HDL, high density lipoprotein; ECG,
electrocardiogram; CVD, cardiovascular disease.
y One unit of alcohol is defined as 12 ounces (3.41 dl) of beer, 5 ounces (1.421 dl) of wine, or
1.5 ounces (0.4262 dl) of hard liquor.
z Hypertension is defined as systolic blood pressure of 140 mmHg or higher, diastolic blood
pressure of 90 mmHg or higher, or previous diagnosis by a physician.
§ Diabetes mellitus is defined as a fasting plasma glucose concentration of 7.0 mmol/liter (126
mg/dl) or higher, previous diagnosis by a physician, or insulin use.
{ Hypercholesterolemia is defined as total cholesterol of 6.20 mmol/liter (240 mg/dl) or higher
or previous diagnosis by a physician.
# Hypertriglyceridemia is defined as triglycerides of 2.26 mmol/liter (200 mg/dl) or higher.
** Low HDL cholesterol is defined as less than 1.03 mmol/liter (40 mg/dl).
yy Dyslipidemia is defined as the presence of one or more of the above lipid abnormalities.

dyslipidemia (ptrend < 0.001). Similar inverse patterns of women with low CRF (ptrend ¼ 0.05). CRF remained in-
association were observed between CRF and each second- versely associated with CVD risk after additional adjust-
ary outcome. ment for intermediate risk factors, although the trend was
In women (table 4), total CVD event rates were inversely not significant (ptrend ¼ 0.30). CRF was inversely associated
associated with CRF (ptrend ¼ 0.002). After adjustment for with coronary heart disease event rates (ptrend ¼ 0.004);
covariables, women with moderate and high CRF had a 26 however, significance was attenuated by adjustment for
percent and a 37 percent lower risk of CVD events than did covariables (ptrend ¼ 0.09) and intermediate risk factors

Am J Epidemiol 2007;165:1413–1423
 Cardiorespiratory Fitness and Cardiovascular Events 1417

TABLE 3. Rates and hazard ratios for cardiovascular disease events by

cardiorespiratory fitness groups in men, Aerobics Center Longitudinal Study, Dallas,
Texas, 1971–2004*

Disease event by 95% 95%

No. of Hazard Hazard
cardiorespiratory Ratey confidence confidence
events ratioz ratio§
fitness group intervalz interval§

Total cardiovascular
disease
Low 345 107.9 1.00 Referent 1.00 Referent
Moderate 664 75.2 0.82 0.72, 0.94 0.89 0.78, 1.02
High 503 50.3 0.61 0.53, 0.71 0.75 0.64, 0.87
p for linear trend <0.001 <0.001 0.001

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

Coronary heart disease
Low 289 88.9 1.00 Referent 1.00 Referent
Moderate 533 60.5 0.81 0.70, 0.94 0.89 0.77, 1.03
High 400 40.3 0.61 0.52, 0.71 0.76 0.64, 0.90
p for linear trend <0.001 <0.001 0.001
Myocardial infarction
Low 123 35.6 1.00 Referent 1.00 Referent
Moderate 212 24.1 0.80 0.64, 1.01 0.87 0.69, 1.09
High 154 16.2 0.60 0.46, 0.77 0.73 0.56, 0.96
p for linear trend <0.001 <0.001 0.02
Stroke
Low 56 19.0 1.00 Referent 1.00 Referent
Moderate 131 14.8 0.86 0.63, 1.18 0.90 0.65, 1.24
High 103 10.0 0.63 0.45, 0.89 0.71 0.49, 1.01
p for linear trend <0.001 0.005 0.04

* There were 66,887, 70,222, and 78,872 man-years of follow-up in the low, moderate, and
high fitness groups, respectively.
y Rate per 10,000 person-years adjusted for age and examination year.
z Adjusted for the above plus current smoking (yes or no), alcohol intake (5 drinks/week or
not), family history of cardiovascular disease (present or not), and abnormal exercise
electrocardiogram responses (present or not).
§ Adjusted for the above plus body mass index (<25 or 25 kg/m2), hypertension, diabetes, or
dyslipidemia (present or not for each).

