Characterizing differences in mortality at the low end of the fitness spectrum in individuals with cardiovascular disease

Abstract

Background

A graded but nonlinear relationship exists between fitness and mortality, with the greatest mortality differences occurring between the least-fit (first, Q1) and the next-least-fit (second, Q2) quintiles of fitness. The purpose of this study was to compare clinical characteristics, exercise test responses, and physical activity (PA) patterns in Q1 versus Q2 in patients with cardiovascular disease (CVD).

Design

Observational retrospective study.

Methods

A total of 5101 patients with a history of CVD underwent clinical treadmill testing and were followed up for 9.1 ± 5.5 years. Patients were classified into quintiles of exercise capacity measured in metabolic equivalents. Clinical characteristics, treadmill test results, and recreational PA patterns were compared between Q1 (n = 923) and Q2 (n = 929).

Results

Q1 had a nearly two-fold increase in age-adjusted relative risk of cardiovascular mortality compared with Q2 (hazard ratio: 3.79 vs. 2.04, P < 0.05; reference: fittest quintile). Q1 patients were older, had more extensive use of medications, and were more likely to have a history of typical angina (35 vs. 28%), myocardial infarction (30 vs. 24%), chronic heart failure (25 vs. 14%), claudication (15 vs. 9%) and stroke (9 vs. 6%) compared with Q2 (all comparisons: P < 0.05). Recent and lifetime recreational PA was not different between the two groups.

Conclusion

Greater severity of disease in the least-fit versus the next-least-fit quintile likely contributes to but cannot fully explain marked differences in mortality rates in CVD patients. To achieve potential survival benefits, our results suggest that unfit CVD patients should engage in exercise programs of sufficient volume and intensity to improve fitness. Eur J Cardiovasc Prev Rehabil 17:289-295 © 2010 The European Society of Cardiology

cardiovascular disease, fitness, mortality, physical activity

Introduction

An inverse, graded, and dose-dependent relationship between physical fitness and mortality has been consistently reported in epidemiological studies, irrespective of the presence or absence of cardiovascular disease (CVD) [1–4]. This relationship is typically nonlinear, with the greatest decline in risk of mortality observed between the least-fit (first, Q1) and the next-least-fit (second, Q2) groups of individuals with and without CVD [1, 3–6]. However, factors that may explain the steep mortality gradient in the lowest end of the fitness spectrum remain largely unknown. We recently observed that lower recent physical activity (PA) rather than differences in clinical characteristics, in part, explained the striking difference in mortality rates between Q1 and Q2 in apparently healthy individuals [7]. It remains unknown whether differences in clinical characteristics reflecting differences in severity of underlying disease, PA patterns, behavior, environmental, or other factors (e.g. genetics) explain the steep mortality gradient between Q1 and Q2 in patients with CVD.

The purpose of this study was to compare clinical characteristics, PA patterns, and exercise test responses between Q1 and Q2 in patients with existing CVD.

Methods

The study population consisted of 5101 CVD patients (134 women) referred for treadmill testing for clinical reasons from 1987 to 2006. Detailed clinical histories, current medications, risk factors, and type of CVD were recorded prospectively on computerized forms [8, 9]. All participants had a history of CVD, an abnormal exercise test response, or both. History of CVD included a history of angiographically documented coronary artery disease (7%), myocardial infarction (28%), coronary bypass surgery (15%), coronary angioplasty (10%), typical angina (28%), chronic heart failure (14%), claudication (8%), stroke (6%), or atrial fibrillation (7%). Abnormal exercise test results were defined as exercise-induced angina (16%) and/or ST-segment abnormalities (24%) (ST-segment depression ≥ 1.0mm downsloping or horizontal during exercise, in recovery, or both). Nine percent of the total population (477 individuals) had a history of mild chronic obstructive pulmonary disease.

The population was divided into quintiles of fitness based on metabolic equivalents (METs) achieved on a symptomlimited exercise test. Cut-off points between the categories were set at every 20th percentile to approximate similar sample sizes in each quintile.