(ptrend ¼ 0.49). Lower myocardial infarction and stroke rates and 6). In men, after adjustment for age and examination
also were observed in women with moderate and high CRF, year, each 1-minute increment of maximal exercise was, on
but these associations were not statistically significant. average, associated with a 3–9 percent (p < 0.05) lower
We also examined whether CRF predicted CVD events CVD risk in each risk factor group, adverse or not. The
independent of reported physical activity status. Age- and consistency in the direction and magnitude of association
examination year-adjusted rates of total CVD events (per between CRF and CVD suggested that there was little effect
10,000 person-years) were inversely associated with physi- modification across risk factor categories. Further adjust-
cal activity status in men (sedentary ¼ 78.4 vs. active ¼ ment for the other risk factors eliminated some but not all
64.8; p < 0.001) but not in women (sedentary ¼ 22.3 vs. of the associations. Results were similar for coronary heart
active ¼ 28.5; p ¼ 0.20). After adjustment for age, exami- disease events and for myocardial infarction (data not
nation year, and physical activity status, hazard ratios in the shown). In women, the pattern of association between
low, moderate, and high CRF groups were 1.00 (referent), CRF and CVD risk was variable across risk factor groups,
0.77 (95 percent confidence interval (CI): 0.68, 0.89), and and statistical power often was limited by a small number of
0.55 (95 percent CI: 0.47, 0.64), ptrend < 0.001, in men events.
and were 1.00 (referent), 0.67 (95 percent CI: 0.44, 1.01), To examine whether CRF had prognostic value beyond an
and 0.57 (95 percent CI: 0.33, 0.81), ptrend ¼ 0.005, in individual’s pretest probability of having a CVD event, we
women. Results were similar for secondary outcomes. computed CVD rates by CRF levels grouped on the number
We next examined whether other risk predictors modified of major CVD risk factors at baseline (figures 1 and 2). By
the association between CRF and total CVD events (tables 5 convention (33), individuals with zero risk factors would be

Am J Epidemiol 2007;165:1413–1423
1418 Sui et al.

TABLE 4. Rates and hazard ratios for cardiovascular disease events by

cardiorespiratory fitness groups in women, Aerobics Center Longitudinal Study, Dallas,
Texas, 1971–2004*

Disease event by 95% 95%

No. of Hazard Hazard
cardiorespiratory Ratey confidence confidence
events ratioz ratio§
fitness group intervalz interval§

Total cardiovascular
disease
Low 35 41.9 1.00 Referent 1.00 Referent
Moderate 63 27.7 0.74 0.49, 1.13 0.83 0.54, 1.28
High 61 20.8 0.63 0.40, 0.98 0.78 0.49, 1.23
p for linear trend 0.002 0.05 0.30

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

Coronary heart disease
Low 22 26.6 1.00 Referent 1.00 Referent
Moderate 40 17.7 0.79 0.47, 1.35 0.93 0.54, 1.60
High 35 11.8 0.61 0.35, 1.09 0.82 0.45, 1.48
p for linear trend 0.004 0.09 0.49
Myocardial infarction
Low 12 13.7 1.00 Referent 1.00 Referent
Moderate 24 10.5 0.92 0.45, 1.88 1.08 0.53, 2.22
High 17 6.1 0.62 0.28, 1.36 0.81 0.36, 1.82
p for linear trend 0.03 0.19 0.55
Stroke
Low 13 15.3 1.00 Referent 1.00 Referent
Moderate 23 10.0 0.65 0.33, 1.31 0.68 0.34, 1.38
High 26 9.1 0.64 0.31, 1.30 0.69 0.33, 1.44
p for linear trend 0.18 0.28 0.40

* There were 19,808, 19,504, and 20,853 woman-years of follow-up in the low, moderate, and
high fitness groups, respectively.
y Rate per 10,000 person-years adjusted for age and examination year.
z Adjusted for the above plus current smoking (yes or no), alcohol intake (5 drinks/week or
not), family history of cardiovascular disease (present or not), and abnormal exercise elec-
trocardiogram responses (present or not).
§ Adjusted for the above plus body mass index (<25 or 25 kg/m2), hypertension, diabetes, or
dyslipidemia (present or not for each).