Exercise testing

Patients performed symptom-limited treadmill testing using an individualized ramp treadmill protocol [10]. Before the exercise tests, exercise capacity was estimated for each patient using a questionnaire such that the protocol was individualized to achieve maximal exercise capacity within 8–12 min in most patients [11]. In the absence of symptoms or other signs of ischemia, patients were encouraged to exercise until volitional fatigue. The use of handrails during exercise was discouraged. Target heart rates were not used as predetermined endpoints. A 12-lead ECG was monitored throughout the test. After exercise, patients were placed in a supine position. Medications were not changed or stopped before testing. The exercise tests were performed, analyzed, and reported according to a standardized protocol using a computerized database [12]. The protocol was approved by the Stanford University Investigational Review Board, and written consent was obtained from all participants before exercise testing.

Peak exercise capacity (in METs) was estimated based on the speed and grade of the treadmill [13]. One MET is defined as the energy expended sitting quietly, which is equivalent to an oxygen consumption of approximately 3.5 ml/kg of body weight per minute for an average adult. Normal standards for age-predicted exercise capacity were derived from regression equations developed from veterans referred for exercise testing [14], using the equation [18.0-(0.15 × age)]. The percentage of normal exercise capacity achieved was defined as follows: (achieved exercise capacity predicted exercise capacity) × 100. Age-predicted maximal heart rate was calculated as [187 - (0.85 × age)] [14].

Physical activity questionnaires

PA data were collected between 1992 and 2006, depending on the availability of research assistants during that period, and were available in a subgroup of 652 patients. PA patterns were quantified using a five-page questionnaire modified from the Harvard Alumni studies of Paffenbarger and colleagues [15], which has been previously described in detail [16]. Briefly, metabolic cost of lifetime and recent recreational PAs were computed, and energy expenditure on activity was expressed in kilocalories per week (kcal/week). Both recent (within the last year) and lifetime (adulthood since 25 years of age) activities were quantified. Patients were categorized as sedentary (< 1000 kcal/week), moderately active (1000-1999 kcal/week), and active (> 2000 kcal/week). These cut-points were based on widely available guidelines on PA and health suggesting that a total weekly activity energy expenditure of ≥ 1000 kcal/week is associated with the prevention of chronic disease [17–20], whereas an energy expenditure of ≥ 2000 kcal/week has been used as the benchmark for ‘physically active’ and is associated with marked reductions in risk [15].

Follow-up

The Veterans Affairs electronic medical records system was used to match all individuals to their records and to confirm clinical and demographic data. Vital status was determined as of October 2007. The Social Security Death Index was used to obtain mortality data. Causes of death were independently determined and confirmed by consensus of two physicians who were blinded to exercise test results. Cardiovascular death was defined as death because of stroke or cardiac reasons.

Statistical analysis

NCSS software (Kayesville, Utah, USA) was used for all statistical analyses. Differences between the five quintiles of exercise capacity were compared using analysis of variance with Bonferroni post-hoc multiple comparisons for continuous variables and χ² tests for discrete variables. Clinical characteristics of Q1 and Q2 were compared using unpaired t-tests and χ² tests where appropriate. All-cause and cardiovascular mortality were used as the endpoints for Kaplan-Meier survival analysis. Cox proportional hazards analysis was used to determine which variables were independently and significantly associated with time to death in multivariate models. Analyses were adjusted for age in years as a continuous variable. The relative risks for both all-cause and cardiovascular mortality were calculated for each quintile of fitness, with the fittest group as the referent. Continuous variables are presented as mean ± SD, whereas categorical variables are expressed as absolute and relative (percent) frequencies. P values less than 0.05 were considered statistically significant.