classified as low risk (e.g., expected 10-year probability of lated to incident events that are survived and not merely to
<10 percent), whereas those with one or more risk factors mortality, as well as whether protection is conferred in both
would have an intermediate to high CVD risk (e.g., 10-year women and men. The present study demonstrated that
probability of 10 percent). In men, we observed inverse higher CRF was associated with significantly lower rates
gradients of CVD rates across CRF categories within each of nonfatal CVD events. The inverse pattern of association
risk factor stratum (p < 0.01 each). Similar inverse patterns was present in women and men and in those with a low or
of association were seen in women, but the rate differences a moderate/high pretest probability of CVD. Significant as-
were not statistically significant. sociations generally persisted after considering the potential
confounding or modifying effects of physical activity status
and other risk factors, although some associations were at-
DISCUSSION tenuated in women because of low statistical power. Inverse
patterns of association also were seen between CRF and
Several prospective studies have shown that CRF is in- nonfatal coronary heart disease events and when myocardial
versely associated with CVD mortality in asymptomatic infarction and stroke were considered separately. This in-
women and men (12–16). Only a few studies in men have vestigation is one of the largest prospective studies and, to
reported on CRF and risk of nonfatal CVD events (17, 18). our knowledge, the first in women to relate an objectively
For evaluation of the true role of CRF in primary CVD measured CRF exposure with the incidence of several non-
prevention, it is important to determine whether CRF is re- fatal CVD endpoints in initially asymptomatic adults.

Am J Epidemiol 2007;165:1413–1423
 Cardiorespiratory Fitness and Cardiovascular Events 1419

TABLE 5. Hazard ratios for total cardiovascular disease events per 1-minute increment in maximal
exercise duration according to cardiovascular disease risk factor categories in men, Aerobics Center
Longitudinal Study, Dallas, Texas, 1971–2004*

95% 95%
No. of Hazard Hazard
Risk factor Total no. confidence p value confidence p value
events ratioy ratioz
intervaly intervalz

Age (years)
<55 17,532 1,025 0.94 0.92, 0.95 <0.001 0.95 0.94, 0.97 <0.001
55 3,196 487 0.96 0.94, 0.98 <0.001 0.98 0.96, 1.00 0.11
Current smoker
No 16,922 1,218 0.95 0.94, 0.96 <0.001 0.97 0.96, 0.99 <0.001

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

Yes 3,806 294 0.93 0.90, 0.96 <0.001 0.94 0.91, 0.96 <0.001
Family history of CVD§
No 17,411 1,233 0.95 0.94, 0.96 <0.001 0.97 0.96, 0.99 <0.001
Yes 3,317 279 0.91 0.88, 0.94 <0.001 0.93 0.90, 0.95 <0.001
Exercise ECG§
responses
Normal 19,667 1,292 0.95 0.94, 0.96 <0.001 0.96 0.95, 0.98 <0.001
Abnormal 1,061 220 0.96 0.93, 0.99 0.007 0.97 0.94, 1.01 0.12
2
Body mass index (kg/m )
18.5–24.9 8,701 573 0.94 0.92, 0.96 <0.001 0.96 0.94, 0.98 <0.001
25 12,027 939 0.95 0.94, 0.97 <0.001 0.97 0.96, 0.99 0.001
Hypertension
No 14,143 832 0.95 0.93, 0.96 <0.001 0.96 0.95, 0.98 <0.001
Yes 6,585 680 0.95 0.94, 0.97 <0.001 0.97 0.95, 0.99 <0.001
Diabetes
No 19,653 1,373 0.95 0.93, 0.96 <0.001 0.97 0.95, 0.98 <0.001
Yes 1,075 139 0.96 0.93, 1.00 0.048 0.98 0.93, 1.02 0.27
Total cholesterol
<6.20 mmol/liter
(<240 mg/dl) 16,668 997 0.94 0.93, 0.95 <0.001 0.96 0.94, 0.97 <0.001
6.20 mmol/liter
(240 mg/dl) 4,060 515 0.97 0.95, 0.99 0.003 0.98 0.96, 1.00 0.08

* The point and interval estimates are the risk of cardiovascular disease events that are associated, on average,
with each 1-minute increment in treadmill exercise duration.
y Adjusted for age and examination year.
z Adjusted for the above plus each of the other risk factors in the table.
§ CVD, cardiovascular disease; ECG, electrocardiogram.