Results

Demographic characteristics

During a mean (± SD) follow-up period of 9.1 ± 5.5 years, there were a total of 1424 (28%) deaths from any cause, of which 468 (32.9%) were cardiovascular in origin. Significant trends were observed for decreasing age and both all-cause and cardiovascular mortality from the least-fit (Q1) to the most-fit (Q5) quintile of fitness (Table 1). Age-adjusted relative risks of all-cause and cardiovascular mortality were progressively lower with increasing quintiles of fitness (Fig. 1). However, the largest reduction in mortality risk was observed between the least-fit (Q1) and the next-least-fit (Q2) quintiles of fitness, with smaller differences observed between other quintiles (Fig. 1). The data fit a polynomial curve with the following equations:

Cardiovascular death: y = 0.26x  ² + 2.4x − 0.41 (R  ² = 0.90)

All-cause death: y = 0.08x  ² + 0.04x + 0.82 (R  ² = 0.99)

Trends across quintiles of fitness

Use of cardiovascular medications and prevalence of risk factors were significantly higher among the least fit compared with the more fit quintiles (Table 1). In addition, the prevalence of history of typical angina, myocardial infarction, chronic heart failure, claudication, stroke, atrial fibrillation, chronic obstructive pulmonary disease, and coronary artery bypass surgery was lower at the low end than at the high end of the fitness spectrum (Table 1). In contrast, the highest prevalence of dysplidemia and percutaneous angioplasty interventions were observed in the most-fit quintile of CVD patients. The prevalence of other conditions such as chronic renal insufficiency, cancer, neurological, liver, or endocrine disease was less than 1% in the total population and was not different across quintiles of fitness.

By using analysis of variance, significant trends were observed for decreased resting heart rate, increased peak exercise heart rate (absolute and age-predicted), systolic blood pressure, and absolute and age-predicted exercise capacity from the lowest to the highest quintile of fitness (Table 2). The prevalence of exercise-induced angina and ST-segment abnormalities were lower in the highest than in the lowest quintiles of fitness.

Recreational PA data showed a significant main effect for higher recent PA (P = 0.004) and a borderline effect for higher lifetime PA (P = 0.06) from the lowest to the highest quintile of fitness (Fig. 2).

Comparison between the least-fit and next-least-fit quintiles

Both all-cause and cardiovascular mortality were significantly higher in Q1 versus Q2 (48 vs. 30% for all-cause and 17 vs. 9% for cardiovascular mortality, respectively, P < 0.001). Patients in Q1 were older, had more extensive use of cardiovascular medications (nitrates and antihypertensive agents), and were more likely to have a history of typical angina, myocardial infarction, claudication, chronic heart failure, stroke, and atrial fibrillation compared with Q2 (Table 1). The prevalence of dyslipidemia, history of smoking, and treatment with angiotensin-converting enzyme inhibitors, Ca²⁺ channel blockers, statins, and diuretics were lower in Q1 than in Q2 (Table 1). Exercise test responses (peak exercise heart rate, peak systolic blood pressure, and exercise capacity) were attenuated in Q1 compared with Q2 (Table 2).

Energy expenditure expressed as age-adjusted recent (1263 ± 1783 vs. 1566 ± 2161 kcal/week, P = 0.22) and lifetime (1323 ± 1847 vs. 1619 ± 1830 kcal/week, P = 0.23) recreational PA were not different between Q1 (n = 122) and Q2 (n = 132). Nearly two-thirds of patients at the lower end of the fitness spectrum did not meet minimal recommendations for activity in the last year (57% in Q1 and 59% in Q2).

Multivariate predictors of mortality in cardiovascular disease

After adjustment for age, risk factors, cardiovascular medications and medical history, exercise capacity was the strongest predictor of all-cause mortality [hazard ratio: 0.89 (95% confidence interval: 0.87-0.91), P < 0.001] and the second strongest predictor of cardiovascular mortality (after history of chronic heart failure) [hazard ratio: 0.87 (95% confidence interval: 0.84-0.90), P < 0.001]. Each 1-MET increase in exercise capacity conferred an 11% reduction in risk for all-cause mortality and a 13% reduction in risk for cardiovascular mortality. When only Q1 and Q2 were considered, risk reductions per MET were 22% for all-cause mortality and 31% for cardiovascular mortality.

Physical activity subgroup versus the entire cohort

The subgroup of patients with PA data had shorter follow-up and lower mortality rates when compared with the total population. In addition, these patients were older, had a higher prevalence of hypertension, dyslipidemia and smoking, and usage of most cardiovascular medications, and were more likely to have a history of stroke, coronary artery bypass grafting, percutaneous coronary angiogram, and abnormal exercise test responses (Table 3). Other clinical characteristics and exercise capacity were not different between the cohorts.

Table 1

Demographic and clinical characteristics between quintiles of fitness

		Quintiles of exercise capacity (estimated METs)
	Total (N=5101)	Q1 Lowest (N = 923)	Q2 (N = 929)	Q3 (N = 993)	Q4 (N = 1045)	Q5 Highest (N = 1211)	P value	P value Q1 versus Q2
Demographics
Age (years)	61 ± 10	67 ± 9∗	64 ± 10∗,∗∗	62 ± 10∗,∗∗	60 ± 10∗,∗∗	56 ± 10∗∗	<0.001	<0.001
Male sex [n (%)]	4967 (97)	893 (97)∗	905 (97)	962 (97)∗	1018 (97)	1189 (98)∗∗	0.242	0.394
Body mass index (kg/m2)	28.3 ± 5.2	28.1 ± 5.4∗	28.6 ± 5.6∗,∗∗	28.9 ± 5.7∗,∗∗	28.6 ± 5.1∗	27.3 ± 4.2∗∗	<0.001	0.044
All-cause mortality [n (%)]	1424 (28)	446 (48)∗	279 (30)∗,∗∗	276 (28)∗,∗∗	225 (22)∗,∗∗	198 (16)∗∗	<0.001	<0.001
CV mortality [n (%)]	468 (9)	158 (17)∗	82 (9)∗,∗∗	104 (11)∗,∗∗	69 (7)∗,∗∗	55 (5)∗∗	<0.001	<0.001
Medications [n (%)]
β-blockers	1465 (29)	288 (31)∗	297 (32)∗	291 (29)∗	282 (27)∗∗	307 (25)∗∗	0.003	0.722
Calcium	1547 (30)	337 (37)∗	299 (32)∗,∗∗	296 (30)∗,∗∗	319 (31)∗,∗∗	296 (24)∗∗	<0.001	0.049
Nitrates	1659 (33)	385 (42)∗	339 (37)∗,∗∗	288 (29)∗∗	312 (30)∗∗	335 (28)∗∗	<0.001	0.021
Antihypertensive	1237 (24)	303 (33)∗	244 (26)∗,∗∗	258 (26)∗,∗∗	227 (22)∗,∗∗	205 (17)∗∗	<0.001	0.002
ACE inhibitors	890 (17)	160 (17)∗	197 (21)∗,∗∗	173 (17)∗	215 (21)∗	145 (12)∗∗	<0.001	0.035
Anticoagulant	1422 (28)	243 (26)	280 (30)∗	258 (26)	332 (32)∗,∗∗	309 (26)	0.002	0.068
Stations	515 (10)	64 (7)∗	96 (10)∗∗	111 (11)∗∗	113 (11)∗∗	131 (11)∗∗	0.012	0.009
Diuretics	217 (4)	34 (4)	54 (6)∗,∗∗	38 (4)	56 (5)∗	35 (3)	0.004	0.031
Risk factors [n (%)]
Hypertension	2810 (57)	566 (61)∗	552 (59)∗	571 (58)∗	577 (57)∗,∗∗	546 (45)∗∗	<0.001	0.403
Dyslipidemia	2170 (43)	328 (36)∗	382 (41)∗∗	446 (45)∗∗	477 (46)∗∗	539 (45)∗∗	<0.001	0.013
Obesity	1555 (31)	278 (31)∗	313 (34)∗	346 (36)∗,∗∗	358 (36)∗,∗∗	260 (22)∗∗	<0.001	0.112
Diabetes mellitus	759 (15)	162 (18)∗	164 (18)∗	177 (18)∗	141 (14)∗,∗∗	115 (10)∗∗	<0.001	0.954
Smoking history	2991 (59)	525 (57)	586 (63)∗,∗∗	576 (58)	644 (62)∗,∗∗	662 (55)	<0.001	0.009
Medical history [n (%)]
Typical angina	1425 (28)	326 (35)∗	264 (28)∗∗	260 (26)∗∗	262 (25)∗∗	313 (26)∗∗	<0.001	0.001
Claudication	425 (8)	136 (15)∗	83 (9)∗,∗∗	93 (9)∗,∗∗	69 (7)∗,∗∗	44 (4)∗∗	<0.001	<0.001
Chronic heart failure	702 (14)	232 (25)∗	128 (14)∗,∗∗	130 (13)∗,∗∗	116 (11)∗,∗∗	96 (8)∗∗	<0.001	<0.001
Stroke	301 (6)	78 (9)∗	54 (6)∗∗	72 (7)∗	47 (5)∗∗	50 (4)∗∗	<0.001	0.027
Myocardial infarction	1420 (28)	281 (30)	226 (24)∗∗	308 (31)	265 (25)∗∗	340 (28)	0.002	0.003
Atrial fibrillation	362 (7)	100 (11)∗	61 (7)∗∗	76 (8)∗,∗∗	65 (6)∗∗	60 (5)∗∗	<0.001	0.001
COPD	477 (9)	119 (13)∗	126 (14)∗	81 (8)∗,∗∗	81 (8)∗∗	70 (6)∗∗	<0.001	0.670
Procedures [n (%)]
CABG	773 (15)	166 (18)∗	146 (16)∗	160 (16)∗	149 (14)∗∗	152 (13)∗∗	0.008	0.192
PCA	500 (10)	70 (8)∗	73 (8)∗	105 (11)∗∗	101 (10)∗	151 (13)∗∗	<0.001	0.825