Three of the study findings deserve further comment. In men, the inverse gradient of CVD risk across CRF
First, CRF predicted primary CVD events independent of groups remained significant after adjustment for confound-
reported physical activity status. Because physical activity ing by age, smoking, family history of CVD, abnormal ex-
assessment was crude in the present study, caution must be ercise electrocardiogram responses, and factors that may be
taken when considering the implications of this finding. intermediate in the causal pathway between CRF and CVD
Accurate questionnaire-based assessment of physical activ- (body mass index, dyslipidemia, hypertension, and diabe-
ity habits is difficult, particularly in women (11). This may tes). The present findings of a strong independent associa-
partly explain the lack of association between physical ac- tion between CRF and nonfatal CVD in men are consistent
tivity and CVD in the present women. Our findings suggest with previous ACLS findings on CRF and CVD mortality
that assessment of CRF in asymptomatic women and men (12, 13), with findings in Finnish men on CRF and nonfatal
may provide important prognostic information above that CVD (17), and with findings from studies that have related
obtained from self-reported physical activity habits. Clini- CRF (15, 20, 21, 34) or reported physical activity (5, 6, 8)
cians should, therefore, consider the benefits and feasibility with combined fatal/nonfatal CVD in men. Similar patterns
of more routine exercise testing. of association generally were seen in women. Lack of

Am J Epidemiol 2007;165:1413–1423
1420 Sui et al.

TABLE 6. Hazard ratios for total cardiovascular disease events per 1-minute increment of maximal
exercise duration according to cardiovascular disease risk factor categories in women, Aerobics Center
Longitudinal Study, Dallas, Texas, 1971–2004*

95% 95%
No. of Hazard Hazard
Risk factor Total no. confidence p value confidence p value
events ratioy ratioz
intervaly intervalz

Age (years)
<55 4,864 93 0.94 0.89, 0.99 0.03 0.96 0.91, 1.02 0.23
55 1,045 66 0.96 0.89, 1.04 0.30 0.98 0.90, 1.06 0.61
Current smoker
No 5,358 136 0.96 0.92, 1.01 0.14 0.98 0.93, 1.03 0.49

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

Yes 551 23 0.89 0.77, 1.01 0.07 0.87 0.75, 1.00 0.05
Family history of CVD§
No 4,820 132 0.95 0.91, 1.00 0.05 0.97 0.92, 1.02 0.23
Yes 1,089 27 0.93 0.83, 1.04 0.22 0.97 0.85, 1.10 0.61
Exercise ECG§
responses
Normal 5,604 138 0.95 0.90, 0.99 0.03 0.96 0.91, 1.01 0.11
Abnormal 305 21 1.00 0.85, 1.17 0.97 1.12 0.93, 1.34 0.24
Body mass index
(kg/m2)
18.5–24.9 4,644 119 0.94 0.89, 0.94 0.01 0.95 0.90, 1.01 0.07
25 1,265 40 1.02 0.91, 1.14 0.80 1.03 0.92, 1.16 0.60
Hypertension
No 4,846 94 0.97 0.92, 1.03 0.33 1.00 0.94, 1.06 0.93
Yes 1,063 65 0.94 0.87, 1.01 0.07 0.92 0.85, 0.99 0.04
Diabetes
No 5,732 152 0.94 0.90, 0.99 0.01 0.96 0.91, 1.01 0.08
Yes 177 7 1.09 0.82, 1.44 0.57 1.15 0.73, 1.79 0.55
Total cholesterol
<6.20 mmol/liter
(240 mg/dl) 5,091 119 0.93 0.89, 0.98 0.009 0.95 0.90, 1.01 0.07
6.20 mmol/liter
(240 mg/dl) 818 40 0.99 0.90, 1.09 0.86 1.03 0.93, 1.13 0.63

* The point and interval estimates are the risk of cardiovascular disease events that are associated, on average,
with each 1-minute increment in treadmill exercise duration.
y Adjusted for age and examination year.
z Adjusted for the above plus each of the other risk factors in the table.
§ CVD, cardiovascular disease; ECG, electrocardiogram.

a significant association in the fully adjusted model that other CVD predictors. The prognostic value of CRF is par-
included biologic intermediates may be due to the small ticularly noteworthy in men who were older and who had
number of cases and is consistent with some (5, 8, 20) but diabetes, exercise electrocardiogram abnormalities, or coex-
not all (7, 9, 10, 16, 34) studies on physical activity or CRF isting risk factors at baseline. A sharp rise in the risk of
and CVD risk in women. For example, CRF predicted CVD a first CVD event occurs in adults aged 45–60 years (1).
mortality risk in women and men in the Lipid Research We observed that men aged 55 years or older had a threefold
Clinics study (16), whereas it was significantly associated higher risk of CVD events than did their younger counter-
with combined fatal/nonfatal coronary heart disease events parts. Diabetes and multiple coexisting risk factors now are
in men but not women in the Framingham Heart Study (20). seen as coronary risk equivalents in asymptomatic adults
Additional prospective data on CRF exposures and nonfatal (33). In our study, 10-year CVD risk was 50 percent greater
CVD events are needed in women to expand on the findings in men with diabetes and was threefold greater in men with
reported here and elsewhere. two or more risk factors than in men without either condi-
A second major finding was that the inverse association tion. Abnormal exercise electrocardiogram responses also
between CRF and CVD generally was consistent in strata of are predictive of CVD events (20, 21, 28) and were