ACE, angiotensin-converting enzyme; CABG, coronary artery bypass surgery; COPD, chronic obstructive pulmonary disease; CV, cardiovascular; PCA, percutaneous coronary angiogram. ∗P<0.05 versus fifth quintile. ∗∗P<0.05 versus first quintile.

Fig. 1

Hazard ratios for all-cause and cardiovascular (CV) mortality across quintiles of fitness. Annual mortality rates are at the top of each bar. CV, cardiovascular; METs, metabolic equivalents; Q, quintile; Q1, the least-fit quintile; Q5, the most-fit quintile.

Discussion

The overall findings of this study are consistent with previous reports showing a nonlinear gradient between fitness and both all-cause and cardiovascular mortality [1, 3–6]. Unique findings include the fact that greater severity of disease and higher presence of comorbidities likely contribute to, but cannot fully explain, the striking mortality difference between the least-fit and the next-least-fit quintiles of fitness in patients with CVD. Importantly, we found no significant differences in recent (last year) and adulthood (lifetime) recreational PA between Q1 and Q2.

Patients in Q1 were older and had between 3 and 9% higher prevalence rates of exercise-induced angina, myocardial infarction, chronic heart failure, stroke, and claudication compared with Q2. Although significant, these differences in clinical characteristics seem to be too small to explain the nearly two-fold increase in the risk of mortality between Q1 and Q2. Recreational PA patterns showed a significant main effect for increased recent PA and a borderline effect for increased lifetime PA from the lowest to the highest quintile of fitness. Average energy expenditures in weekly recreational activities were lower by 20% or more in Q1 compared with Q2, but these differences did not reach statistical significance. These findings are different from our previous report in apparently healthy US veterans in which we observed significantly lower recent (last year) PA in Q1 versus Q2 [7].