Am J Epidemiol 2007;165:1413–1423
 Cardiorespiratory Fitness and Cardiovascular Events 1421

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

FIGURE 1. Age- and examination year-adjusted rates of total FIGURE 2. Age- and examination year-adjusted rates of total
cardiovascular disease (CVD) events (per 10,000 person-years) by cardiovascular disease (CVD) events (per 10,000 person-years) by
levels of cardiorespiratory fitness and number of major CVD risk levels of cardiorespiratory fitness and number of major CVD risk
factors (current smoking, hypertension, hypercholesterolemia, diabe- factors (current smoking, hypertension, hypercholesterolemia, diabe-
tes, and family history of CVD) in 20,728 men, Aerobics Center tes, and family history of CVD) in 5,909 women, Aerobics Center
Longitudinal Study, Dallas, Texas, 1971–2004. White bars represent Longitudinal Study, Dallas, Texas, 1971–2004. White bars represent
low fitness; striped bars, moderate fitness; and black bars, high low fitness; striped bars, moderate fitness; and black bars, high
fitness. The p values are for a test of linear trend across cardiorespi- fitness. The p values are for a test of linear trend across cardiorespi-
ratory fitness groups. The numbers of men (and cases) in the low, ratory fitness groups. The numbers of women (and cases) in the low,
moderate, and high fitness groups were 878 (n ¼ 42), 2,886 (n ¼ 143), moderate, and high fitness groups were 355 (n ¼ 8), 1,068 (n ¼ 21),
and 4,099 (n ¼ 154) in those with zero risk factors; 1,548 (n ¼ 120), and 1,690 (n ¼ 25) in those with zero risk factors; 349 (n ¼ 14), 725
3,422 (n ¼ 266), and 3,131 (n ¼ 226) in those with one risk factor; and (n ¼ 22), and 963 (n ¼ 24) in those with one risk factor; and 178
1,541 (n ¼ 183), 1,987 (n ¼ 255), and 1,236 (n ¼ 123) in those with (n ¼ 13), 303 (n ¼ 20), and 278 (n ¼ 12) in those with two or more risk
two or more risk factors. factors.

but also with myocardial infarction and with myocardial

associated with a twofold higher risk of CVD events among infarction and coronary revascularization combined. Myo-
men in our study. Even in these high-risk subgroups of men, cardial infarction or sudden death is the first clinical mani-
higher functional capacity was associated with significantly festation in many adults, among whom risk factors often are
lower CVD event rates. Stratified analyses were more vari- normal or only slightly elevated (33). The findings reported
able in women; however, greater functional capacity tended herein and elsewhere (13, 16, 17, 20) suggest that low CRF
to be associated with lower CVD risk across risk factor is a significant predictor of atherothrombotic CVD events in-
strata. CVD rates also were lower across incremental CRF dependent of the presence or absence of traditional risk fac-
groups in women with two or more risk factors. The statis- tors. Assessment of CRF in clinical settings could, therefore,
tical significance of these cross-tabulations in women was be an important tool to facilitate more effective primary
limited by the small number of events. CVD prevention. Effective strategies are needed to better
Collectively, the present results suggest that CRF is an integrate exercise testing into CVD risk assessment (36).
important prognostic factor for nonfatal CVD in asymptom- CRF also was inversely associated with stroke incidence
atic men beyond information obtained from the exercise in men, which is consistent with findings on CRF and stroke
electrocardiogram and traditional risk factors. Higher CRF mortality in the ACLS (37) and in Finnish studies (19).
is protective against CVD events in those with a moderate/ Others have reported inverse associations between physical
high or a low pretest probability of CVD. Assessing func- activity and stroke in women (4, 5). The inverse trend in
tional capacity in asymptomatic women likely is of similar stroke events across CRF groups was not significant in the
benefit to CVD risk assessment as in men (8); however, present women, which may partly be due to the small num-
additional data are needed to confirm the suggestive findings ber of stroke events. We were not able to differentiate be-
reported here. tween hemorrhagic and ischemic strokes, and stroke
A third noteworthy issue is the variety of CVD endpoints subtype modifies the association between physical activity
that were related to baseline CRF levels. A recent review of and stroke risk (4, 38). Additional studies on activity, fitness,
published prospective data on physical activity, CRF, and and stroke are needed to expand on our suggestive findings
CVD outcomes indicated that the strongest inverse associ- of an inverse association.
ations were for CVD mortality in men, and that additional Strengths of the current study include the extensive base-
data are needed in women and on nonfatal endpoints such as line examination to detect subclinical disease, the use of
myocardial infarction and stroke (35). In the current study, measured risk factors and of maximal exercise testing to
CRF was not only inversely related with total CVD events quantify CRF, the large person-years of follow-up, and the

Am J Epidemiol 2007;165:1413–1423
1422 Sui et al.

variety of CVD endpoints. We also accounted for variable Clinic physicians and technicians for collecting the baseline
patterns of survey responses in our analyses, an approach data, staff at the Cooper Institute for data entry and data
not typically used in cohort studies such as ours (4, 9, 38). management, and Melba Morrow for editorial assistance.
The inverse associations generally were graded and inde- Conflict of interest: none declared.
pendent of traditional risk factors, which strengthens causal
inferences. Biologic plausibility for these associations may,
for example, be through enhanced endothelial cell function
and coronary flow reserve, reduced myocardial oxygen de- REFERENCES
mand under a variety of circumstances, a higher myocardial
arrhythmia threshold, improved endogenous thrombolytic 1. Thom T, Haase N, Rosamond W, et al. Heart disease and stroke
activity, and lower levels of circulating atherothrombotic statistics—2006 update: a report from the American Heart
cytokines that may promote coronary plaque stabilization Association Statistics Committee and Stroke Statistics

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

(11). Subcommittee. Circulation 2006;113:e85–151.
The homogeneity of our population sample in sociode- 2. Thompson PD, Buchner D, Pina IL, et al. Exercise and phys-
ical activity in the prevention and treatment of atherosclerotic
mographic factors enhances the internal validity of our find- cardiovascular disease: a statement from the Council on
ings by reducing confounding by these factors. Although the Clinical Cardiology (Subcommittee on Exercise, Rehabilita-
self-referred origin and homogeneity of our cohort also may tion, and Prevention) and the Council on Nutrition, Physi-
be seen as a weakness, we believe that our data are no less cal Activity, and Metabolism (Subcommittee on Physical
meaningful than those from population samples of adults Activity). Circulation 2003;107:3109–16.
referred to exercise testing for clinical reasons (39) or data 3. Leon AS, Connett J, Jacobs DR Jr, et al. Leisure-time physical
from other selected cohorts that have been influential in activity levels and risk of coronary heart disease and death.
preventive cardiology (6, 14, 20). Our findings should be The Multiple Risk Factor Intervention Trial. JAMA 1987;
generalized carefully to other adult populations. We did not 258:2388–95.
have sufficient information on medication usage, meno- 4. Hu G, Sarti C, Jousilahti P, et al. Leisure time, occupational,
and commuting physical activity and the risk of stroke. Stroke
pausal status, or dietary habits to include in our analysis. 2005;36:1994–9.
It is possible that residual confounding by these factors may 5. Salonen JT, Puska P, Tuomilehto J. Physical activity and risk
exist, although it seems unlikely that it would account for all of myocardial infarction, cerebral stroke and death: a longitu-
of the observed association between CRF and CVD. Future dinal study in eastern Finland. Am J Epidemiol 1982;115:
studies should include such information to expand on the 526–37.
findings reported here. Women tend to manifest CVD events 6. Paffenbarger RS Jr, Hyde RT, Wing AL, et al. A natural history
10 years later than men. In the present study, the age distri- of athleticism and cardiovascular health. JAMA 1984;252:
bution in women was insufficient for grouped analysis 491–5.
beyond 55 years of age. Genetics clearly contribute to max- 7. Rockhill B, Willett WC, Manson JE, et al. Physical activity
imal CRF (40, 41). Nonetheless, CRF can be enhanced in and mortality: a prospective study among women. Am J Public
Health 2001;91:578–83.
most individuals through participation in moderate and vig- 8. O’Connor GT, Hennekens CH, Willett WC, et al. Physical
orous physical activities, such as brisk walking, bicycling, exercise and reduced risk of nonfatal myocardial infarction.