Table 2

Resting, exercise, and hemodynamic data for quintiles of fitness

		Quintiles of exercise capacity (estimated METs)
	Total (N = 5101)	Q1 (N = 923)	Q2 (N = 929)	Q3 (N = 993)	Q4 (N = 1045)	Q5 (N = 1211)	P value for trend	P value first versus second
Resting values
HR (bpm)	76 ± 14	77 ± 15∗	76 ± 14∗	76 ± 15∗	76 ± 14∗	74 ± 14∗∗	<0.001	0.332
Systolic BP (mmHg)	133 ± 21	135 ± 23∗	133 ± 21∗	135 ± 21∗	133 ± 19∗	130 ± 19∗∗	<0.001	0.097
Peak exercise values
HR (bpm)	132 ± 24	116 ± 22∗	123 ± 21∗,∗∗	129 ± 21∗,∗∗	138 ± 21∗,∗∗	147 ± 21∗∗	<0.001	<0.001
Systolic BP (mmHg)	173 ± 29	156 ± 31∗	167 ± 28∗,∗∗	175 ± 28∗,∗∗	177 ± 27∗,∗∗	181 ± 26∗∗	<0.001	<0.001
Exercise capacity (METs)	7.4 ± 3.2	3.3 ± 0.6∗	5.0 ± 0.4∗,∗∗	6.6 ± 0.5∗,∗∗	8.5 ± 0.6∗,∗∗	11.9 ± 2.0∗∗	<0.001	<0.001
Age-predicted METs (%)	84 ± 34	43 ± 11∗	61 ± 12∗,∗∗	78 ± 14∗,∗∗	96 ± 17∗,∗∗	127 ± 26∗∗	<0.001	<0.001
Age-predicted HR (%)	94 ± 16	86 ± 17∗	89 ± 16∗,∗∗	93 ± 16∗,∗∗	97 ± 15∗,∗∗	101 ± 14∗∗	<0.001	<0.001
Recovery
Two-min HR recovery (bpm)	41 ± 16	29 ± 17∗	36 ± 16∗,∗∗	41 ± 15∗,∗∗	44 ± 15∗,∗∗	49 ± 14∗∗	<0.001	<0.001
Abnormal exercise response
Angina	793 (16)	184 (20)∗	163 (18)∗	171 (17)∗	160 (15)∗,∗∗	115 (10)∗∗	<0.001	0.188
Abnormal ST-segment depression	1228 (24)	213 (23)	253 (27)∗,∗∗	253 (26)∗	267 (26)∗	242 (20)	<0.001	0.039

BP, blood pressure; HR, heart rate; METs, metabolic equivalents. ∗P<0.05 versus fifth quintile. ∗∗P<0.05 versus first quintile.

Fig. 2

Age-adjusted weekly energy expenditure in recent (a) and lifetime (b) recreational physical activities for quintiles of fitness in patients with cardiovascular disease. METs, metabolic equivalents; Q1, the least-fit quintile; Q5, the most-fit quintile. ∗P<0.05 versus Q5; †P<0.05 versus Q1.

Nearly 60% of our patients in both Q1 and Q2 did not meet current minimal activity guidelines (energy expenditure ≥ 1000 kcal/week) during the previous year. An additional 28% of patients in Q1 and 18% in Q2 were insufficiently active to reach the benchmark of ≥ 2000 kcal/week. Although subtle differences in PA patterns between Q1 and Q2 may, in part, have contributed to the differences in fitness in CVD patients, our results support the concept that disease itself rather than PA patterns may be the principal cause of low fitness in clinical populations [21], particularly among inactive patients. However, the extent to which the onset of chronic disease, its severity, and duration of therapy influenced fitness levels in Q1 versus Q2 remains unknown.

It is important to emphasize that our findings do not diminish the value of regular PA in CVD patients, but rather highlight the importance of engaging patients in exercise programs of sufficient volume and intensity to improve fitness in this population. Consistent with recent studies [1–4, 16, 22–24], exercise capacity was a strong independent predictor of all-cause and cardiovascular mortality after adjusting for age, medications, cardiovascular risk factors, and medical history. Every 1-MET increase in exercise capacity in the lowest end of the fitness spectrum (Q1 and Q2) was associated with a 22% reduction in risk of all-cause mortality and a 31% reduction in risk of cardiovascular mortality. These risk reductions were two times greater than the respective values in the total population. Similar results were previously observed in our laboratory in apparently healthy US veterans [7]. Taken together, these results support the concept that higher fitness is associated with greater survival, irrespective of the presence or absence of CVD. As PA is a primary means of improving fitness, it is important to encourage patients, especially those who are poorly fit with CVD, to engage in a sufficient amount of PA to improve fitness, attenuate disease progression, and ultimately reduce mortality risk.