and jogging, for 30 minutes or more on most days of the Am J Epidemiol 1995;142:1147–56.
week (2). 9. Lee IM, Rexrode KM, Cook NR, et al. Physical activity and
We conclude that CRF is a significant determinant of non- coronary heart disease in women: is ‘‘no pain, no gain’’ passe?
fatal primary CVD events in women and men. Assessment of JAMA 2001;285:1447–54.
CRF provides important prognostic information indepen- 10. Manson JE, Greenland P, LaCroix AZ, et al. Walking com-
dent of exercise electrocardiogram responses and traditional pared with vigorous exercise for the prevention of cardiovas-
risk factors, and in those with high and low pretest probabil- cular events in women. N Engl J Med 2002;347:716–25.
ities of CVD. Exercise testing to assess functional capacity 11. Haskell WL, Leon AS, Caspersen CJ, et al. Cardiovascular
benefits and assessment of physical activity and physical
may enhance CVD risk stratification beyond conventional fitness in adults. Med Sci Sports Exerc 1992;24(suppl 6):
office-based methods in asymptomatic adults. We believe S201–20.
that clinicians should consider the benefits of assessing 12. Blair SN, Kohl HW 3rd, Paffenbarger RS Jr, et al. Physical
CRF and that they should vigilantly counsel their sedentary fitness and all-cause mortality: a prospective study of healthy
patients to become more physically active and to improve men and women. JAMA 1989;262:2395–401.
their CRF as a cornerstone of primary CVD prevention. 13. Blair SN, Kampert JB, Kohl HW, et al. Influences of cardio-
respiratory fitness and other precursors on cardiovascular
disease and all-cause mortality in men and women. JAMA
1996;276:205–10.
14. Ekelund LG, Haskell WL, Johnson JL, et al. Physical fitness as
ACKNOWLEDGMENTS a predictor of cardiovascular mortality in asymptomatic North
American men. The Lipid Research Clinics Mortality Follow-up
Supported by National Institute of Health grants AG06945 Study. N Engl J Med 1988;319:1379–84.
and HL62508 and by the Communities Foundation of Texas 15. Lakka TA, Venalainen JM, Rauramaa R, et al. Relation of
on recommendation of Nancy Ann and Ray L. Hunt. leisure-time physical activity and cardiorespiratory fitness to
The authors thank Dr. Kenneth H. Cooper for establish- the risk of acute myocardial infarction in men. N Engl J Med
ing the Aerobics Center Longitudinal Study, the Cooper 1994;330:1549–54.

Am J Epidemiol 2007;165:1413–1423
 Cardiorespiratory Fitness and Cardiovascular Events 1423

16. Mora S, Redberg RF, Cui Y, et al. Ability of exercise testing to 28. Gibbons LW, Mitchell TL, Wei M, et al. Maximal exercise test
predict cardiovascular and all-cause death in asymptomatic as a predictor of risk for mortality from coronary heart disease
women: a 20-year follow-up of the lipid research clinics in asymptomatic men. Am J Cardiol 2000;86:53–8.
prevalence study. JAMA 2003;290:1600–7. 29. Macera CA, Jackson KL, Davis DR, et al. Patterns of non-
17. Laukkanen JA, Kurl S, Salonen R, et al. The predictive value response to a mail survey. J Clin Epidemiol 1990;43:1427–30.
of cardiorespiratory fitness for cardiovascular events in men 30. Luepker RV, Apple FS, Christenson RH, et al. Case definitions
with various risk profiles: a prospective population-based for acute coronary heart disease in epidemiology and clinical
cohort study. Eur Heart J 2004;25:1428–37. research studies: a statement from the AHA Council on Epi-
18. Miller GJ, Cooper JA, Beckles GL. Cardiorespiratory fitness, demiology and Prevention; AHA Statistics Committee; World
all-cause mortality, and risk of cardiovascular disease in Tri- Heart Federation Council on Epidemiology and Prevention;
nidadian men—the St James survey. Int J Epidemiol 2005; the European Society of Cardiology Working Group on Epi-
34:1387–94. demiology and Prevention; Centers for Disease Control and
19. Kurl S, Laukkanen JA, Rauramaa R, et al. Cardiorespiratory Prevention; and the National Heart, Lung, and Blood Institute.