Table 3

Multivariate predictors of all-cause and cardiovascular mortality

	All-cause mortality		Cardiovascular mortality
Variables	HR (95% CI)	P value	HR (95% CI)	P value
Age	1.04 (1.03-1.04)	<0.001	1.02 (1.01-1.03)	0.003
METs	0.89 (0.88-0.91)	<0.001	0.86 (0.83-0.89)	<0.001
Chronic heart failure	1.66 (1.44-1.90)	<0.001	2.19 (1.77-2.70)	<0.001
Myocardial infarction	1.21 (1.08-1.36)	0.001	1.43 (1.18-1.73)	<0.001
CABG	1.35 (1.19-1.55)	<0.001	1.79 (1.45-2.21)	<0.001
COPD			1.72 (1.03-2.94)	0.040
Exercise-induced angina	1.26 (1.08-1.48)	0.004
Exercise-induced ST-depression	1.23 (1.05-1.44)	0.009	1.82 (1.33-2.50)	<0.001
Obesity	1.34 (1.18-1.53)	<0.001	1.56 (1.24-1.95)	<0.001
Smoking - history			1.30 (1.07-1.58)	0.009
Smoking - current	1.36 (1.15-1.60)	<0.001
Calcium channel blockers			1.28 (1.06-1.55)	0.010
Nitrates	1.27 (1.14-1.43)	<0.001	1.49 (1.23-1.81)	<0.001
Antihypertensive medications	1.39 (1.24-1.56)	<0.001	1.67 (1.38-2.03)	<0.001
Anticoagulants	1.38 (1.16-1.64)	<0.001	3.57 (2.41-5.44)	<0.001
Statins	2.87 (1.52-5.41)	0.001

CABG, coronary artery bypass surgery; CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; METs, metabolic equivalents.

Clinical implications

Our results extend the public health message emphasizing the importance of improving fitness, particularly among poorly fit CVD patients. Health professionals as a practice should encourage deconditioned individuals to adopt exercise programs of moderate intensity and other health-promoting behaviors to improve fitness and reduce mortality risk.

Limitations and future directions

This study has several limitations. First, we used a single measurement of exercise capacity and PA at baseline. It is unknown whether changes in fitness because of changes in PA or other health habits influenced the results. Second, PA data were collected in more recent years and were available only in a subset of patients. This subset of patients was older and had a less favorable medical profile but similar exercise capacity compared with the entire population. Although more than two-third of patients reported an energy expenditure of less than 2000 kcal/week, the high prevalence of sedentary individuals and high variability in the data likely limited the power to detect differences in PA patterns between Q1 and Q2. Third, there are inherent limitations to self-reporting of PA, including potential measurement errors, recall bias, variation in the level of activity during adulthood, and how attentive patients may have been in their responses. Fourth, earlier studies have shown that physically active individuals (quintiles 4 and 5) tend to adopt other kinds of healthpromoting behavior such as healthier diets [25] that may influence disease severity, comorbidities, or mortality. Fifth, the effects of environmental factors and genetics on an individual's fitness level are difficult to quantify. Finally, the results are based on US veterans and may not necessarily apply to the general population. Future studies need to consider other factors such as healthpromoting behaviors, environmental factors, and genetics to further explore factors that contribute to the differences in mortality rates between Q1 and Q2.

In summary, our results suggest that a greater severity of disease and prevalence of comorbidities in Q1 versus Q2 likely contribute to differences in fitness but cannot fully explain the nearly two-fold increase in mortality risk observed between Q1 and Q2. In contrast, recent (last year) and adulthood (lifetime) recreational PA patterns were not different between Q1 and Q2. Therefore, although increasing daily PA is important in CVD patients, it may not be sufficient to achieve the potential survival benefits in the least-fit CVD patients. Health professionals should encourage poorly fit CVD patients to improve fitness by engaging in exercise programs of sufficient volume and intensity to achieve this goal.

Acknowledgements

No funding sources to be disclosed and the authors have no conflicts of interest.

Author notes

Part of this work was presented as abstracts at the European Congress of Cardiology 2008 (Munich, Germany) and American Heart Association Annual Meeting 2008 (New Orleans, USA).