Downloaded from https://academic.oup.com/aje/article-abstract/165/12/1413/125453 by guest on 14 May 2020

fitness and the risk for stroke in men. Arch Intern Med 2003; Circulation 2003;108:2543–9.
163:1682–8. 31. Kelly-Hayes M, Robertson JT, Broderick JP, et al. The
20. Balady GJ, Larson MG, Vasan RS, et al. Usefulness of exercise American Heart Association Stroke Outcome Classification.
testing in the prediction of coronary disease risk among Stroke 1998;29:1274–80.
asymptomatic persons as a function of the Framingham risk 32. Manson JE, Stampfer MJ, Hennekens CH, et al. Body weight
score. Circulation 2004;110:1920–5. and longevity. A reassessment. JAMA 1987;257:353–8.
21. Bodegard J, Erikssen G, Bjornholt JV, et al. Reasons for ter- 33. Greenland P, Smith SC Jr, Grundy SM. Improving coronary
minating an exercise test provide independent prognostic in- heart disease risk assessment in asymptomatic people: role of
formation: 2014 apparently healthy men followed for 26 years. traditional risk factors and noninvasive cardiovascular tests.
Eur Heart J 2005;26:1394–401. Circulation 2001;104:1863–7.
22. Pickering TG, Hall JE, Appel LJ, et al. Recommendations for 34. Roger VL, Jacobsen SJ, Pellikka PA, et al. Prognostic value of
blood pressure measurement in humans and experimental treadmill exercise testing: a population-based study in
animals. Part 1. Blood pressure measurement in humans: Olmsted County, Minnesota. Circulation 1998;98:2836–41.
a statement for professionals from the Subcommittee of 35. Kohl HW 3rd. Physical activity and cardiovascular disease:
Professional and Public Education of the American Heart evidence for a dose response. Med Sci Sports Exerc
Association Council on High Blood Pressure Research. 2001;33(suppl 6):S472–83.
Hypertension 2005;45:142–61. 36. Lauer M, Froelicher ES, Williams M, et al. Exercise testing in
23. Balke B, Ware RW. An experimental study of physical fitness asymptomatic adults: a statement for professionals from the
in Air Force personnel. U S Armed Forces Med J 1959;10: American Heart Association Council on Clinical Cardiology,
675–88. Subcommittee on Exercise, Cardiac Rehabilitation, and Pre-
24. Pollock ML, Bohannon RL, Cooper KH, et al. A comparative vention. Circulation 2005;112:771–6.
analysis of four protocols for maximal treadmill stress testing. 37. Lee CD, Blair SN. Cardiorespiratory fitness and stroke mor-
Am Heart J 1976;92:39–46. tality in men. Med Sci Sports Exerc 2002;34:592–5.
25. Pollock ML, Foster C, Schmidt D, et al. Comparative analysis 38. Hu FB, Stampfer MJ, Colditz GA, et al. Physical activity and
of physiologic responses to three different maximal graded risk of stroke in women. JAMA 2000;283:2961–7.
exercise test protocols in healthy women. Am Heart J 1982; 39. Myers J, Prakash M, Froelicher V, et al. Exercise capacity and
103:363–73. mortality among men referred for exercise testing. N Engl J
26. American College of Sports Medicine. ACSM’s guidelines for Med 2002;346:793–801.
exercise testing and prescription. 6th ed. Philadelphia, PA: 40. Bouchard C, Daw EW, Rice T, et al. Familial resemblance for
Lippincott Williams & Wilkins, 2000. VO2max in the sedentary state: the HERITAGE family study.
27. Barlow CE, LaMonte MJ, FitzGerald SJ, et al. Cardiorespi- Med Sci Sports Exerc 1998;30:252–8.
ratory fitness is an independent predictor of hypertension in- 41. Bouchard C, An P, Rice T, et al. Familial aggregation of
cidence among initially normotensive healthy women. Am J VO2max response to exercise training: results from the HERI-
Epidemiol 2006;163:142–50. TAGE family study. J Appl Physiol 1999;87:1003–8.

Am J Epidemiol 2007;165:1413–1423